height = cap.get(cv2.CAP_PROP_FRAME_HEIGHT)
fps = cap.get(cv2.CAP_PROP_FPS)
fourcc = cv2.VideoWriter_fourcc(*'DIVX') # 코덱 정의
out = cv2.VideoWriter('out.avi', fourcc, fps,
(int(width), int(height))) # VideoWriter 객체 정의
while (cap.isOpened()):
ret, frame = cap.read()
top_frame = frame[:470, :, :]
vmid_frame = frame[470:662, :, :]
bot_frame = frame[662:, :, :]
gray = cv2.cvtColor(vmid_frame, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray, 200, 300)
lines = cv2.HoughLines(edges, 1, np.pi / 180, 100) # 마지막게 Threshold
for i in range(len(lines)):
for rho, theta in lines[i]:
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(vmid_frame, (x1, y1), (x2, y2), (0, 0, 255), 2)
result = np.vstack((top_frame, vmid_frame, bot_frame))
import cv2.cv as cv
import time
cap = cv2.VideoCapture(0)
cap.set(3, 640)
cap.set(4, 480)
#http://opencv-code.com/tutorials/automatic-perspective-correction-for-quadrilateral-objects/
while (True):
# Capture frame-by-frame
ret, frame = cap.read()
gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray, 100, 200, 3)
lines = cv2.HoughLines(edges, 1, np.pi / 180, 130,
150) #minLineLength,maxLineGap
if lines == None:
continue
for rho, theta in lines[0]:
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
extend = 1000
x1 = int(x0 + extend * (-b))
y1 = int(y0 + extend * (a))
x2 = int(x0 - extend * (-b))
y2 = int(y0 - extend * (a))
cv2.line(gray, (x1, y1), (x2, y2), (0, 0, 255), 2)
# Img Difference (Assume objects station /slow motion)
resized = resized.astype("int16")
resized0 = resized0.astype("int16")
frame1 = np.absolute(resized - resized0)
frame1 = frame1.astype("uint8")
# Non-maximum Suppression (replace with thresholding)
# frame1[frame1<15] = 0
# cv2.imshow('Frame',frame1)
# Hough transform
edges = cv2.Canny(frame1,150,230, apertureSize = 3)
# cv2.imshow('edges', edges)
lines = cv2.HoughLines(edges,1,np.pi/180,22) # Or try HoughLinesP
if lines is not None:
temp = [0,0,0,0]
for rho,theta in lines[:,0,:]:
a = np.cos(theta)
b = np.sin(theta)
x0 = a*rho
y0 = b*rho
temp[0] += (x0 + 1000*(-b))
temp[1] += (y0 + 1000*(a))
temp[2] += (x0 - 1000*(-b))
temp[3] += (y0 - 1000*(a))
no_lines = len(lines[:,0,:])
x1 = int((temp[0]/no_lines)*8)
y1 = int((temp[1]/no_lines)*8)+10 # adjustment (shift)
x2 = int((temp[2]/no_lines)*8)
def align(filename):
im = cv2.imread(filename)
gray = cv2.cvtColor(im, cv2.COLOR_BGR2GRAY)
gray = cv2.GaussianBlur(gray, (5, 5), 0)
cv2.imshow("gray", R(gray))
edges = cv2.Canny(gray, 220, 300, apertureSize=3)
cv2.imshow("edges", R(edges))
cv2.waitKey(0)
#pic5 200 else 60
lines = cv2.HoughLines(edges, 1, np.pi / 360, 200)
print("lines: ", len(lines))
def merge_lines(lines):
nlines = [line[0] for line in lines]
km = KMeans(4)
km.fit(nlines)
return km.cluster_centers_
lines = merge_lines(lines)
print(lines)
result = im.copy()
h, w, c = result.shape
for line in lines:
# line: xcos(t) + ysin(y) - p = 0
rho, theta = line
_cos = np.cos(theta)
_sin = np.sin(theta)
if _cos != 0:
pt1 = (int(rho / _cos), 0)
pt2 = (int((rho - h * _sin) / _cos), h)
else:
#horizontal
pt1 = (0, rho)
pt2 = (w, rho)
cv2.line(result, pt1, pt2, (0, 0, 255), 3)
# 求交点
def get_intersection(a, b):
#a: (rho1, theta1)
#b: (rho2, theta2)
_cos0 = np.cos(a[1])
_sin0 = np.sin(a[1])
_cos1 = np.cos(b[1])
_sin1 = np.sin(b[1])
if _sin1 != 0:
t = _sin0 / _sin1
x = (a[0] - t * b[0]) / (_cos0 - _cos1 * t)
y = (b[0] - x * _cos1) / _sin1
return np.array([x, y])
x = b[0] / _cos1
y = (a[0] - x * _cos0) / _sin0
return np.array([x, y])
tts = np.vstack([get_intersection(lines[3], lines[j]) for j in range(3)])
rec = np.sum(np.multiply(tts, tts), 1)
ai = np.argmax(rec)
#lines[3] 和 lines[ai] 几乎平行
pts = []
for j in range(3):
if j != ai:
pts.append(tts[j])
pts.extend(
[get_intersection(lines[ai], lines[j]) for j in range(3) if j != ai])
for pt in pts:
xy = (int(pt[0]), int(pt[1]))
cv2.circle(result, xy, 32, (255, 0, 0), 32)
cnts = [0] * 4
for i in range(4):
for j in range(4):
if pts[i][0] > pts[j][0]:
cnts[i] += 1
if pts[i][1] > pts[j][1]:
cnts[i] += 1
lt = np.argmin(cnts)
rb = np.argmax(cnts)
# 计算右上角
maxdx = -np.inf
rt = -1
for i in range(4):
if i != lt and i != rb:
dx = pts[i][0] - pts[lt][0]
if dx > maxdx:
maxdx = dx
rt = i
# 计算左下角
for i in range(4):
if i != lt and i != rb and i != rt:
lb = i
paper_p = np.array([(0, 0), (PAPER_WIDTH, 0), (PAPER_WIDTH, PAPER_HEIGHT),
(0, PAPER_HEIGHT)]).astype(np.float32)
'''
pts = [pts[i].tolist() for i in range(4)]
pts.sort()
[lt, rt, rb, lb] = [1,3,2,0]
sp = np.array([pts[i] for i in [lt,rt,rb,lb]]).astype(np.float32)
#print (sp)
'''
sp = np.array([pts[i] for i in [lt, rt, rb, lb]]).astype(np.float32)
M = cv2.getPerspectiveTransform(sp, paper_p)
paper = cv2.warpPerspective(im, M, (int(PAPER_WIDTH), int(PAPER_HEIGHT)))
result = cv2.resize(result, (1080, 720))
cv2.imshow("source", R(im))
cv2.waitKey(0)
cv2.imshow("lines", R(result))
cv2.waitKey(0)
cv2.imshow("paper", R(paper))
cv2.waitKey(0)
return paper
verticalsize = rows // 30 # 30 is the origin one; choose the resolution
# Create structure element for extracting vertical lines through morphology operations
verticalStructure = cv2.getStructuringElement(cv2.MORPH_RECT, (1, verticalsize))
# Apply morphology operations
vertical = cv2.erode(vertical, verticalStructure)
vertical = cv2.dilate(vertical, verticalStructure)
# print(len(vertical[0]))
# # implt(vertical)
gray = cv2.addWeighted(horizontal,0.5,vertical,0.5,0)
# gray = cv2.addWeighted(horizontal,0.5,vertical,0.5,0)
implt(gray,'gray')
cv2.imwrite('../data/pages/separated_lines.jpg',gray)
horizontal_lines = cv2.HoughLines(horizontal,1,np.pi/180,700) # h800
horizontal_boundary = []
print(len(horizontal_lines))
# lines = cv2.HoughLinesP(edges,1,np.pi/180,100,100,10)
for i in range(len(horizontal_lines)):
for rho,theta in horizontal_lines[i]:
a = np.cos(theta)
b = np.sin(theta)
x0 = a*rho
y0 = b*rho
x1 = int(x0 + 1000*(-b))
y1 = int(y0 + 1000*(a))
x2 = int(x0 - 1000*(-b))
y2 = int(y0 - 1000*(a))
# print(x1,x2,y1,y2)
if (abs(y1-y2) <= 2):
def identifica_cor(frame):
'''
Segmenta o maior objeto cuja cor é parecida com cor_h (HUE da cor, no espaço HSV).
'''
hsv1_M = np.array([ 0, 0, 0], dtype=np.uint8)
hsv2_M= np.array([0, 0, 255], dtype=np.uint8)
# placeholdersmaior
valores_esq = { "a_esq" : [], "b_esq" : [], "rho_esq" : [],"aMed_esq" : 1.0, "bMed_esq" : 1.0, "rhoMed_esq" : 1.0}
valores_dir = { "a_dir" : [], "b_dir" : [], "rho_dir" : [],"aMed_dir" : 1.0, "bMed_dir" : 1.0, "rhoMed_dir" : 1.0}
#aMed_esq = 1
#bMed_esq = 1
#rhoMed_esq = 1
#aMed_dir = 1
#bMed_dir = 1
#rhoMed_dir = 1
min_length = 50 # Melhorar mascara e aumentar min_len
lista_ab = []
#a_esq = []
#b_esq = []
#rho_esq = []
#a_dir = []
#b_dir = []
#rho_dir = []
hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
mask1 = cv2.inRange(hsv, hsv1_M, hsv2_M)
seg = cv2.morphologyEx(mask1,cv2.MORPH_CLOSE,np.ones((1, 1)))
selecao = cv2.bitwise_and(frame, frame, mask=seg)
blur = cv2.GaussianBlur(selecao,(5,5),0)
min_contrast = 50
max_contrast = 250
linhas = cv2.Canny(blur, min_contrast, max_contrast )
bordas_color = cv2.cvtColor(linhas, cv2.COLOR_RGB2BGR)
lines = cv2.HoughLines(linhas, 1, np.pi/180, min_length)
if lines is not None:
for line in lines:
rho, theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
lista_ab.append([a, b, rho])
for abrho in lista_ab:
if -18 < abrho[0] < -0.1 :
valores_esq["a_esq"].append(abrho[0])
valores_esq["b_esq"].append(abrho[1])
valores_esq["rho_esq"].append(abrho[2])
elif 18 > abrho[0] > 0.1 :
valores_dir["a_dir"].append(abrho[0])
valores_dir["b_dir"].append(abrho[1])
valores_dir["rho_dir"].append(abrho[2])
if (len(valores_esq["a_esq"]) & len(valores_esq["b_esq"]) & len(valores_esq["rho_esq"])) != 0:
valores_esq["aMed_esq"] = sum(valores_esq["a_esq"]) / len(valores_esq["a_esq"])
valores_esq["bMed_esq"] = sum(valores_esq["b_esq"]) / len(valores_esq["b_esq"])
valores_esq["rhoMed_esq"] = sum(valores_esq["rho_esq"]) / len(valores_esq["rho_esq"])
if (len(valores_dir["a_dir"]) & len(valores_dir["b_dir"]) & len(valores_dir["rho_dir"])) != 0:
valores_dir["aMed_dir"] = sum(valores_dir["a_dir"]) / len(valores_dir["a_dir"])
valores_dir["bMed_dir"] = sum(valores_dir["b_dir"]) / len(valores_dir["b_dir"])
valores_dir["rhoMed_dir"] = sum(valores_dir["rho_dir"]) / len(valores_dir["rho_dir"])
desenhar_reta_media(frame, valores_esq["aMed_esq"], valores_esq["bMed_esq"], valores_esq["rhoMed_esq"])
desenhar_reta_media(frame, valores_dir["aMed_dir"], valores_dir["bMed_dir"], valores_dir["rhoMed_dir"])
x_ponto, y_ponto = interseccao(frame, valores_esq["aMed_esq"], valores_esq["bMed_esq"], valores_esq["rhoMed_esq"], valores_dir["aMed_dir"], valores_dir["bMed_dir"], valores_dir["rhoMed_dir"])
media = (x_ponto,y_ponto)
font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(mask1,'Press q to quit',(0,50), font, 1,(255,255,255),2,cv2.LINE_AA)
cv2.imshow ('Frame', frame)
centro = (frame.shape[1]//2, frame.shape[0]//2)
return media, centro
dst = cv2.Canny(src, 50, 200) # aplica o detector de bordas de Canny à imagem src
cdst = cv2.cvtColor(dst, cv2.COLOR_GRAY2BGR) # Converte a imagem para BGR para permitir desenho colorido
if True: # HoughLinesP
lines = cv2.HoughLinesP(dst, 10, math.pi/180.0, 100, np.array([]), 5, 5)
print("Used Probabilistic Rough Transform")
print("The probabilistic hough transform returns the end points of the detected lines")
a,b,c = lines.shape
print("Valor de A",a, "valor de lines.shape", lines.shape)
for i in range(a):
# Faz uma linha ligando o ponto inicial ao ponto final, com a cor vermelha (BGR)
cv2.line(cdst, (lines[i][0][0], lines[i][0][1]), (lines[i][0][2], lines[i][0][3]), (0, 0, 255), 3, cv2.LINE_AA)
else: # HoughLines
# Esperemos nao cair neste caso
lines = cv2.HoughLines(dst, 1, math.pi/180.0, 50, np.array([]), 0, 0)
a,b,c = lines.shape
for i in range(a):
rho = lines[i][0][0]
theta = lines[i][0][1]
a = math.cos(theta)
b = math.sin(theta)
x0, y0 = a*rho, b*rho
pt1 = ( int(x0+1000*(-b)), int(y0+1000*(a)) )
pt2 = ( int(x0-1000*(-b)), int(y0-1000*(a)) )
cv2.line(cdst, pt1, pt2, (0, 0, 255), 3, cv2.LINE_AA)
print("Used old vanilla Hough transform")
print("Returned points will be radius and angles")
cv2.imshow("source", src)
cv2.imshow("detected lines", cdst)
def line_detection_low(image):
gray = cv.cvtColor(image, cv.COLOR_RGB2GRAY)
edges = cv.Canny(gray, 50, 310) # apertureSize参数默认其实就是3 # 50 310
# cv.imshow("edges", edges)
edge = Image.fromarray(edges)
edge.save("edge.jpeg")
lines = cv.HoughLines(edges, 1, np.pi / 180, 30) # 68
# l1 = lines[:, 0, :]
# print(l1)
mink = float('inf')
maxk = -float('inf')
for line in lines:
rho, theta = line[0] # line[0]存储的是点到直线的极径和极角，其中极角是弧度表示的。
a = np.cos(theta) # theta是弧度
b = np.sin(theta)
x0 = a * rho # 代表x = r * cos（theta）
y0 = b * rho # 代表y = r * sin（theta）
x1 = int(x0 + 1000 * (-b)) # 计算直线起点横坐标
y1 = int(y0 + 1000 * a) # 计算起始起点纵坐标
x2 = int(x0 - 1000 * (-b)) # 计算直线终点横坐标
y2 = int(
y0 - 1000 * a
) # 计算直线终点纵坐标 注：这里的数值1000给出了画出的线段长度范围大小，数值越小，画出的线段越短，数值越大，画出的线段越长
print("x1: %s, y1:%s, x2:%s, y2:%s" % (x1, y1, x2, y2))
k = (y2 - y1) / (x2 - x1)
if k > maxk:
maxk = k
xmax1 = x1
ymax1 = y1
xmax2 = x2
ymax2 = y2
lineMax = line
if k < mink:
mink = k
xmin1 = x1
ymin1 = y1
xmin2 = x2
ymin2 = y2
lineMin = line
cv.line(image, (xmax1, ymax1), (xmax2, ymax2), (255, 0, 0),
2) # 点的坐标必须是元组，不能是列表。
cv.line(image, (xmin1, ymin1), (xmin2, ymin2), (255, 0, 0),
2) # 点的坐标必须是元组，不能是列表。
crossX = int((maxk * xmax1 - ymax1 - mink * xmin1 + ymin1) / (maxk - mink))
crossY = int(
(maxk * mink *
(xmax1 - xmin1) + maxk * ymin1 - mink * ymax1) / (maxk - mink))
print(crossX, 250 - crossY)
height = 250 - crossY
print("顶点高度：" + str(int(height)))
x1 = (-height) / mink + crossX
x2 = (-height) / maxk + crossX
print("与x轴交点：%f,%f" % (x1, x2))
# 底边长度
xl = abs(x1 - x2)
cv.circle(image, (crossX, crossY), 3, (0, 255, 0), -1) # 两直线交点
cv.circle(image, (xmax2, ymax2), 3, (0, 0, 255), -1)
cv.circle(image, (xmin1, ymin1), 3, (0, 0, 255), -1)
vector1 = np.array([xmax2 - crossX, ymax2 - crossY])
vector2 = np.array([xmin1 - crossX, ymin1 - crossY])
L1 = np.sqrt(vector1.dot(vector1))
L2 = np.sqrt(vector2.dot(vector2))
cos_angle = vector1.dot(vector2) / (L1 * L2)
angle = np.arccos(cos_angle)
angle2 = angle * 360 / 2 / np.pi
print(angle2)
# cv.imshow("image-lines", image)
im = Image.fromarray(image)
# im.save(image_line)
cv.waitKey(0)
return angle2, edge, image, height, xl
# Codes inspired by:
# Github.com/imvinod/
# Official Documentation
#==============================================================================
import cv2
import numpy as np
#HOUGHLINES LINE DETECTION
image = cv2.imread('imgs/demo1.jpg')
cv2.imshow('Original', image)
gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray, 100, 170)
rho_accuracy = 1
theta_accuracy = np.pi / 180
threshold = 210
lines = cv2.HoughLines(edges, rho_accuracy, theta_accuracy, threshold)
for line in lines:
rho, theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(image, (x1, y1), (x2, y2), (255, 255, 0), 2)
#cv2.imwrite('houghlines.jpg',image)
cv2.imshow('Hough Lines', image)
def get_hough_lines(self, edges):
kernel = np.ones((11, 11), np.uint8)
edges = cv2.dilate(edges, kernel, iterations=1)
# cv2.imshow('dilated edges in hough', edges)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
# kernel = np.ones((2,2),np.uint8)
# dilated_edges = cv2.dilate(edges,kernel,iterations = 1)
minLength = 80
maxLineGap = 50
lines = cv2.HoughLines(edges, 1, np.pi / 180, 80)
# lines = np.squeeze(lines)
# lines = cv2.HoughLinesP(edges,1,np.pi/120,10, minLength, maxLineGap)
if (lines is not None):
lines = np.squeeze(lines)
print("houghlines shape =", lines.shape)
edges3CH = np.dstack((edges, edges, edges))
new_lines = []
rect_line = np.zeros((1, 2))
for rho, theta in lines:
a = np.cos(np.float(theta))
b = np.sin(np.float(theta))
##### convert line from polar coordinates to cartesian coordinattes
slope = -a / b
intercept = rho / b
# print("intercept, slope = ",intercept,slope)
rect_line = np.vstack((rect_line, np.array([intercept,
slope])))
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(edges3CH, (x1, y1), (x2, y2), (0, 0, 255), 2)
# new_lines.append([rh,th])
# cv2.imshow('all houg_lines', edges3CH)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
print("new_lines = ", new_lines)
np_lines = np.array(new_lines)
status = False
line_clusters = self.get_line_clusters(np_lines, 25, 0.4)
print("lines clusters = ", line_clusters)
if (line_clusters.shape[0] >= 4):
mod_line_cluster = self.augment_lines(line_clusters)
# print ("mod_line_cluster",mod_line_cluster)
rect_line = np.zeros((1, 2))
# print line_clusters
# for rho,theta in mod_line_cluster:
for rho, theta in new_lines:
# rho = rho+1
a = np.cos(np.float(theta))
b = np.sin(np.float(theta))
##### convert line from polar coordinates to cartesian coordinattes
slope = -a / b
intercept = rho / b
# print("intercept, slope = ",intercept,slope)
rect_line = np.vstack(
(rect_line, np.array([intercept, slope])))
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(edges3CH, (x1, y1), (x2, y2), (0, 0, 255), 2)
rect_line = rect_line[1:, :]
# print ("rect_line",rect_line)
# cv2.imshow('clustered hough lines', edges3CH)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
status = True
return status, rect_line
else:
# print('no hough lines')
status = False
return status, lines
else:
status = False
return status, lines
#plt.title(titles[i])
#plt.xticks([]), plt.yticks([])
#plt.show()
##检测线段
# input_img = canny_img
# lines = cv2.HoughLinesP(input_img, 1, np.pi / 180, 100, 100, 10)
# for x1, y1, x2, y2 in lines[0]:
# cv2.line(input_img, (x1, y1), (x2, y2), (0, 255, 0), 100)
# win4 = cv2.namedWindow('xianduan', flags=0)
# cv2.imshow('xianduan', input_img)
###霍夫变换
#最后说明多少个点决定一条直线
input_img = canny_img1
lines = cv2.HoughLines(input_img, 1, np.pi / 10, 40) #这里对最后一个参数使用了经验型的值
lines1 = lines[:, 0, :] #提取为为二维
for rho, theta in lines1[:]:
print(rho, theta)
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
#print(x1,y1,x2,y2)
input_img = cv2.line(input_img, (x1, y1), (x2, y2), (255, 0, 0), 2)
win5 = cv2.namedWindow('HF', flags=0)
import cv2
import numpy as np
""" hough transform is apopular technique to detect any shape,
if you can represent that shape in a mathematical form. it can
detect the shape even if it is broken or distorted a little bit"""
#steps
#Edge detection
#Mapping of edge points to the hough space snd store in an acumulator
# interpretation of the accumulator toyeild lines of infinite lenght
#the interpretation is done by thresholding and other possible constraints
#conversion of infinite lines to finite lines
img = cv2.imread(r'data/sudoku.png')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
edge = cv2.Canny(gray, 50, 150, apertureSize=3)
lines = cv2.HoughLines(edge, 1, np.pi / 180, 200)
for line in lines:
rho, theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(img, (x1, y1), (x2, y2), (0, 0, 255), 2)
print("Horitzontal lines: ", len(horitzontalLines))
for i in range(0, len(verticalLines)):
#print VERTICAL in GREEN
cv2.line(cdstP, (verticalLines[i][0][0], verticalLines[i][0][1]),
(verticalLines[i][0][2], verticalLines[i][0][3]),
(0, 255, 0), 3, cv2.LINE_AA)
print("LINE ", i, ": ")
print("--------------")
print(verticalLines[i])
print("Vertical lines: ", len(verticalLines))
elif method == 2:
lines = cv2.HoughLines(dst, 1, np.pi / 180, 100, None, 0, 0)
#linesP = cv2.HoughLinesP(dst, 1, np.pi / 180, 50, None, 50, 10)
# SOURCE size = 140 lines -- ANotacio utilitzada per veure si anava millorant la detecció de linies
# mathematicalLines es un array que intenta expressar les linies d'una manera més humana
# amb punts d'inici, final, graus, distàncies....
valid_lines = []
for line in lines:
rho, theta = line[0]
if theta > math.pi:
theta = theta - math.pi
if theta > math.pi / 2 + math.pi / 4:
theta = theta - math.pi
assigned = False
def hough(self, picture):
img = cv2.cvtColor(picture, cv2.COLOR_RGB2GRAY)
# uu=rgb2gray(picture)
#cv2.imshow("gray", img)
# gradX = cv2.Sobel(gray, ddepth=cv2.CV_32F, dx = 1, dy = 0, ksize = -1)
# gradY = cv2.Sobel(gray, ddepth = cv2.CV_32F, dx = 0, dy = 1, ksize = -1)
# cv2.imshow("gradX", gradX)
# x=cv2.waitKey(1)& 0xFF
# cv2.imshow("gradY", gradY)
# x=cv2.waitKey(1)
# subtract the y-gradient from the x-gradient
# gradient = cv2.subtract(gradX, gradY)
# img = cv2.convertScaleAbs(gradient)
img = cv2.blur(img, (3, 3))
edges = cv2.Canny(img, 50, 150, apertureSize=3)
# cv2.imshow('cannyuu',edges)
lines = cv2.HoughLines(edges, 1, np.pi / 180,
118) # ÕâÀï¶Ô×îºóÒ»¸ö²ÎÊýʹÓÃÁ˾ÑéÐ͵ÄÖµ
if (lines is not None):
for line in lines[0]:
line_pix = line[0]
else:
line_pix = -1
result = img.copy()
shuipingx = []
if (lines is not None):
for lop in range(int(lines.size / 2)):
for line in lines[lop]:
rho = line[0] # µÚÒ»¸öÔªËØÊǾàÀërho
theta = line[1] # µÚ¶þ¸öÔªËØÊǽǶÈtheta
#print(rho)
#print(theta)
if (theta <
(np.pi / 4.)) or (theta >
(3. * np.pi / 4.0)): # ´¹Ö±Ö±Ïß
# ¸ÃÖ±ÏßÓëµÚÒ»ÐеĽ»µã
pt1 = (int(rho / np.cos(theta)), 0)
# ¸ÃÖ±ÏßÓë×îºóÒ»ÐеĽ¹µã
pt2 = (int((rho - result.shape[0] * np.sin(theta)) /
np.cos(theta)), result.shape[0])
# »æÖÆÒ»Ìõ°×Ïß
cv2.line(result, pt1, pt2, 0, 1)
else: # ˮƽֱÏß
# ¸ÃÖ±ÏßÓëµÚÒ»ÁеĽ»µã
pt1 = (0, int(rho / np.sin(theta)))
# ¸ÃÖ±ÏßÓë×îºóÒ»ÁеĽ»µã
pt2 = (result.shape[1],
int((rho - result.shape[1] * np.cos(theta)) /
np.sin(theta)))
# »æÖÆÒ»ÌõÖ±Ïß
if (pt2[0] - pt1[0] > 0):
cv2.line(result, pt1, pt2, 0, 5)
shuipingx.append(pt1[1])
#print(('pt2=', pt2, 'pt1 =', pt1))
# if shuipingx>0:
# break
nowiq = datetime.datetime.now()
if not os.path.isdir('hough'):
os.makedirs('hough')
print('create dir picture')
cv2.imwrite(
'hough/%s_%s_%s_%s_%s_%s.jpg' %
(nowiq.year, nowiq.month, nowiq.day, nowiq.hour, nowiq.minute,
nowiq.second), result)
return result, shuipingx, line_pix
def detectTurn(self):
### Params for region of interest
bot_left = [0, 480]
bot_right = [640, 480]
apex_right = [640, 170]
apex_left = [0, 170]
v = [np.array([bot_left, bot_right, apex_right, apex_left], dtype=np.int32)]
cropped_raw_image = self.region_of_interest(cf.img_rgb_raw, v)
# cropped_raw_image = cf.img_rgb_raw[self.crop_top:self.crop_bottom, :]
### Run canny edge dection and mask region of interest
# gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
hsv = cv2.cvtColor(cropped_raw_image, cv2.COLOR_BGR2HSV)
lower_white = np.array([0,0,255], dtype=np.uint8)
upper_white = np.array([179,255,255], dtype=np.uint8)
mask = cv2.inRange(hsv, lower_white, upper_white)
dilation = cv2.dilate(mask, self.kernel, iterations=1)
closing = cv2.morphologyEx(dilation, cv2.MORPH_GRADIENT, self.kernel)
closing = cv2.morphologyEx(dilation, cv2.MORPH_CLOSE, self.kernel)
blur = cv2.GaussianBlur(closing, (9,9), 0)
edge = cv2.Canny(blur, 150,255)
cropped_image = self.region_of_interest(edge, v)
# cropped_image = edge[self.crop_top:self.crop_bottom, :]
# blank_image = np.zeros(cropped_raw_image.shape)
# turnSignal = False
lines = cv2.HoughLines(cropped_image, rho=0.2, theta=np.pi/80, threshold=70)
if lines is not None:
# print('lines', len(lines))
for line in lines:
for rho,theta in line:
a = np.cos(theta)
b = np.sin(theta)
x0 = a*rho
y0 = b*rho
x1 = int(x0 + 1000*(-b))
y1 = int(y0 + 1000*(a))
x2 = int(x0 - 1000*(-b))
y2 = int(y0 - 1000*(a))
cv2.line(cropped_raw_image, (x1,y1), (x2,y2), cf.listColor[0], 2)
# cv2.line(blank_image, (x1,y1), (x2,y2), cf.listColor[0], 2)
if abs(y1-y2) < 40:
# turnSignal = True
# break
return True
# cv2.imshow('hsv', hsv)
# cv2.imshow('closing', closing)
# cv2.imshow('cropped_image', cropped_image)
# cv2.imshow('cropped_raw_image', cropped_raw_image)
# cv2.imshow('blank_image', blank_image)
return False
gray = cv2.cvtColor(output, cv2.COLOR_BGR2GRAY)
low_threshold, high_threshold = 100, 500
edges = cv2.Canny(gray, low_threshold, high_threshold)
cv2.imshow('edges', gray)
cv2.waitKey(0)
rho = 3 # distance resolution in pixels of the Hough grid
theta = np.pi / 180 # angular resolution in radians of the Hough grid
threshold = 200 # minimum number of votes (intersections in Hough grid cell)
# min_line_length = 10 # minimum number of pixels making up a line
# max_line_gap = 5 # maximum gap in pixels between connectable line segments
line_image = np.copy(img) * 0 # creating a blank to draw lines on
lines = cv2.HoughLines(edges, rho, theta, threshold)
print(len(lines))
for line in lines:
rho, theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(img, (x1, y1), (x2, y2), (0, 255, 0), 1)
def main(argv):
## [load]
default_file = 'paashaas.jpg'
maxlen = 7
maxhoek = np.pi / 2 # in radialen
offset_y = 200 + 200
offset_x = 180
filename = argv[0] if len(argv) > 0 else default_file
# Loads an image
src = cv.imread(cv.samples.findFile(filename), 0) # cv.IMREAD_GRAYSCALE)
# Check if image is loaded fine
if src is None:
print('Error opening image!')
print('Usage: hough_lines.py [image_name -- default ' + default_file +
'] \n')
return -1
## [load]
## [edge_detection]
# Edge detection
dst = cv.Canny(src, 50, 200, None, 3)
## [edge_detection]
# Copy edges to the images that will display the results in BGR
cdst = cv.cvtColor(dst, cv.COLOR_GRAY2BGR)
cdstP = np.copy(cdst)
## [hough_lines]
# Standard Hough Line Transform
lines = cv.HoughLines(dst, 1, np.pi / 180, 150, None, 0, 0)
## [hough_lines]
## [draw_lines]
# Draw the lines
if lines is not None:
for i in range(0, len(lines)):
rho = lines[i][0][0]
theta = lines[i][0][1]
a = math.cos(theta)
b = math.sin(theta)
x0 = a * rho
y0 = b * rho
pt1 = (int(x0 + 1000 * (-b)), int(y0 + 1000 * (a)))
pt2 = (int(x0 - 1000 * (-b)), int(y0 - 1000 * (a)))
cv.line(cdst, pt1, pt2, (0, 0, 255), 3, cv.LINE_AA)
## [draw_lines]
## [hough_lines_p]
# Probabilistic Line Transform
linesP = cv.HoughLinesP(dst, 1, np.pi / 2, 1, None, 0, 0)
## [hough_lines_p]
## [draw_lines_p]
# Draw the lines
coorda = [] # array van coordinaten
if linesP is not None:
File1 = open("tabel2.txt", "w")
for i in range(0, len(linesP)):
l = linesP[i][0]
cv.line(cdstP, (l[0], l[1]), (l[2], l[3]), (0, 0, 255), 3,
cv.LINE_AA)
coorda.append([[l[0] + offset_x, -l[1] + offset_y],
[l[2] + offset_x, -l[3] + offset_y]])
File1.write("[" + str(l[0]) + " " + str(l[1]) + "]" + "\n")
File1.close()
coorda = sorted(coorda) # array of coordinates
groepnr = 0
coordgroepen = [coorda[0]] # array of coordinates
while len(coorda) > 0:
afstanden = []
weg = []
for i in coorda:
if len(coordgroepen[groepnr]) >= 3:
afst = afstand(coordgroepen[groepnr][-3],
coordgroepen[groepnr][-1], i[1], maxlen,
maxhoek)
else:
afst = afstand(coordgroepen[groepnr][1],
coordgroepen[groepnr][-1], i[1], maxlen,
maxhoek)
if afst < maxlen:
afstanden.append([afst, i])
if len(afstanden) == 0:
if len(coorda) > 0:
coordgroepen.extend([coorda[0]])
groepnr += 1
else:
afstanden.sort()
coordgroepen[groepnr].extend(afstanden[0][1])
coorda.remove(afstanden[0][1])
print(groepnr)
plt.axis('off')
plt.axis('equal')
for i in coordgroepen:
xs = [x[0] for x in i]
ys = [x[1] for x in i]
plt.plot(xs, ys)
# plt.savefig('dobot\\dobot_min' + minlen.__str__() + '_max' + maxlen.__str__() + '_' + default_file)
plt.show()
## [draw_lines_p]
## [imshow]
# Show results
# cv.imshow("Source", src)
# cv.imshow("Detected Lines (in red) - Standard Hough Line Transform", cdst)
cv.imshow("Detected Lines (in red) - Probabilistic Line Transform", cdstP)
## [imshow]
## [exit]
# Wait and Exit
cv.waitKey()
return 0
def find_road_boundary(img):
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
img_blur = cv2.GaussianBlur(img_gray, (3, 3), 0)
contour = cv2.Canny(img_blur, 20, 60)
lines = cv2.HoughLines(contour, 1, np.pi / 180, 100, 100, 5)
lines_direction = []
lines_intercept = []
lines_dataset2 = []
lines1 = lines[:, 0, :]
#for line in lines1:
# [x1,y1,x2,y2] = line
#print(x1)
#print(x2)
#print(y1)
#print(y2)
# if((x1 in range(0,480)) and (x2 in range(0,480)) and (y1 in range(0,360)) and (y2 in range(0,360))):
# dx,dy = x2-x1,y2-y1
# angle = np.arctan2(dy,dx) * (180/np.pi)
# if (abs(angle)>10 and abs(angle) < 75):
# print("%d %d %d %d"% (x1,y1,x2,y2))
# direction = dy/(dx+0.000001)
# intercept = y1 - direction*x1
# if(abs(direction)>=40):
# direction = 50
# intercept = 100000
# print(direction)
# print(intercept)
# lines_direction.append(direction)
# lines_intercept.append(intercept)
# cv2.line(img,(x1,y1),(x2,y2),(0,255,0),2)
# lines_dataset2.append(line)
# plt.figure()
# plt.scatter(lines_direction,lines_intercept)
# plt.show()
# lines_dataset = np.array([[lines_direction[i],lines_intercept[i]] for i in range(len(lines_direction))])
# print(lines_dataset)
#K = range(1,4)
#for k in K:
# k = 2
# kmeans = KMeans(n_clusters=k)
# kmeans.fit(lines_dataset)
# print(kmeans.cluster_centers_)
# boundaries = []
# for i in range (kmeans.cluster_centers_.shape[0]):
# boundary = []
# for j in range(360):
# x = int((j - kmeans.cluster_centers_[i][1])/kmeans.cluster_centers_[i][0])
#cv2.circle (img,(x,j),2,(0,0,255),2)
#meandistortion = sum
lines_dataset = []
for rho, theta in (lines1[:]):
print(theta * 180 / np.pi)
if ((theta * 180 / np.pi < 80) or (theta * 180 / np.pi > 105)):
lines_dataset.append([rho, theta])
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
#if()
cv2.line(img, (x1, y1), (x2, y2), (0, 0, 255), 1)
#cv2.imshow("line_detector",img)
#cv2.imshow('Canny',contour)
#cv2.waitKey(50000)
lines_dataset = np.array(lines_dataset)
k = 3
kmeans = KMeans(n_clusters=k)
kmeans.fit(lines_dataset)
print(kmeans.cluster_centers_)
boundaries = []
for i in range(kmeans.cluster_centers_.shape[0]):
boundary = []
for j in range(360):
x = int((j - kmeans.cluster_centers_[i][1]) /
kmeans.cluster_centers_[i][0])
cv2.circle(img, (x, j), 2, (0, 255, 0), 2)
cv2.imshow("line_detector", img)
#cv2.imshow('Canny',contour)
cv2.waitKey(50000)
clip_img = new_img[left_top_w:right_bottle_w,
left_top_l:right_bottle_l, :] #图片切割的结果
#cv.imwrite('pic/clip.jpg', clip_img)
#cv.imshow('clip_img', clip_img)
gray_img = cv.cvtColor(clip_img, cv.COLOR_BGR2GRAY) #灰度化
edges_img = cv.Canny(gray_img, 50, 150, apertureSize=3) #边缘检测
ret, binary = cv.threshold(gray_img, 0, 255,
cv.THRESH_BINARY | cv.THRESH_TRIANGLE)
#cv.imshow('gray_img', gray_img)
#cv.imshow('edges', edges_img)
#cv.imshow('thresh', binary)
lines = cv.HoughLines(edges_img, 1, np.pi / 180, 150) #houghline直线检测
#print(lines.shape)
ls = [] #保存直线上的点
for line in lines:
rho, theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.imshow('original', img)
b_channel = np.array(img[:,:,0]).astype('float')
g_channel = np.array(img[:,:,1]).astype('float')
r_channel = np.array(img[:,:,2]).astype('float')
#cv2.imshow('b_chan', b_channel)
#cv2.imshow('g_chan', g_channel)
#cv2.imshow('r_chan', r_channel)
bgr_channel = np.add((np.add(b_channel, g_channel)), r_channel)
img_rec_red = np.divide(r_channel, bgr_channel)
img_rec_red = img_rec_red * 255
img_rec_red = np.floor(img_rec_red).astype('uint8')
#gray = cv2.cvtColor(img_rec_red,cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(img_rec_red,50,150,apertureSize = 3)
lines = cv2.HoughLines(edges,1,np.pi/40,40)
print("raw lines:")
print(lines)
# convert the grayscale image to binary image
ret,thresh = cv2.threshold(img_rec_red,127,255,0)
cv2.imshow('thresh', thresh)
# calculate moments of binary image
im2, contours, hierarchy = cv2.findContours(thresh,cv2.RETR_TREE,cv2.CHAIN_APPROX_SIMPLE)
try:
for c in contours:
# calculate moments for each contour
M = cv2.moments(c)
cX = int(M["m10"] / M["m00"])
cY = int(M["m01"] / M["m00"])
imlow = cv2.cvtColor(imlow1, cv2.COLOR_BGR2GRAY)
imlow = cv2.normalize(imlow.astype('float'), None, 0.0, 1.0,
cv2.NORM_MINMAX)
## laplacian = cv2.Laplacian(imlow,cv2.CV_64F)
## laplacian1 = np.absolute(laplacian)
## laplacian2 = np.uint8(laplacian1)
kernel = np.array([[-1, 0, 1]])
dst = cv2.filter2D(imlow, -1, kernel)
dst[dst < 0] = 0 # Where values are low
dst[dst > 1] = 1 # Where values are high
dst = dst * 255
dst = np.uint8(dst)
ret, th3 = cv2.threshold(dst, 20, 255, cv2.THRESH_BINARY)
lines = cv2.HoughLines(th3, 1, np.pi / 180, 1)
lines = np.double(lines)
condition11 = 1
condition12 = 1
if lines != []:
for rho, theta in lines[:, 0, :]:
if (theta < 80 * npi) & (theta > 0 * npi) & (condition11 == 1):
thetaL = theta
rhoL = rho
condition11 = 0
if (theta > 120 * npi) & (theta < 180 * npi) & (condition12 == 1):
thetaR = theta
rhoR = rho
condition12 = 0
def hough_line(edges, min_line_length=100, max_line_gap=10):
lines = cv2.HoughLines(edges, 1, np.pi / 180, 125, min_line_length,
max_line_gap)
lines = np.reshape(lines, (-1, 2))
return lines
def main():
#for eachArg in sys.argv:
# print eachArg
filename = sys.argv[1]
image = cv2.imread(filename)
if image is None:
print 'Unable to open file ', filename
return
rows=image.shape[0]
cols=image.shape[1]
gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
edges = cv2.Canny(gray,150,250,apertureSize = 3)
# Arguments are distance resolution, angle resolution, threshold
# Large distance resolution yields larger bins so more lines meeting
# threshold. Larger angular resolution yeilds fewer lines with similar
# lines counting as the same line
lines = cv2.HoughLines(edges,2,2*np.pi/180,100)
if lines is None:
print 'No lines found'
return
for line in lines:
for rho, theta in line:
if theta != 0: # ignore verticals
a = np.cos(theta)
b = np.sin(theta)
x0 = a*rho
y0 = b*rho
seglength = 1000
x1 = int(x0 + seglength*(-b))
y1 = int(y0 + seglength*(a))
x2 = int(x0 - seglength*(-b))
y2 = int(y0 - seglength*(a))
cv2.line(image,(x1,y1),(x2,y2),(0,0,255),2)
xtotal=0
xcount=0
ytotal=0
ycount=0
for line1,line2 in combinations(lines,2):
rho1=line1[0][0]
theta1=line1[0][1]
rho2=line2[0][0]
theta2=line2[0][1]
#print rho1, theta1, rho2, theta2
if theta1 != 0 and theta2 != 0: # ignore verticals
x0, y0 = intersection(rho1,theta1,rho2,theta2)
xtotal+=x0
xcount+=1
ytotal+=y0
ycount+=1
cv2.circle(image, (x0,y0),3,(255,0,0),-1)
cv2.circle(image, (xtotal/xcount,ytotal/ycount),50,(0,255,255),3)
cv2.namedWindow('Hall with Line', cv2.WINDOW_NORMAL)
cv2.imshow('Hall with Line',image)
cv2.waitKey(0)
cv2.destroyAllWindows()
def enderezar(entrada, salida):
# Leer la imagen
imagen = cv2.imread(entrada)
# Convertirla a gris y detectar bordes
gray = cv2.cvtColor(imagen, cv2.COLOR_BGR2GRAY)
binaria = cv2.Canny(gray,50,150,apertureSize = 3)
#cv2.imshow('Grayscale', imagen)#, binaria)
# Usar la transformada de Hough para encontrar líneas
# en la imagen binarizada, con una resolución de medio
# grado (pi/720) y quedándose sólo con las líneas que
# alcancen puntuación de 1000 o más (que serán las
# más largas)
lineas = cv2.HoughLines(binaria, 1, np.pi/720, 100)
# Recopilemos qué ángulos ha encontrado la transformada
# de hough para cada una de las líneas halladas
angulos = []
try:
for linea in lineas:
rho, theta = linea[0]
if rho<0:
theta = -theta
# Quedarse solo con las rayas próximas a la horizontal
# (con un error de +-10 grados)
if not estan_cercanos(theta, np.pi/2, np.deg2rad(20)):
continue;
angulos.append(theta)
from collections import Counter
veces = Counter(angulos)
# Quedémonos con los tres casos más frecuentes
frecuentes = veces.most_common(3)
# Y calculemos el promedio de esos tres casos
suma = sum(angulo*repeticion for angulo,repeticion in frecuentes)
repeticiones = sum(repeticion for angulo, repeticion in frecuentes)
angulo = suma/repeticiones
angulo = np.rad2deg(angulo - np.pi/2)
print("[INFO] angulo: {:.5f}".format(angulo))
W = 1200.
height, width, depth = imagen.shape
imgScale = W/width
newX,newY = imagen.shape[1]*imgScale, imagen.shape[0]*imgScale
# Ahora enderecemos la imagen, girando el ángulo detectado
(h, w) = imagen.shape[:2]
centro = (w // 2, h // 2)
M = cv2.getRotationMatrix2D(centro, angulo, 1.0)
girada = cv2.warpAffine(imagen, M, (w, h),
flags=cv2.INTER_CUBIC, borderMode=cv2.BORDER_REPLICATE)
girada = cv2.resize(girada,(int(newX),int(newY)), interpolation = cv2.INTER_AREA)
# Y volcamos a disco el resultado
cv2.imwrite(salida, girada)
except:
W = 1200.
height, width, depth = imagen.shape
imgScale = W/width
newX,newY = imagen.shape[1]*imgScale, imagen.shape[0]*imgScale
print("Solo resize")
girada = cv2.resize(imagen,(int(newX),int(newY)), interpolation = cv2.INTER_AREA)
cv2.imwrite(salida, girada)
import cv2
import numpy as np
img = cv2.imread('test.jpg')
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
cv2.imshow('grey', gray)
cv2.waitKey(0)
edges = cv2.Canny(gray, 50, 150, apertureSize=3)
cv2.imshow('edges', edges)
cv2.waitKey(0)
lines = cv2.HoughLines(edges, 1, np.pi / 180, 120)
print len(lines)
print len(lines[0])
print len(lines[0][0])
for i in lines:
rho = i[0][0]
theta = i[0][1]
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(img, (x1, y1), (x2, y2), (0, 0, 255), 2)
cv2.imshow('lines', img)
cv2.waitKey(0)
cv2.imwrite('houghlines3.jpg', img)
def applyHough(image, original):
img = cv2.imread(image)
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
#edges = cv2.Canny(gray,50,150,apertureSize = 3) #23,55
edges = cv2.Canny(gray, 23, 60, apertureSize=3)
lines = cv2.HoughLines(edges, 1, np.pi / 720, 120) #150 #90
if lines is not None:
# number = 0;
# x1_mean = y1_mean = x2_mean = y2_mean = 0;
x1_list = []
x2_list = []
y1_list = []
y2_list = []
for line in lines:
rho, theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
if (y1 <= y2):
x1_list.append(x1)
x2_list.append(x2)
y1_list.append(y1)
y2_list.append(y2)
else:
x1_list.append(x2)
x2_list.append(x1)
y1_list.append(y2)
y2_list.append(y1)
# number = number + 1;
# x1_mean = (x1_mean)+(x1-x1_mean)/number
# y1_mean = (y1_mean)+(y1-y1_mean)/number
# x2_mean = (x2_mean)+(x2-x2_mean)/number
# y2_mean = (y2_mean)+(y2-y2_mean)/number
x1_median = int(np.ma.median(x1_list))
x2_median = int(np.ma.median(x2_list))
y1_median = int(np.ma.median(y1_list))
y2_median = int(np.ma.median(y2_list))
# cv2.line(original,(int(x1_mean),int(y1_mean)),(int(x2_mean),int(y2_mean)),(0,0,255),2)
lines = cv2.HoughLinesP(edges, 0.5, np.pi / 720, 75, 70, 8)
l = 0
xmin = ymin = 100000
xmax = ymax = -100000
if lines is not None:
for line in lines:
x1, y1, x2, y2 = line[0]
if (y1 < ymin): ymin = y1
if (x1 < xmin): xmin = x1
if (x2 > xmax): xmax = x2
if (y2 > ymax): ymax = y2
if (y2 < ymin): ymin = y2
if (x2 < xmin): xmin = x2
if (x1 > xmax): xmax = x1
if (y1 > ymax): ymax = y1
# cv2.line(original,(x1,y1),(x2,y2),(0,0,255),2)
# l = math.sqrt((xmin-xmax)**2 + (ymin-ymax)**2)
# cv2.line(original,(int(x1_mean)+int(x2_mean)-xmin,int(y2_mean)+int(y1_mean)-ymin),(int(xmin),ymin),(0,0,255),2)
# slope_mean = abs((y2_mean - y1_mean)/(x2_mean-x1_mean))
# slope_median = abs((y2_median - y1_median)/(x2_median-x1_median))
# print(slope_mean-slope_median);
# cv2.line(original,(int(x1_mean),int(y1_mean)),(int(x2_mean),int(y2_mean)),(0,0,255),2)
global cutoff, prev_x1, prev_x2, prev_y1, prev_y2, last5_slopes
if (x2_median - x1_median is not 0):
slope = (y2_median - y1_median) / (x2_median - x1_median)
else:
slope = last5_slopes[len(last5_slopes) - 1]
prev_slope = slope
use_this = True
all_neg = True
all_pos = True
for x in last5_slopes:
if (x > 0):
all_neg = False
elif (x < 0):
all_pos = False
if (len(last5_slopes) > 4) and (
(all_pos and slope < 0
and last5_slopes[len(last5_slopes) - 1] < 200) or
(all_neg and slope > 0
and last5_slopes[len(last5_slopes) - 1] > -200)):
use_this = False
else:
last5_slopes.append(slope)
while (len(last5_slopes) > 5):
last5_slopes.pop(0)
# print(str(prev_slope) + " " + str(slope));
# if(prev_slope != 0):
# if(slope - prev_slope > epsilon or slope -prev_slope < epsilon):
# slope = prev_slope;
x_lim_up = (0 - y1_median) / slope + x1_median
if (cutoff == 0):
cutoff = ymax
elif (cutoff > 0 and ymax > 0):
cutoff = 0.9 * cutoff + 0.1 * ymax
# print(cutoff)
x_lim_down = (cutoff - y1_median) / slope + x1_median
# print(str(x1_median) + " " + str(y1_median) + " " + str(x2_median) + " " + str(y2_median))
# cv2.line(original,(int(x1_median),int(y1_median)),(int(x2_median),int(y2_median)),(0,255,255),3)
if use_this:
cv2.line(original, (int(x_lim_up), int(0)),
(int(x_lim_down), int(cutoff)), (0, 0, 255), 3)
else:
cv2.line(original, (int(prev_x1), int(prev_y1)),
(int(prev_x2), int(prev_y2)), (0, 0, 255), 3)
prev_x1 = x_lim_up
prev_y1 = 0
prev_x2 = x_lim_down
prev_y2 = cutoff
# cv2.imshow('asd',img)
elif persist:
cv2.line(original, (int(prev_x1), int(prev_y1)),
(int(prev_x2), int(prev_y2)), (0, 0, 255), 3)
# prev_slope = 0;
return original
#Applying hough lines
minLineLength = 70
maxLineGap = 0.1
lines = cv2.HoughLinesP(total_gradient, 1, np.pi / 180, 5, minLineLength,
maxLineGap)
for x1, y1, x2, y2 in lines[0]:
cv2.line(image_1, (x1, y1), (x2, y2), (0, 255, 0),
thickness=2,
lineType=8,
shift=0)
lines = cv2.HoughLines(total_gradient, 1, np.pi / 180, 200)
for rho, theta in lines[0]:
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(image, (x1, y1), (x2, y2), (0, 0, 255), 1)
#for var in range(-135, 150, 45):
cv2.imshow("Figure: Sobel Edges", total_gradient)
def analyzeGameBoard(image, debug=False):
"""
Determine the current state of the game board. Return a 2d array specifiying the contents of each of the nine
spaces, one of {'X', 'O', ' '}
:param image: Image of the game board on a blank background
:param debug: If true, displays step-by-step visuals for debugging
:return: A 2d array specifiying the contents of each of the nine spaces, one of {'X', 'O', ''}
"""
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
image = cv2.GaussianBlur(image, (5, 5), 0)
binary = cv2.threshold(image, 0, 255,
cv2.THRESH_BINARY + cv2.THRESH_OTSU)[1]
edges = cv2.Canny(binary, 1, 254)
if debug:
cv2.imshow("Image", image)
cv2.imshow("Binary", binary)
cv2.imshow("Edges", edges)
cv2.waitKey(0)
cv2.destroyAllWindows()
# Find contours based on edges
contours = cv2.findContours(edges.copy(), cv2.RETR_LIST,
cv2.CHAIN_APPROX_SIMPLE)
# Normalize format of contours between different versions of OpenCV
contours = imutils.grab_contours(contours)
# Find the contour with the largest area, which should be the game board
board2 = max(contours, key=cv2.contourArea)
#contours.remove(board2)
contours = sorted(contours, key=cv2.contourArea, reverse=True)
contours = contours[1:]
board = max(contours, key=cv2.contourArea)
contours = sorted(contours, key=cv2.contourArea, reverse=True)
contours = contours[1:]
#contours.remove(board)
mask = np.zeros_like(binary)
cv2.drawContours(mask, [board], 0, 255, -1)
out = np.full_like(binary, 255)
out[mask == 255] = binary[mask == 255]
for contour in contours:
mask = np.zeros_like(binary)
cv2.drawContours(mask, [contour], 0, 255, -1)
out[mask == 255] = 255
if debug:
cv2.imshow('t', mask)
cv2.imshow('h', out)
cv2.waitKey(0)
cv2.destroyAllWindows()
if debug:
cv2.imshow('Original', binary)
cv2.imshow('Mask', mask)
cv2.imshow('Output', out)
cv2.waitKey(0)
cv2.destroyAllWindows()
boardEdges = cv2.Canny(out, 1, 254)
lines = cv2.HoughLines(boardEdges, 2, np.pi / 90, 100)
lines = mergeLines(lines)
vLines, hLines = findExtremeLines(lines)
lines = vLines + hLines
if debug:
for line in lines:
for rho, theta in line:
a = np.cos(theta)
b = np.sin(theta)
x0 = a * rho
y0 = b * rho
x1 = int(x0 + 1000 * (-b))
y1 = int(y0 + 1000 * (a))
x2 = int(x0 - 1000 * (-b))
y2 = int(y0 - 1000 * (a))
cv2.line(image, (x1, y1), (x2, y2), (0, 0, 255), 2)
cv2.imshow("i", image)
cv2.waitKey(0)
cv2.destroyAllWindows()
# Remove the game board from the image
binary[out == 0] = 255
if debug:
cv2.imshow('mask', mask)
cv2.imshow('out', out)
cv2.imshow('binary', binary)
cv2.waitKey(0)
cv2.destroyAllWindows()
tlPoint, trPoint, blPoint, brPoint = getAllIntersections(vLines, hLines)
upperMiddle = int((tlPoint[0] + trPoint[0]) / 2)
middleLeft = int((tlPoint[1] + blPoint[1]) / 2)
middleRight = int((trPoint[1] + brPoint[1]) / 2)
lowerMiddle = int((blPoint[0] + brPoint[0]) / 2)
yMax = binary.shape[0] - 1
xMax = binary.shape[1] - 1
spaces = np.empty((3, 3), dtype=object)
if debug:
image[tlPoint[0], tlPoint[1]] = 255
image[trPoint[0], trPoint[1]] = 255
image[blPoint[0], blPoint[1]] = 255
image[brPoint[0], brPoint[1]] = 255
cv2.imshow('h', image)
cv2.waitKey(0)
cv2.destroyAllWindows()
spaces[0][0] = binary[0:tlPoint[0], 0:tlPoint[1]]
spaces[0][1] = binary[0:upperMiddle, tlPoint[1]:trPoint[1]]
spaces[0][2] = binary[0:trPoint[0], trPoint[1]:xMax]
spaces[1][0] = binary[tlPoint[0]:blPoint[0], 0:middleLeft]
spaces[1][1] = binary[upperMiddle:lowerMiddle, middleLeft:middleRight]
spaces[1][2] = binary[trPoint[0]:brPoint[0], middleRight:xMax]
spaces[2][0] = binary[blPoint[0]:yMax, 0:blPoint[1]]
spaces[2][1] = binary[lowerMiddle:yMax, blPoint[1]:brPoint[1]]
spaces[2][2] = binary[brPoint[0]:yMax, brPoint[1]:xMax]
gameState = np.full((3, 3), ' ')
for i in range(3):
for j in range(3):
gameState[i][j] = analyzeSpace(spaces[i][j], debug)
return gameState
def run(self):
if self.active:
#pub.sendMessage("transectline", message = [1, 2, 3, 4, 5, 6])
#pub_to_manager("transectline", message = [1, 2, 3, 4, 5, 6])
#print("active: ",self.active)
#print("show: ", self.show)
if self.captureON == False:
if self.simulation==False:
self.cap = cv2.VideoCapture(0)
self.captureON = True
'''
cv2.namedWindow('Sliders')
cv2.createTrackbar('H','Sliders',0,255,nothing)
cv2.createTrackbar('H_Range','Sliders',0,255,nothing)
cv2.createTrackbar('S','Sliders',0,255,nothing)
cv2.createTrackbar('S_Range','Sliders',0,255,nothing)
cv2.createTrackbar('V','Sliders',0,255,nothing)
cv2.createTrackbar('V_Range','Sliders',0,255,nothing)
'''
ret, frame = self.cap.read()
height = frame.shape[0]
width = frame.shape[1]
origin = (0, 0)
center = (width//2, height//2)
process = np.zeros([height,width,1],dtype=np.uint8)
process.fill(255)
hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
H = cv2.getTrackbarPos('H','Sliders')
H = 102
H_Range = cv2.getTrackbarPos('H_Range','Sliders')
H_Range = 51
S = cv2.getTrackbarPos('S','Sliders')
S = 51
S_Range = cv2.getTrackbarPos('S_Range','Sliders')
S_Range = 128
V = cv2.getTrackbarPos('V','Sliders')
V = 102
V_Range = cv2.getTrackbarPos('V_Range','Sliders')
V_Range = 71
lower_blue = np.array([H,S,V]) #110-130
upper_blue = np.array([SliderLimit(H, H_Range),SliderLimit(S, S_Range),SliderLimit(V, V_Range)])
mask = cv2.inRange(hsv, lower_blue, upper_blue)
res = cv2.bitwise_and(frame,frame, mask= mask)
median = cv2.medianBlur(res,5)
#cv2.imshow('res',res)
grayscaled = cv2.cvtColor(median,cv2.COLOR_BGR2GRAY)
#cv2.imshow('grayscaled',grayscaled)
'''
kernel = np.ones((10,10),np.uint8)
erosion = cv2.erode(grayscaled,kernel,iterations = 1)
cv2.imshow('erosion',erosion)
'''
#retval, threshold = cv2.threshold(grayscaled, 10, 255, cv2.THRESH_BINARY)
#cv2.imshow('gray', grayscaled)
# The bitwise and of the frame and mask is done so
# that only the blue coloured objects are highlighted
# and stored in res
#res = cv2.bitwise_and(frame,frame, mask= mask)
#Filter out all colours except for a range of blue
#gray = cv2.cvtColor(res, cv2.COLOR_BGR2GRAY)
#cv2.imshow('gray',gray)
edges = cv2.Canny(grayscaled,50,150,apertureSize = 3)
#cv2.imshow('edges',edges)
'''
theta = np.pi / 180
rho = 50
threshold = 15 # minimum number of votes (intersections in Hough grid cell)
min_line_length = 50 # minimum number of pixels making up a line
max_line_gap = 20
lines = cv2.HoughLinesP(edges, rho, theta, threshold, np.array([]),min_line_length, max_line_gap)
'''
point1_x, point1_y, point2_x, point2_y = width//2, height, width//2, 0
lines = cv2.HoughLines(edges,10,np.pi/180, 200, 0)
line_amount = 4
coordinates = []
lines_seen = line_amount
#Lines = [line_attr() for i in range(line_amount)]
Lines = []
pos_length = 0
pos_len_count = 1
neg_length = 0
neg_len_count = 1
for i in range(0, line_amount):
try:
for r,theta in lines[i]:
# Stores the value of cos(theta) in a
a = np.cos(theta)
# Stores the value of sin(theta) in b
b = np.sin(theta)
# x0 stores the value rcos(theta)
x0 = a*r
# y0 stores the value rsin(theta)
y0 = b*r
# x1 stores the rounded off value of (rcos(theta)-1000sin(theta))
x1 = int(x0 + 1000*(-b))
# y1 stores the rounded off value of (rsin(theta)+1000cos(theta))
y1 = int(y0 + 1000*(a))
# x2 stores the rounded off value of (rcos(theta)+1000sin(theta))
x2 = int(x0 - 1000*(-b))
# y2 stores the rounded off value of (rsin(theta)-1000cos(theta))
y2 = int(y0 - 1000*(a))
###
#y - y2 = s(x - x2)
#y = sx - sx2 + y2
s = get_gradient((x1,y1), (x2, y2))
t = -x2*s+y2
###
cv2.line(process, (x1, y1), (x2,y2), (0, 0, 255), 1)
m = -1/s
k = -m*center[0] + center[1]
###
#y - y2 = s(x - x2)
#y = sx - sx2 + y2
#y = sx + t
#y - centery = m(x - centerx)
#y = mx - mcenterx + centery
#y = mx + k
#sx + t = mx + k
#x(m-s) = t-k
#x = (t-k)/(m-s)
x_intersect = round((t-k)/(m-s))
y_intersect = round(s*x_intersect + t)
intersect = (x_intersect, y_intersect)
remove = False
for coordinate in coordinates:
if within_range(x_intersect, coordinate[0]-50, coordinate[0]+50) and within_range(y_intersect, coordinate[1]-50, coordinate[1]+50):
line_amount -= 1
lines_seen -= 1
remove = True
continue
if remove == True:
continue
coordinates.append(intersect)
###
cv2.line(process,(center),(intersect),(0,0,255),1)
cv2.circle(process, (intersect), 2, (0, 0, 255), 2)
cv2.circle(process, (center), 2, (0, 0, 255), 2)
if x_intersect < center[0]:
length = -1*get_length((center), (intersect))
neg_length = neg_length + length
neg_len_count +=1
else:
length = get_length((center), (intersect))
pos_length = pos_length + length
pos_len_count += 1
abs_length = abs(length)
angle = np.arctan(self.relative(intersect, center)[1]/self.relative(intersect, center)[0])
#print(angle)
if self.relative(intersect, center)[1]>0 and self.relative(intersect, center)[0] < 0:
angle = np.pi + angle
if self.relative(intersect, center)[1]<0 and self.relative(intersect, center)[0] < 0:
angle = angle + np.pi
if self.relative(intersect, center)[1]<0 and self.relative(intersect, center)[0] > 0:
angle = 2*np.pi - angle*-1
###
#y = sx + t
#0 = sx + t
#-t = sx
#x = -t/s
top_intersect_x = -t/s
top_intersect = (top_intersect_x, 0)
###
#Lines[i].set((length, angle, top_intersect_x))
Lines = Lines + [line_attr()]
Lines[-1].set((length, angle, top_intersect_x))
except:
lines_seen = lines_seen - 1
if lines_seen > 0:
neg_length = neg_length/neg_len_count
pos_length = pos_length/pos_len_count
Length_Deviation = pos_length + neg_length
if pos_length == 0 or neg_length == 0: #If one side gone
pos_length = pos_length *2
neg_length = neg_length *2
Line_Distance = (abs(neg_length) + pos_length)/2
Total_Top = 0
Top_pos = [0, 0] #value, amount
Top_neg = [0, 0]
Total_Angle = 0
Angles = lines_seen
#Total_Acute_Angle = 0
#Total_Obtuse_Angle = 0
#Total_Right_Angle = 0
#Angle_Counter = [1, 1, 1, 3] #Acute, Obtuse, Right, Types
for i in range(line_amount):
try:
if Lines[i].top_intersect_x > center[0]:
Top_pos[0] = Top_pos[0] + Lines[i].top_intersect_x
Top_pos[1] = Top_pos[1] + 1
elif Lines[i].top_intersect_x < center[0]:
Top_neg[0] = Top_neg[0] + Lines[i].top_intersect_x
Top_neg[1] = Top_neg[1] + 1
if Lines[i].angle < np.pi/2:
Angle = Lines[i].angle + np.pi/2
elif Lines[i].angle > np.pi/2 and Lines[i].angle < np.pi:
Angle = Lines[i].angle - np.pi/2
elif Lines[i].angle> np.pi and Lines[i].angle< np.pi*3/2:
Angle = Lines[i].angle - np.pi/2
elif Lines[i].angle> np.pi*3/2:
Angle = Lines[i].angle - np.pi*3/2
else:
Angle = 0
#print(Angle)
Total_Angle = Total_Angle + Angle
'''
if Angle < np.pi/2:
Total_Acute_Angle = Total_Acute_Angle + Angle
Angle_Counter[0] += 1
elif Angle > np.pi/2:
Total_Obtuse_Angle = Total_Obtuse_Angle + Angle
Angle_Counter[1] += 1
elif Angle == np.pi/2:
Total_Right_Angle = Total_Right_Angle + Angle
Angle_Counter[2] += 1
'''
except:
pass
Total_Angle = Total_Angle/Angles
if Top_pos[1] != 0 and Top_neg[1] != 0:
Total_Top = (Top_pos[0]/Top_pos[1] + Top_neg[0]/Top_neg[1])/2
elif Top_pos[1] == 0:
Total_Top = (Top_neg[0]/Top_neg[1])
elif Top_neg[1] == 0:
Total_Top = (Top_pos[0]/Top_pos[1])
'''
Angles.append(Total_Acute_Angle/Angle_Counter[0])
Angles.append(Total_Obtuse_Angle/Angle_Counter[1])
Angles.append(Total_Right_Angle/Angle_Counter[2])
for i in range(3):
if Angle_Counter[i] == 1:
Angle_Counter[3] -= 1
else:
pass
Total_Angle = Total_Angle + Angles[i]
Total_Angle = Total_Angle/Angle_Counter[3]
'''
cv2.circle(process, (round(Total_Top), 0), 2, (0, 0, 255), 2)
else:
Total_Top = center[0]
Line_Distance = (Updown_Deadzone[0]+Updown_Deadzone[1])/2
Total_Angle = np.pi/2
Length_Deviation = 0
'''
if Total_Top <center[0] - 30:
print('Move Right')
elif Total_Top > center[0] - 30 and Total_Top<center[0] + 30:
print("Don't Move")
else:
print('Move Left')
if Total_Angle > np.pi/2 - 0.2 and Total_Angle < np.pi/2 + 0.2:
print("Don't turn")
elif Total_Angle > np.pi/2 + 0.2:
print('Turn Left')
elif Total_Angle < np.pi/2 - 0.2:
print('Turn Right')
if Line_Distance > 100:
print("Go Higher")
elif Line_Distance<70:
print("Go Lower")
else:
print("Height OK")
print('\n\n\n\n\n\n')
'''
'''
#Strafe_Power = Length_Deviation/(width/2)
Yaw_Power =(Total_Top-width//2)/(width//2)
Updown_Power = 0
if Line_Distance > 150:
Updown_Power = (150-Line_Distance)/100
if Updown_Power > 1:
Updown_Power = 1
elif Line_Distance < 100:
Updown_Power = (100-Line_Distance)/100
Message = (Strafe_Power,Drive_Power,Yaw_Power,Updown_Power,0,0) #Strafe, drive, yaw, updown, 0, 0
'''
if Total_Top > center[0] - Strafe_Deadzone and Total_Top<center[0] + Strafe_Deadzone:
Total_Top = width/2
Strafe_Power = (Total_Top-width/2)/(width/2)
if Strafe_Power > 1:
Strafe_Power = 1
elif Strafe_Power < -1:
Strafe_Power = -1
if Total_Angle > np.pi/2 - Yaw_Deadzone and Total_Angle < np.pi/2 + Yaw_Deadzone:
Total_Angle = np.pi/2
Yaw_Power = (-Total_Angle+np.pi/2)/(np.pi/2)
if Line_Distance < Updown_Deadzone[1] and Line_Distance>Updown_Deadzone[0]:
Line_Distance = max_line_distance/2
elif Line_Distance<=Updown_Deadzone[0]:
Line_Distance = Line_Distance*((max_line_distance/2)/Updown_Deadzone[0])
elif Line_Distance >= Updown_Deadzone[1]:
Line_Distance = ((Line_Distance - Updown_Deadzone[1])*((max_line_distance/2)/(max_line_distance-Updown_Deadzone[1])))+max_line_distance/2
Updown_Power = (Line_Distance-max_line_distance/2)/(max_line_distance/2)
if Updown_Power > 1:
Updown_Power = 1
elif Updown_Power < -1:
Updown_Power = -1
#Value Modifiers
Drive_Power = self.drive_power #0-1
Strafe_Power = PowerFunction(Strafe_Power, self.strafe_mod)
Yaw_Power = PowerFunction(Yaw_Power, self.yaw_mod)
Updown_Power = PowerFunction(Updown_Power, self.updown_mod)
if self.show:
cv2.imshow('frame',frame)
cv2.imshow('process', process)
Powers = [Strafe_Power,Drive_Power,Yaw_Power,Updown_Power,0,0] #Strafe, drive, yaw, updown, 0, 0
#print(Powers)
cv2.waitKey(1)
#pub.sendMessage("transectline", message = Powers)
#a = [1, 2, 3, 4, 5, 6]
#pub_to_manager("control-movement", message = ("transectline", Powers))
pub.sendMessage("control-movement", message = ("transectline", Powers))
else:
if self.captureON == True:
self.cap.release()
cv2.destroyAllWindows()
self.captureON = False
for line in lines:
x1,y1,x2,y2 = line[0]
cv.line(img,(x1,y1),(x2,y2),(0,255,0),2)
cv.imshow("result", img)
cv.waitKey(0)
# HoughLines Code
img = cv.imread('hallway.jpg')
gray = cv.cvtColor(img,cv.COLOR_BGR2GRAY)
edges = cv.Canny(gray,50,150)
lines = cv.HoughLines(edges,1,np.pi/180,200)
for line in lines:
rho,theta = line[0]
a = np.cos(theta)
b = np.sin(theta)
x0 = a*rho
y0 = b*rho
x1 = int(x0 + 1000*(-b))
y1 = int(y0 + 1000*(a))
x2 = int(x0 - 1000*(-b))
y2 = int(y0 - 1000*(a))
cv.line(img,(x1,y1),(x2,y2),(0,0,255),2)
cv.imshow("result", img)
cv.waitKey(0)