def calculateAffinityMatrixAndDraw(self, bestImage, inliersDataBase, inliersWebCam, imgout):
# The affinity matrix A
A = cv2.estimateRigidTransform(inliersDataBase, inliersWebCam, fullAffine=True)
A = np.vstack((A, [0, 0, 1]))
# Calculate the points of the rectangle occupied by the recognized object
a = np.array([0, 0, 1], np.float)
b = np.array([bestImage.shape[1], 0, 1], np.float)
c = np.array([bestImage.shape[1], bestImage.shape[0], 1], np.float)
d = np.array([0, bestImage.shape[0], 1], np.float)
centro = np.array([float(bestImage.shape[0]) / 2,
float(bestImage.shape[1]) / 2, 1], np.float)
# Multiply the points of the virtual space, to convert them into Real image points
a = np.dot(A, a)
b = np.dot(A, b)
c = np.dot(A, c)
d = np.dot(A, d)
centro = np.dot(A, centro)
# The points are dehomogenized
areal = (int(a[0] / a[2]), int(a[1] / b[2]))
breal = (int(b[0] / b[2]), int(b[1] / b[2]))
creal = (int(c[0] / c[2]), int(c[1] / c[2]))
dreal = (int(d[0] / d[2]), int(d[1] / d[2]))
centroreal = (int(centro[0] / centro[2]), int(centro[1] / centro[2]))
# The polygon and the file name of the image are painted in the center of the polygon
points = np.array([areal, breal, creal, dreal], np.int32)
cv2.polylines(imgout, np.int32([points]), 1, (255, 255, 255), thickness=2)
utilscv.draw_str(imgout, centroreal, bestImage.nameFile.upper())
def p4_object_recog(self, frame):
try:
t1 = time.time()
image_in = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
image_out = frame.copy()
kp, desc = self.detector.detectAndCompute(image_in, None)
selectedDataBase = self.dataBase[self.method_str]
if len(selectedDataBase) > 0:
# We perform the mutual matching
imgsMatchingMutuos = self.find_matching_mutuos_optimum(self.method_str, desc, kp)
minInliers = int(cv2.getTrackbarPos('inliers', 'Features'))
projer = float(cv2.getTrackbarPos('projer', 'Features'))
LOG.info("Line: %s :: minInliers: %s, projer: %s", print_frame(), minInliers, projer)
# The best image is calculated based on the number of inliers.
# The best image is one that has more number of inliers,
# but always exceeding the minimum that is indicated in the trackbar 'minInliers'
bestImage, inliersWebCam, inliersDataBase = self.calculate_best_image_by_num_inliers(self.method_str,
projer, minInliers)
if not bestImage is None:
# If we find a good image, we calculate the affinity matrix
# and paint the recognized object on the screen.
self.calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam, image_out)
# Get descriptor dimension of each feature:
if desc is not None:
if len(desc) > 0:
dim = len(desc[0])
else:
dim = -1
# We draw features, and write informative text about the image
# Only the features are dibuban if the slider indicates it
if (int(cv2.getTrackbarPos('drawKP', 'Features')) > 0):
cv2.drawKeypoints(image_out, kp, image_out,
flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
t1 = 1000 * (time.time() - t1)
utilscv.draw_str(image_out, (20, 20),
"Method {0}, {1} features found, desc. dim. = {2} ".
format(self.method_str, len(kp), dim))
utilscv.draw_str(image_out, (20, 40), "Time (ms): {0}".format(str(t1)))
# Show results and check keys:
# self.display_image('Features', imgage_out, resize_max=640, wait_time=1)
return image_out
except Exception as e:
LOG.info("* ** p4_object_recog Error: %s\n" % str(e))
traceback.print_exc()
def calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam,
imgout):
#Se calcula la matriz de afinidad A
A = cv2.estimateRigidTransform(inliersDataBase,
inliersWebCam,
fullAffine=True)
A = np.vstack((A, [0, 0, 1]))
#Se calculan los puntos del rectangulo que ocupa el objeto reconocido
a = np.array([0, 0, 1], np.float)
b = np.array([bestImage.shape[1], 0, 1], np.float)
c = np.array([bestImage.shape[1], bestImage.shape[0], 1], np.float)
d = np.array([0, bestImage.shape[0], 1], np.float)
centro = np.array(
[float(bestImage.shape[0]) / 2,
float(bestImage.shape[1]) / 2, 1], np.float)
#Se multiplican los puntos del espacio virtual, para convertirlos en
#puntos reales de la imagen
a = np.dot(A, a)
b = np.dot(A, b)
c = np.dot(A, c)
d = np.dot(A, d)
centro = np.dot(A, centro)
#Se deshomogeneizan los puntos
areal = (int(a[0] / a[2]), int(a[1] / b[2]))
breal = (int(b[0] / b[2]), int(b[1] / b[2]))
creal = (int(c[0] / c[2]), int(c[1] / c[2]))
dreal = (int(d[0] / d[2]), int(d[1] / d[2]))
centroreal = (int(centro[0] / centro[2]), int(centro[1] / centro[2]))
#Se pinta el polígono y el nombre del fichero de la imagen en el centro del polígono
points = np.array([areal, breal, creal, dreal], np.int32)
cv2.polylines(imgout, np.int32([points]), 1, (255, 255, 255), thickness=2)
utilscv.draw_str(imgout, centroreal, bestImage.nameFile.upper())
#Se visualiza el objeto detectado en una ventana a parte
cv2.imshow('ImageDetector', bestImage.imageBinary)
def callback(self,data):
try:
frame = self.bridge.imgmsg_to_cv2(data, "bgr8")
except CvBridgeError as e:
print(e)
if (not self.paused) and (frame is not None):
# self.curFrame = 0
# convert color to gray
imgin = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
# output image for display
imgout = frame.copy()
t1 = time.time()
kp, desc = self.detector.detectAndCompute(imgin, None)
selectedDataBase = self.dataBaseDictionary[self.methodstr]
if len(selectedDataBase) > 0:
# Mutual Matching
imgsMatchingMutuos = orec.findMatchingMutuosOptimizado(selectedDataBase, desc, kp)
minInliers = int(20)
projer = float(5)
# Calculating the best image according to the inliers numbers
# The best image is the one with the greater inliers number,
# Always taking into account that it should exceed the minimum indicated in
# the trackbar as 'minInliers'
bestImage, inliersWebCam, inliersDataBase = orec.calculateBestImageByNumInliers(selectedDataBase, projer, minInliers)
rate = rospy.Rate(10)
motion = Twist();
if not bestImage is None:
#If we find a good image, we calculate the affinity matrix and mark it on the screen as a recognized object
self.depthObj, center = orec.calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam, imgout)
self.msg_recog.variance = 1
self.msg_recog.temperature = self.depthObj
self.msg_recog.header.frame_id = bestImage.nameFile.upper().split('.')[0]
self.msg_recog.header.stamp = rospy.get_rostime();
xc = 160
dx = xc - center[0]
motion.angular.z = dx/500.0;
if self.depthObj > 0.4:
motion.linear.x = 0.3
if self.depthObj < 0.2:
motion.linear.x = -0.3
self.motionPub.publish(motion);
else:
motion.linear.x = 0
self.motionPub.publish(motion);
rate.sleep();
self.depthObj = 0
self.msg_recog.variance = 0
self.msg_recog.temperature = 0.00
self.msg_recog.header.frame_id = 'None'
self.msg_recog.header.stamp = rospy.get_rostime();
t1 = 1000 * (time.time() - t1) # Time in milliseconds
# Obtain the descriptors dimension in each features
if desc is not None:
if len(desc) > 0:
dim = len(desc[0])
else:
dim = -1
imgout = cv2.resize(imgout, (0, 0), fx = 1.45, fy = 1.45)
# Drawing features, writing the informative text on the image
# And, only draw features if the slider indicates it
if self.drawKP > 0:
cv2.drawKeypoints(imgout, kp, imgout, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
utilscv.draw_str(imgout, (20, 20),
"Method {0}, {1} features found ".
format(self.methodstr, len(kp)))
utilscv.draw_str(imgout, (20, 40), "Time (ms): {0}".format(str(t1)))
if self.depthObj == 0:
utilscv.draw_str(imgout, (20, 60),
"Depth to the object: None")
else:
utilscv.draw_str(imgout, (20, 60),
"Depth to the object: {0} m".
format(np.around(self.depthObj, decimals = 3)))
# Show results and check the keyboards
cv2.imshow('Pepper Object Recognition', imgout)
ch = cv2.waitKey(5) & 0xFF
if ch == ord(' '): # Pause space bar
self.paused = not self.paused
self.recog_pub.publish(self.msg_recog)
orec.calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam, imgout)
t1 = 1000 * (time.time() - t1) # Tiempo en milisegundos
# Obtener dimensión de descriptores de cada feature:
if desc is not None:
if len(desc) > 0:
dim = len(desc[0])
else:
dim = -1
# Dibujamos features, y escribimos texto informativo sobre la imagen
# Solo se dibuban las features si el slider lo indica
if (int(cv2.getTrackbarPos('drawKP', 'Features')) > 0):
cv2.drawKeypoints(imgout, kp, imgout,
flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
utilscv.draw_str(imgout, (20, 20),
"Method {0}, {1} features found, desc. dim. = {2} ".
format(methodstr, len(kp), dim))
utilscv.draw_str(imgout, (20, 40), "Time (ms): {0}".format(str(t1)))
# Mostrar resultados y comprobar teclas:
cv2.imshow('Features', imgout)
ch = cv2.waitKey(5) & 0xFF
if ch == 27: # Escape termina
break
elif ch == ord(' '): # Barra espaciadora pausa
paused = not paused
elif ch == ord('.'): # Punto avanza un sólo frame
paused = True
frame = videoinput.read()
# Cerrar ventana(s) y fuente(s) de vídeo:
videoinput.close()
def calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam, imgout):
#A Calculates the affinity matrix A
A = cv2.estimateRigidTransform(inliersDataBase, inliersWebCam, fullAffine=True)
A = np.vstack((A, [0, 0, 1]))
#A Calculated points of the rectangle that occupies the recognized object
a = np.array([0, 0, 1], np.float)
b = np.array([bestImage.shape[1], 0, 1], np.float)
c = np.array([bestImage.shape[1], bestImage.shape[0], 1], np.float)
d = np.array([0, bestImage.shape[0], 1], np.float)
centro = np.array([float(bestImage.shape[0])/2,
float(bestImage.shape[1])/2, 1], np.float)
#A Multiply the points of the virtual space, to become
#puntos real image
a = np.dot(A, a)
b = np.dot(A, b)
c = np.dot(A, c)
d = np.dot(A, d)
centro = np.dot(A, centro)
#A Deshomogeneizan points
areal = (int(a[0]/a[2]), int(a[1]/b[2]))
breal = (int(b[0]/b[2]), int(b[1]/b[2]))
creal = (int(c[0]/c[2]), int(c[1]/c[2]))
dreal = (int(d[0]/d[2]), int(d[1]/d[2]))
centroreal = (int(centro[0]/centro[2]), int(centro[1]/centro[2]))
#height = imgout.shape[0]
#width = imgout.shape[1]
#if areal[0]<0:
# areal = (0, areal[1])
#elif areal[0]> width:
# areal = (width, areal[1])
#if creal[0]<0:
# creal = (0, creal[1])
#elif creal[0]> width:
# creal = (width, creal[1])
#if areal[1]<0:
# areal = (areal[0], 0)
#elif areal[1]> height:
# areal = (areal[0], height)
#if creal[1]<0:
# creal = (creal[0], 0)
#elif creal[1]> height:
# creal = (creal[0], height)
#print(areal,breal,creal,dreal)
#if areal[1] >= creal[1] and areal[0] >= creal[0]:
# imgshow = imgout[creal[1]:areal[1], creal[0]:areal[0]]
#if areal[1] <= creal[1] and areal[0] > creal[0]:
# imgshow = imgout[areal[1]:creal[1], creal[0]:areal[0]]
#if areal[1] < creal[1] and areal[0] <= creal[0]:
# imgshow = imgout[areal[1]:creal[1], areal[0]:creal[0]]
#if areal[1] > creal[1] and areal[0] < creal[0]:
# imgshow = imgout[creal[1]:areal[1], areal[0]:creal[0]]
#cv2.imshow('Test', imgshow)
#cv2.imwrite("modelos/CRYSIS"+ str(areal[1])+".jpg",imgshow)
#A Paints the polygon and the file name of the image in the center of the polygon
points = np.array([areal, breal, creal, dreal], np.int32)
cv2.polylines(imgout, np.int32([points]),1, (0,255,255), thickness=2)
utilscv.draw_str(imgout, centroreal, bestImage.nameFile.upper())
#A Displays the detected object in a window part
cv2.imshow('ImageDetector', bestImage.imageBinary)
# If we find a good image, the affinity matrix is calculated and the recognized object is painted on the screen.
orec.calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam, imgout)
t1 = 1000 * (time.time() - t1) # Time in milliseconds
# Get dimension of descriptors for each feature:
if desc is not None:
if len(desc) > 0:
dim = len(desc[0])
else:
dim = -1
# Features draw and write text about the image
# Only features are drawn if the slider indicates
if int(cv2.getTrackbarPos("drawKP", "Features")) > 0:
cv2.drawKeypoints(imgout, kp, imgout, flags=cv2.DRAW_MATCHES_FLAGS_DRAW_RICH_KEYPOINTS)
utilscv.draw_str(
imgout, (20, 20), "Method {0}, {1} features found, desc. dim. = {2} ".format(methodstr, len(kp), dim)
)
utilscv.draw_str(imgout, (20, 40), "Time (ms): {0}".format(str(t1)))
# Show results and check keys:
cv2.imshow("Features", imgout)
ch = cv2.waitKey(5) & 0xFF
if ch == 27: # escape ends
break
elif ch == ord(" "): # Spacebar pause
paused = not paused
elif ch == ord("."): # Point moves forward one frame
paused = True
ret, frame = cap.read()
# Close windows and fonts video:
cap.release()
def calculateAffinityMatrixAndDraw(bestImage, inliersDataBase, inliersWebCam,
imgout):
#A Calculates the affinity matrix A
A = cv2.estimateRigidTransform(inliersDataBase,
inliersWebCam,
fullAffine=True)
A = np.vstack((A, [0, 0, 1]))
#A Calculated points of the rectangle that occupies the recognized object
a = np.array([0, 0, 1], np.float)
b = np.array([bestImage.shape[1], 0, 1], np.float)
c = np.array([bestImage.shape[1], bestImage.shape[0], 1], np.float)
d = np.array([0, bestImage.shape[0], 1], np.float)
centro = np.array(
[float(bestImage.shape[0]) / 2,
float(bestImage.shape[1]) / 2, 1], np.float)
#A Multiply the points of the virtual space, to become
#puntos real image
a = np.dot(A, a)
b = np.dot(A, b)
c = np.dot(A, c)
d = np.dot(A, d)
centro = np.dot(A, centro)
#A Deshomogeneizan points
areal = (int(a[0] / a[2]), int(a[1] / b[2]))
breal = (int(b[0] / b[2]), int(b[1] / b[2]))
creal = (int(c[0] / c[2]), int(c[1] / c[2]))
dreal = (int(d[0] / d[2]), int(d[1] / d[2]))
centroreal = (int(centro[0] / centro[2]), int(centro[1] / centro[2]))
#height = imgout.shape[0]
#width = imgout.shape[1]
#if areal[0]<0:
# areal = (0, areal[1])
#elif areal[0]> width:
# areal = (width, areal[1])
#if creal[0]<0:
# creal = (0, creal[1])
#elif creal[0]> width:
# creal = (width, creal[1])
#if areal[1]<0:
# areal = (areal[0], 0)
#elif areal[1]> height:
# areal = (areal[0], height)
#if creal[1]<0:
# creal = (creal[0], 0)
#elif creal[1]> height:
# creal = (creal[0], height)
#print(areal,breal,creal,dreal)
#if areal[1] >= creal[1] and areal[0] >= creal[0]:
# imgshow = imgout[creal[1]:areal[1], creal[0]:areal[0]]
#if areal[1] <= creal[1] and areal[0] > creal[0]:
# imgshow = imgout[areal[1]:creal[1], creal[0]:areal[0]]
#if areal[1] < creal[1] and areal[0] <= creal[0]:
# imgshow = imgout[areal[1]:creal[1], areal[0]:creal[0]]
#if areal[1] > creal[1] and areal[0] < creal[0]:
# imgshow = imgout[creal[1]:areal[1], areal[0]:creal[0]]
#cv2.imshow('Test', imgshow)
#cv2.imwrite("modelos/CRYSIS"+ str(areal[1])+".jpg",imgshow)
#A Paints the polygon and the file name of the image in the center of the polygon
points = np.array([areal, breal, creal, dreal], np.int32)
cv2.polylines(imgout, np.int32([points]), 1, (0, 255, 255), thickness=2)
utilscv.draw_str(imgout, centroreal, bestImage.nameFile.upper())
#A Displays the detected object in a window part
cv2.imshow('ImageDetector', bestImage.imageBinary)
