仿人视觉对抗A 自动驾驶.zip

import numpy as np import time import tensorflow as tf from robotPi import robotPi import cv2 import os from rev_cam import rev_cam # 图片翻转 width = 480 height = 180 channel = 1 inference_path = tf.Graph() filepath = os.getcwd() + '/model/auto_drive_model/old/-339' temp_image = np.zeros(width * height * channel, 'uint8') def auto_pilot(): cap = cv2.VideoCapture("/dev/video0") with tf.Session(graph=inference_path) as sess: init = tf.global_variables_initializer() sess.run(init) saver = tf.train.import_meta_graph(filepath + '.meta') saver.restore(sess, filepath) tf_X = sess.graph.get_tensor_by_name('input:0') pred = sess.graph.get_operation_by_name('pred') number = pred.outputs[0] prediction = tf.argmax(number, 1) frame_num = 0 zhong_dian = 0 robot.movement.start_hit() while cap.isOpened(): ret, frame = cap.read() frame = rev_cam(frame) resized_height = int(width * 0.75) # 计算缩放比例 frame = cv2.resize(frame, (width, resized_height)) frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) # 选取图像的下半部分 [480*0.75-180]*480（高*宽） res = frame[resized_height - height:, :] #cv2.imshow("frame", res) # 二值化 ret,thresh1 = cv2.threshold(res,95,255,cv2.THRESH_BINARY) #kernel = np.ones((12, 12), dtype=np.uint8) #closing = cv2.morphologyEx(thresh1, cv2.MORPH_CLOSE, kernel) cv2.imshow("thresh", thresh1) cv2.moveWindow("thresh",700,0) cv2.waitKey(1) frame = np.array(thresh1, dtype=np.float32) value = prediction.eval(feed_dict={tf_X: np.reshape(frame, [-1, height, width, channel])}) print('img_out:', value) if value == 0: print("forward") frame_num += 1 robot.movement.move_forward(speed=yby_speed,times=yby_delay) elif value == 1: print("left") frame_num += 1 if (time.time() - start) <= yby_time: robot.movement.turn_left(speed=5, times=300) #robot.movement.left_ward() else: robot.movement.turn_left(speed=10, times=50) time.sleep(0.2) elif value == 2: print("right") frame_num += 1 if (time.time() - start) <= yby_time: robot.movement.turn_right(speed=5, times=300) #robot.movement.right_ward() else: robot.movement.turn_right(speed=10, times=50) time.sleep(0.2) elif value == 3: frame_num += 1 if (time.time() - start) <= yby_time: robot.movement.move_forward(speed=yby_speed,times=yby_delay) else: zhong_dian = zhong_dian + 1 if(zhong_dian == 2): print("stop sign") print(frame_num,"Frame") robot.movement.move_forward(speed=30,times=350) time.sleep(0.5) cap.release() cv2.destroyAllWindows() elif value == 4: print("Banner forward") frame_num += 1 robot.movement.move_forward(speed=yby_speed,times=yby_delay) elif value == 5: print("Banner left") frame_num += 1 robot.movement.left_ward() #robot.movement.turn_left(speed=5, times=300) elif value == 6: print("Banner right") frame_num += 1 robot.movement.right_ward() #robot.movement.turn_right(speed=5, times=300) elif cv2.waitKey(1) & 0xFF ==ord('q'): cap.release() cv2.destroyAllWindows() break def ding_wei(): cap = cv2.VideoCapture("/dev/video1") time.sleep(0.5) while(1): ret, frame = cap.read() frame = rev_cam(frame) # 重设图像尺寸 frame_old = cv2.resize(frame, (480,360)) # RGB转HSV frame_hsv1 = cv2.cvtColor(frame_old,cv2.COLOR_BGR2HSV) frame_hsv2 = cv2.cvtColor(frame_hsv1,cv2.COLOR_BGR2HSV) # 红色范围 low_range = np.array([0, 0, 208]) high_range = np.array([177, 166, 255]) frame = cv2.inRange(frame_hsv2, low_range, high_range) # 图像腐蚀和扩张 kernel_1 = np.ones((5,5),np.uint8) kernel_2 = np.ones((12,12),np.uint8) frame = cv2.erode(frame,kernel_1) #腐蚀图像 frame = cv2.dilate(frame,kernel_2) # 查找符合的轮廓 _,contours, hierarchy = cv2.findContours(frame, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE) if len(contours) != 0: target = contours[0] (x, y), radius = cv2.minEnclosingCircle(target) print(int(x),int(radius)) # 左右调整 if x > 320: robot.movement.move_right(speed=10, times=50) elif x < 250: robot.movement.move_left(speed=10, times=50) else: robot.movement.move_forward(speed=10, times=100) if radius >= 46: robot.movement.start_hit() robot.movement.hit() time.sleep(1) robot.movement.start_hit() cv2.destroyAllWindows() cap.release() break cv2.namedWindow("old",cv2.WINDOW_NORMAL) #cv2.namedWindow("hsv1",cv2.WINDOW_NORMAL) #cv2.namedWindow("hsv2",cv2.WINDOW_NORMAL) cv2.namedWindow("Y",cv2.WINDOW_NORMAL) #调整窗口大小及位置 cv2.imshow("old", frame_old) cv2.moveWindow("old",700,0) #cv2.imshow("hsv1", frame_hsv1) #cv2.imshow("hsv2", frame_hsv2) cv2.imshow("Y", frame) cv2.moveWindow("Y",700,280) k = cv2.waitKey(1) if k & 0xFF ==ord('q'): cv2.destroyAllWindows() cap.release() break cv2.destroyAllWindows() cap.release() if __name__ == '__main__': yby_time = 40 yby_speed = 33 yby_delay = 155 robot = robotPi() start = time.time() auto_pilot() ding_wei() end = time.time() print(end-start,"S") #print(frame_number/(end-start),"f/s")