def __init__(self):
        # Define camera settings
        # put bright, contrast, saturation, hue into image_processing_param.yaml file
        self.vid = cv2.VideoCapture(
            rospy.get_param("/videoDevicePath"
                            ))  # Sets the port /dev/video6 as the video device
        self.vid.set(10, rospy.get_param("/brightness"))  # brightness
        self.vid.set(11, rospy.get_param("/contrast"))  # contrast
        self.vid.set(12, rospy.get_param("/saturation"))  # saturation
        self.vid.set(13, rospy.get_param("/hue"))  # hue

        # Decalre calibration matrices to rectify the image

        self.mtx = np.array(rospy.get_param("/mtx"))
        self.dist = np.array(rospy.get_param("/dist"))

        # Camera resolution
        self.width = rospy.get_param("/width")
        self.height = rospy.get_param("/height")

        # Reference velocity
        self.v_ref = rospy.get_param("/reference_velocity")

        # Number of points for moving average filter
        self.numMovingAveragePoints = rospy.get_param(
            "/numMovingAveragePoints")
        self.movingAverageValue = np.zeros([2, self.numMovingAveragePoints])

        # Number of sample points at the reference velocity to check along the path
        self.numpoints = rospy.get_param("/numStepsToLookAhead")

        # Set node loop rate (30 hz)
        self.loop_rate = rospy.get_param("/loop_rate")
        self.dt = 1.0 / self.loop_rate
        self.rate = rospy.Rate(self.loop_rate)

        self.f1Matrix = rospy.get_param("/f1Matrix")
        self.f2Matrix = rospy.get_param("/f2Matrix")
        self.bMatrix = rospy.get_param("/bMatrix")
        self.yPixel_to_xInertial_Matrix = rospy.get_param(
            "/yPixel_to_xInertial_Matrix")
        self.xInertial_to_yPixel_Matrix = rospy.get_param(
            "/xInertial_to_yPixel_Matrix")
        self.furthest_distance = rospy.get_param("/furthest_distance")
        self.camera_offset_distance = rospy.get_param(
            "/camera_offset_distance")
        self.flipped_camera = rospy.get_param("/flipped_camera")

        # Compute the udistortion and rectification transformation map
        self.newcameramtx, self.roi = cv2.getOptimalNewCameraMatrix(
            self.mtx, self.dist, (self.width, self.height), 0,
            (self.width, self.height))
        self.mapx, self.mapy = cv2.initUndistortRectifyMap(
            self.mtx, self.dist, None, self.newcameramtx,
            (self.width, self.height), 5)

        # Messages to be filled
        self.state_constraints = barc_state()
        self.reference_trajectory = barc_state()

        self.bridge = CvBridge()

        # Initialize publishers and subscribers
        self.moving_pub = rospy.Publisher("moving", Moving, queue_size=1)
        self.hold_previous_turn_pub = rospy.Publisher("hold_previous_turn",
                                                      Bool,
                                                      queue_size=1)
        self.moving_pub.publish(True)
        self.reference_trajectory_pub = rospy.Publisher("reference_trajectory",
                                                        barc_state,
                                                        queue_size=1)
        self.reference_image_pub = rospy.Publisher("image_raw",
                                                   Image,
                                                   queue_size=1)
        self.uOpt_pub = rospy.Publisher("uOpt", Input, queue_size=1)
        self.optimal_state_sub = rospy.Subscriber("optimal_state_trajectory",
                                                  barc_state,
                                                  self.convertDistanceToPixels)
        self.dt_pub = rospy.Publisher("dt", Float64, queue_size=1)

        # The boolean messages passed to these topics are not used, we only want them for the independently threaded callback function.
        self.draw_lines_pub = rospy.Publisher("draw_lines", Bool, queue_size=1)
        self.draw_lines_sub = rospy.Subscriber("draw_lines",
                                               Bool,
                                               self.draw_lines,
                                               queue_size=1)
        self.publish_states_pub = rospy.Publisher("publish_states",
                                                  Bool,
                                                  queue_size=1)
        self.publish_states_sub = rospy.Subscriber("publish_states",
                                                   Bool,
                                                   self.publish_states,
                                                   queue_size=1)
        self.show_Image_pub = rospy.Publisher("show_Image", Bool, queue_size=1)
        self.show_Image_sub = rospy.Subscriber("show_Image",
                                               Bool,
                                               self.show_Image,
                                               queue_size=1)

        self.count = 0
        self.totalTimeCounter = 0
        self.totalTime = 0
        self.averageTime = 0
        self.publish_image = True
        self.previousTime = time.time()
        self.printme = False
        self.statepoints = ''
        self.camera_distance_calibrated = False
        print(
            "Press Up Arrow to start moving. Press Down Arrow to stop moving.")
        while not rospy.is_shutdown():
            try:
                self.count = self.count + 1  # updates the count
                self.rel, self.dst = self.vid.read(
                )  # gets the current frame from the camera

                # Updates the sample time
                self.dt = time.time() - self.previousTime
                self.previousTime = time.time()
                if self.flipped_camera:
                    self.cv_image = cv2.flip(
                        cv2.remap(self.dst, self.mapx, self.mapy,
                                  cv2.INTER_LINEAR),
                        -1)  #Undistorts the fisheye image to rectangular
                else:
                    self.cv_image = cv2.remap(
                        self.dst, self.mapx, self.mapy, cv2.INTER_LINEAR
                    )  #Undistorts the fisheye image to rectangular
                self.x, self.y, self.width, self.height = self.roi

                # colorFilter = True makes the edge detection search for a red/white track using HSV. False will use grayscale and search for any edge regardless of color
                # colorFilter =  rospy.get_param("/colorFilter")
                colorFilter = False
                kernel_size = rospy.get_param("/kernel_size")
                if colorFilter:
                    imageToFilter = self.cv_image
                    imageToFilter[
                        0:280, 0:self.
                        width] = 0  #blacks out the top portion of the image (not used)

                    #self.hsv = cv2.cvtColor(imageToFilter, cv2.COLOR_BGR2HSV) #.004

                    # define range of color thresholds in (B,G,R)
                    lower_red = np.flipud(
                        np.array(rospy.get_param("/lower_red")))
                    upper_red = np.flipud(
                        np.array(rospy.get_param("/upper_red")))

                    lower_white = np.flipud(
                        np.array(rospy.get_param("/lower_white")))
                    upper_white = np.flipud(
                        np.array(rospy.get_param("/upper_white")))

                    # Threshold the image to only have the red/white track appear
                    self.reds = cv2.inRange(imageToFilter, lower_red,
                                            upper_red)
                    self.whites = cv2.inRange(imageToFilter, lower_white,
                                              upper_white)

                    self.edges = cv2.bitwise_or(
                        self.reds, self.whites
                    )  # combines the red filter and white filter images
                    self.edges = cv2.GaussianBlur(self.edges,
                                                  (kernel_size, kernel_size),
                                                  0)  # blurs the image
                    retval, self.edges = cv2.threshold(
                        self.edges, 127, 255, cv2.THRESH_BINARY
                    )  # converts the blurred greyscale to binary to filter once more

                else:
                    # Convert Color Image to Grayscale
                    gray_image = cv2.cvtColor(self.cv_image,
                                              cv2.COLOR_BGR2GRAY)
                    gray_image[0:270, 0:self.width] = 0
                    gray_image = cv2.GaussianBlur(gray_image,
                                                  (kernel_size, kernel_size),
                                                  0)
                    self.edges = cv2.Canny(
                        gray_image, 40, 80)  # openCV edge detection function

                # Parameters to combine image: dst = alpha*img1+beta*img2+gamma
                alpha = 0.6
                beta = 1
                gamma = 0

                # overlayPointsOnColoredImage = True makes the path show up on top of the colored image.
                # Draw lanes over image (color or black/white)
                overlayPointsOnColoredImage = rospy.get_param(
                    "/overlayPointsOnColoredImage")
                if overlayPointsOnColoredImage:
                    self.line_img_color = np.zeros(self.cv_image.shape,
                                                   dtype=np.uint8)
                    self.pathOverlayedImage = cv2.addWeighted(
                        self.cv_image, alpha, self.line_img_color, beta, gamma)
                else:
                    self.edges_color = cv2.cvtColor(self.edges,
                                                    cv2.COLOR_GRAY2RGB)
                    self.line_img_color = np.zeros(self.edges_color.shape,
                                                   dtype=np.uint8)
                    self.pathOverlayedImage = cv2.addWeighted(
                        self.edges_color, alpha, self.line_img_color, beta,
                        gamma)

                # Collect 100 images before running image processing
                if self.count > 100:
                    self.draw_lines_pub.publish(True)

                # Publish image with lanes
                if self.publish_image:
                    if True:
                        self.reference_image_pub.publish(
                            self.bridge.cv2_to_imgmsg(
                                cv2.cvtColor(self.edges, cv2.COLOR_GRAY2RGB),
                                "bgr8"))
                    else:
                        print('Could not publish reference image')

                # Check the true loop rate of incoming images from camera (i.e. frames per second should match parameter specified in launch file)
                if (self.count > 100 and self.count % 3 == 0):
                    self.totalTimeCounter += 1
                    self.timenext = time.time()
                    self.timeElapsed = self.timenext - self.previousTime
                    self.totalTime = self.totalTime + self.timeElapsed
                    self.averageTime = self.totalTime / (self.totalTimeCounter)
                    self.dt_pub.publish(self.averageTime)
                    #print('Average Time: ',self.averageTime)
                self.rate.sleep()

            except IOError, (ErrorNumber, ErrorMessage):
                print('HERE')
                print('HERE')
                print(ErrorMessage)
                pass
Пример #2
0
    def __init__(self):

        self.vid = cv2.VideoCapture("/dev/video6")
        self.vid.set(12, 5)  #contrast
        self.vid.set(13, 255)  #saturation

        # Calibration Matrices
        self.mtx = np.array([[592.156892, 0.000000, 326.689246],
                             [0.000000, 584.923917, 282.822026],
                             [0.000000, 0.000000, 1.000000]])
        self.dist = np.array(
            [-0.585868, 0.248490, -0.023236, -0.002907, 0.000000])
        self.rel, self.dst = self.vid.read()
        # Camera resolution
        self.w = 640
        self.h = 480

        # Reference velocity
        self.v_ref = 1.2

        # Number of moving average points
        self.sx = 5
        self.movmean = np.zeros([2, self.sx])

        # Set node rate
        self.loop_rate = 30
        self.ts = 1.0 / self.loop_rate
        self.rate = rospy.Rate(self.loop_rate)
        self.t0 = time.time()

        self.dt = self.ts
        self.count = 0
        self.incount = 0
        self.total = 0
        self.avg = 0
        self.total2 = 0
        self.avg2 = 0
        self.publish_image = False
        self.timeprev = time.time() - self.dt

        time.sleep(.5)

        # Compute the udistortion and rectification transformation map
        self.newcameramtx, self.roi = cv2.getOptimalNewCameraMatrix(
            self.mtx, self.dist, (self.w, self.h), 0, (self.w, self.h))
        self.mapx, self.mapy = cv2.initUndistortRectifyMap(
            self.mtx, self.dist, None, self.newcameramtx, (self.w, self.h), 5)

        self.statepoints = ''
        self.printme = True

        self.state_constraints = barc_state()
        self.reference_trajectory = barc_state()

        # Initialize publishers and subscribers
        self.moving_pub = rospy.Publisher("moving", Moving, queue_size=1)
        self.moving_pub.publish(True)
        self.reference_trajectory_pub = rospy.Publisher("reference_trajectory",
                                                        barc_state,
                                                        queue_size=1)
        self.reference_image_pub = rospy.Publisher(
            "image_reference_trajectory", Image, queue_size=10)
        self.uOpt_pub = rospy.Publisher("uOpt", Input, queue_size=1)

        while not rospy.is_shutdown():
            try:
                self.count = self.count + 1
                self.rel, self.dst = self.vid.read(
                )  # gets the current frame from the camera
                self.dt = time.time() - self.timeprev
                self.timeprev = time.time()

                self.cv_image = cv2.remap(
                    self.dst, self.mapx, self.mapy, cv2.INTER_LINEAR
                )  #Undistorts the fisheye image to rectangular
                self.x, self.y, self.w, self.h = self.roi
                self.dst = self.dst[self.y:self.y + self.h,
                                    self.x:self.x + self.w]

                # yellow = True makes the edge detection search for a yellow track using HSV. False will use grayscale and search for any edge regardless of color
                yellow = True
                kernel_size = 5
                if yellow:
                    cropped = self.cv_image
                    cropped[0:280, 0:640] = 0
                    #cropped = cv2.GaussianBlur(cropped,(kernel_size,kernel_size),0) #0.017s
                    #cropped = cv2.medianBlur(cropped,kernel_size)

                    hsv = cv2.cvtColor(cropped, cv2.COLOR_BGR2HSV)  #.004

                    #hsv = cv2.GaussianBlur(hsv,(kernel_size,kernel_size),0)
                    #######cv2.imshow('hsv',hsv[270:480,:])
                    """ THIS IS THE YELLLOW LINE DETECION CODE
                    # define range of blue color in HSV (B,G,R)
                    lower_yellow = np.array([0,180,100])
                    upper_yellow = np.array([50,255,255])

                    # Threshold the HSV image to get only blue colors
                    edges = cv2.inRange(hsv, lower_yellow, upper_yellow) #0.03s
                    
                    """

                    # UNCOMMENT THIS FOR THE REDWHITE TRACK
                    lower_red = np.array([0, 0, 180])
                    upper_red = np.array([50, 120, 255])

                    lower_white = np.array([150, 150, 150])
                    upper_white = np.array([255, 255, 255])

                    # Threshold the HSV image to get only blue colors
                    self.reds = cv2.inRange(self.cv_image, lower_red,
                                            upper_red)
                    self.whites = cv2.inRange(self.cv_image, lower_white,
                                              upper_white)  #0.03s
                    edges = cv2.bitwise_or(self.reds, self.whites)
                    edges = cv2.GaussianBlur(edges, (kernel_size, kernel_size),
                                             0)

                    #edges = cv2.cvtColor(edges, cv2.COLOR_BGR2GRAY)
                    #cv2.imshow("hsv to gray",edges)
                    #cv2.imshow("Edges",edges[270:480,:])
                    ####edges = cv2.GaussianBlur(edges,(kernel_size,kernel_size),0)
                    #edges = cv2.Canny(edges,10,200)
                    #######cv2.imshow("Edges2",edges[270:480,:])
                    edgescropped = edges
                else:
                    # Convert Color Image to Grayscale
                    gray_image = cv2.cvtColor(self.cv_image,
                                              cv2.COLOR_BGR2GRAY)
                    gray_image[0:270, 0:640] = 0
                    gray_image = cv2.GaussianBlur(gray_image,
                                                  (kernel_size, kernel_size),
                                                  0)
                    #cv2.imshow("blurred Image", gray_image)
                    #in MPC lab, 13 15 100
                    edges = cv2.Canny(gray_image, 40, 80)
                    cv2.imshow("Advanced Lane Detection ed", edges[270:480, :])
                    # whitecount = cv2.countNonZero(edges)

                    edgescropped = edges

                #print(time.time() - self.timeprev)

                alpha = .6
                beta = 1.
                gamma = 0

                # Colored = True makes the path show up on top of the colored image.
                colored = False
                if colored:
                    line_img_color = np.zeros(self.cv_image.shape,
                                              dtype=np.uint8)
                    midpointlist, leftlist, rightlist = self.draw_lines(
                        line_img_color, edgescropped)
                    LinesDrawn2_color = cv2.addWeighted(
                        self.cv_image, alpha, line_img_color, beta, gamma)
                else:
                    edges_color = cv2.cvtColor(edgescropped,
                                               cv2.COLOR_GRAY2RGB)
                    line_img_color = np.zeros(edges_color.shape,
                                              dtype=np.uint8)
                    midpointlist, leftlist, rightlist = self.draw_lines(
                        line_img_color, edges)
                    LinesDrawn2_color = cv2.addWeighted(
                        edges_color, alpha, line_img_color, beta, gamma)

                fontFace = cv2.FONT_HERSHEY_TRIPLEX
                self.publish_states(midpointlist, leftlist, rightlist)
                cv2.imshow("Advanced Lane Detection",
                           LinesDrawn2_color[270:480, :])

                if self.publish_image:
                    try:
                        self.reference_image_pub.publish(
                            self.bridge.cv2_to_imgmsg(LinesDrawn2_color,
                                                      "bgr8"))
                    except:  # CvBridgeError as e:
                        pass  #print(e)

                if (self.count > 100 and self.count % 50 == 0):
                    self.incount += 1
                    self.timenext = time.time()
                    self.timeElapsed = self.timenext - self.timeprev
                    self.total2 = self.total2 + self.timeElapsed
                    self.avg2 = self.total2 / (self.incount)
                    print('Average Time: ', self.avg2)

                # Waitkey is necesarry to update image
                cv2.waitKey(3)

                self.rate.sleep()
            except IOError, (ErrorNumber, ErrorMessage):
                print('HERE')
                print('HERE')
                print(ErrorMessage)
                pass
Пример #3
0
    def __init__(self):    
        
        self.vid = cv2.VideoCapture("/dev/video6")
        self.vid.set(12,5) #contrast
        self.vid.set(13,255) #saturation

        # Calibration Matrices
        self.mtx = np.array([[592.156892, 0.000000, 326.689246], [0.000000, 584.923917, 282.822026], [0.000000, 0.000000, 1.000000]])
        self.dist = np.array([-0.585868, 0.248490, -0.023236, -0.002907, 0.000000])

        # Camera resolution
        self.w = 640
        self.h = 480

        # Reference velocity
        self.v_ref = 2

        # Set node rate
        self.loop_rate   = 30
        self.ts          = 1.0 / self.loop_rate
        self.rate        = rospy.Rate(self.loop_rate)
        self.t0          = time.time()

        self.dt = self.ts
        self.count = 0
        self.total = 0 
        self.avg = 0 
        self.total2 = 0 
        self.avg2 = 0 
        self.publish_image = True;
        self.timeprev = time.time()-self.dt
        time.sleep(0.2)
        # Compute the udistortion and rectification transformation map
        self.newcameramtx, self.roi = cv2.getOptimalNewCameraMatrix(self.mtx,self.dist,(self.w,self.h),0,(self.w,self.h))
        self.mapx,self.mapy = cv2.initUndistortRectifyMap(self.mtx,self.dist,None,self.newcameramtx,(self.w,self.h),5)


        self.statepoints = ''
        self.printme = True

        self.state_constraints = barc_state()
        self.reference_trajectory = barc_state()

        # Initialize publishers and subscribers
        self.moving_pub = rospy.Publisher("moving", Moving, queue_size=1)
        self.moving_pub.publish(True)

        self.state_constraints_pub = rospy.Publisher("state_constraints", barc_state, queue_size = 1)
        self.reference_trajectory_pub = rospy.Publisher("reference_trajectory", barc_state, queue_size = 1)
        self.reference_image_pub = rospy.Publisher("image_reference_trajectory", Image, queue_size = 10)
        #self.optimal_state_sub = rospy.Subscriber("optimal_state_trajectory", barc_state, self.convertDistanceToPixels,queue_size=1)
        self.uOpt_pub = rospy.Publisher("uOpt", Input, queue_size=1)

        while not rospy.is_shutdown():
            self.count = self.count +1
            self.rel,self.dst = self.vid.read() # gets the current frame from the camera

            self.dt = time.time() - self.timeprev
            self.timeprev = time.time()

            self.cv_image = cv2.remap(self.dst,self.mapx,self.mapy,cv2.INTER_LINEAR) #Undistorts the fisheye image to rectangular
            self.x,self.y,self.w,self.h = self.roi
            self.dst = self.dst[self.y:self.y+self.h, self.x:self.x+self.w]

            # True makes the edge detection search for a yellow track using HSV. False will use grayscale and search for any edge regardless of color
            yellow = True
            kernel_size = 7
            if yellow:
                cropped = self.cv_image
                cropped[0:280,0:640] = 0
                hsv = cv2.cvtColor(cropped, cv2.COLOR_BGR2HSV)
                #########cv2.imshow('hsv',hsv[270:480,:])

                # define range of blue color in HSV
                lower_yellow = np.array([0,200,100])
                upper_yellow = np.array([50,255,255])

                # Threshold the HSV image to get only blue colors
                edges = cv2.inRange(hsv, lower_yellow, upper_yellow)
                edges = cv2.GaussianBlur(edges,(kernel_size,kernel_size),0)
                ####################cv2.imshow("Edges",edges[270:480,:])
                edgescropped = edges
            else:
                # Convert Color Image to Grayscale
                gray_image = cv2.cvtColor(self.cv_image, cv2.COLOR_BGR2GRAY)
                gray_image[0:270,0:640] = 0
                gray_image = cv2.GaussianBlur(gray_image, (kernel_size, kernel_size), 0)
                edges = cv2.Canny(gray_image,40,80)
                edgescropped = edges

            
            alpha = .6
            beta = 1.
            gamma = 0

            # Colored = True makes the path show up on top of the colored image. 
            colored = True
            if colored:
                line_img_color = np.zeros(self.cv_image.shape, dtype=np.uint8)
                midpointlist,leftlist,rightlist = self.draw_lines(line_img_color,edgescropped)
                LinesDrawn2_color = cv2.addWeighted(self.cv_image,alpha,line_img_color,beta,gamma)
            else: 
                edges_color = cv2.cvtColor(edgescropped, cv2.COLOR_GRAY2RGB)
                line_img_color = np.zeros(edges_color.shape, dtype=np.uint8)
                midpointlist,leftlist,rightlist = self.draw_lines(line_img_color,edges)
                LinesDrawn2_color = cv2.addWeighted(edges_color,alpha,line_img_color,beta,gamma)

            fontFace = cv2.FONT_HERSHEY_TRIPLEX
            self.publish_states(midpointlist,leftlist,rightlist)
            cv2.imshow("Advanced Lane Detection", LinesDrawn2_color[270:480,:])

            if self.publish_image:
                try:
                    self.reference_image_pub.publish(self.bridge.cv2_to_imgmsg(LinesDrawn2_color, "bgr8"))
                except:# CvBridgeError as e:
                    pass#print(e)

            self.timenext = time.time()
            self.timeElapsed = self.timenext - self.timeprev
            self.total2 = self.total2+self.timeElapsed
            self.avg2 = self.total2/self.count
            # Waitkey is necesarry to update image
            cv2.waitKey(3)
 
            self.rate.sleep()