图像识别
这个代码的功能是根据摄像头输入的图像，识别红色块，绿色块的坐标，并返回红绿色块的中心点的图像坐标（单位为像素）
当然，识别色可以自己更改，如蓝色，白色，黑色
运行平台：
硬件：奥比中光usb摄像头，30FPS
软件：pycharm,opencv库
识别绿色色块代码如下
识别绿色
def green_point_video(frame):kernel_2 = np.ones((2, 2), np.uint8) # 2x2的卷积核kernel_3 = np.ones((3, 3), np.uint8) # 3x3的卷积核kernel_4 = np.ones((4, 4), np.uint8) # 4x4的卷积核kernel_8 = np.ones((8, 8), np.uint8)Lower_red = np.array([156, 128, 46])Upper_red = np.array([180, 255, 255])red = [Lower_red, Upper_red, 'red']Lower_white = np.array([0, 0, 200]) # 221Upper_white = np.array([180, 30, 255])white = [Lower_white, Upper_white, 'white']Lower_black = np.array([0, 0, 0])Upper_black = np.array([180, 255, 46])black = [Lower_black, Upper_black, 'black']Lower_green = np.array([35, 95, 46])Upper_green = np.array([70, 255, 255])green = [Lower_green, Upper_green, 'green']Lower_blue = np.array([80, 80, 150])Upper_blue = np.array([110, 255, 255])blue = [Lower_blue, Upper_blue, 'blue']HSV = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV) # 把BGR图像转换为HSV格式color = greenif color[2] == 'green':mask_red = cv2.inRange(HSV, color[0], color[1])erosion = cv2.erode(mask_red, kernel_2, iterations=1)erosion = cv2.erode(erosion, kernel_4, iterations=1)erosion = cv2.erode(erosion, kernel_4, iterations=1)dilation = cv2.dilate(erosion, kernel_2, iterations=1)dilation = cv2.dilate(dilation, kernel_4, iterations=1)dilation = cv2.dilate(dilation, kernel_8, iterations=1)ret, binary = cv2.threshold(dilation, 127, 255, cv2.THRESH_BINARY) # 二值化，将滤波后的图像变成二值图像放在binary中cnts = cv2.findContours(binary.copy(), cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)cnts = imutils.grab_contours(cnts) # 寻找轮廓green_list = []for c in cnts:# compute the center of the contourM0 = cv2.moments(c)# cv2.drawContours(frame, [c], -1, (0, 255, 0), 2)car_video_location = [int(M0["m10"] / M0["m00"]), int(M0["m01"] / M0["m00"])]cv2.circle(frame, (car_video_location[0], car_video_location[1]), 7, (0, 255, 0), -1)cv2.putText(frame, "green", (car_video_location[0] - 20, car_video_location[1] - 20),cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 255, 0), 2)cv2.imshow("Image", frame)green_point = [int(M0["m10"] / M0["m00"]), int(M0["m01"] / M0["m00"])]green_list = green_list + green_point# if len(green_list) == 6: # 起始可能识别不到3个红色块，等都识别到，再输出坐标# print('红色块1', green_list[0], green_list[1])# print('红色块2', green_list[2], green_list[3])# print('红色块3', green_list[4], green_list[5])return green_list识别红色
def red_point_video(frame):kernel_2 = np.ones((2, 2), np.uint8) # 2x2的卷积核kernel_3 = np.ones((3, 3), np.uint8) # 3x3的卷积核kernel_4 = np.ones((4, 4), np.uint8) # 4x4的卷积核kernel_8 = np.ones((8, 8), np.uint8)Lower_red = np.array([156, 128, 46])Upper_red = np.array([180, 255, 255])red = [Lower_red, Upper_red, 'red']HSV = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV) # 把BGR图像转换为HSV格式color = redmask_red = cv2.inRange(HSV, color[0], color[1])erosion = cv2.erode(mask_red, kernel_2, iterations=1)erosion = cv2.erode(erosion, kernel_4, iterations=1)erosion = cv2.erode(erosion, kernel_4, iterations=1)dilation = cv2.dilate(erosion, kernel_2, iterations=1)dilation = cv2.dilate(dilation, kernel_4, iterations=1)dilation = cv2.dilate(dilation, kernel_8, iterations=1)ret, binary = cv2.threshold(dilation, 127, 255, cv2.THRESH_BINARY) # 二值化，将滤波后的图像变成二值图像放在binary中cnts = cv2.findContours(binary.copy(), cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)cnts = imutils.grab_contours(cnts) red_list = []for c in cnts:# compute the center of the contourM0_red = cv2.moments(c)
# cv2.drawContours(frame, [c], -1, (0, 255, 0), 2)car_video_location = [int(M0_red["m10"] / M0_red["m00"]), int(M0_red["m01"] / M0_red["m00"])]cv2.circle(frame, (car_video_location[0], car_video_location[1]), 7, (0, 0, 255), -1)cv2.putText(frame, "red", (car_video_location[0] - 20, car_video_location[1] - 20),cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)cv2.imshow("Image", frame)red_point = [int(M0_red["m10"] / M0_red["m00"]), int(M0_red["m01"] / M0_red["m00"])]# red_point = [int(M0["m10"] / M0["m00"]), int(M0["m01"] / M0["m00"])red_list = red_list + red_point# print('红色', red_list)# if len(red_list) == 6: # 起始可能识别不到3个红色块，等都识别到，再输出坐标# print('红色块1', red_list[0], red_list[1])# print('红色块2', red_list[2], red_list[3])# print('红色块3', red_list[4], red_list[5])return red_list
# 主函数：识别红色和绿色并在输出视频实时显示标记中心
import cv2
import numpy as np
import imutilsif __name__ == "__main__":cap = cv2.VideoCapture(0)while True:ret, sframe = cap.read()green_video_point = green_point_video(sframe) # 绿色点在视频中的坐标，并在输出标记green_video_point = red_point_video(sframe)cv2.imshow("Image", sframe)if cv2.waitKey(1) & 0xFF == ord('q'):breakcv2.waitKey(300) # 延时30ms运行效果
