基于深度学习的列车监控系统的设计与实现

import os import random import cv2 as cv import numpy as np import pandas as pd BODY_PARTS = {"Nose": 0, "Neck": 1, "RShoulder": 2, "RElbow": 3, "RWrist": 4, "LShoulder": 5, "LElbow": 6, "LWrist": 7, "RHip": 8, "RKnee": 9, "RAnkle": 10, "LHip": 11, "LKnee": 12, "LAnkle": 13, "REye": 14, "LEye": 15, "REar": 16, "LEar": 17, "Background": 18} POSE_PAIRS = [["Neck", "RShoulder"], ["Neck", "LShoulder"], ["RShoulder", "RElbow"], ["RElbow", "RWrist"], ["LShoulder", "LElbow"], ["LElbow", "LWrist"], ["Neck", "RHip"], ["RHip", "RKnee"], ["RKnee", "RAnkle"], ["Neck", "LHip"], ["LHip", "LKnee"], ["LKnee", "LAnkle"], ["Neck", "Nose"], ["Nose", "REye"], ["REye", "REar"], ["Nose", "LEye"], ["LEye", "LEar"]] thr = 0.2 inWidth = 368 inHeight = 368 net = cv.dnn.readNetFromTensorflow("dat\\graph_opt.pb") cap = cv.VideoCapture(0) def get_points(frame): frameWidth = frame.shape[1] frameHeight = frame.shape[0] net.setInput(cv.dnn.blobFromImage(frame, 1.0, (inWidth, inHeight), (127.5, 127.5, 127.5), swapRB=True, crop=False)) out = net.forward() out = out[:, :19, :, :] # MobileNet output [1, 57, -1, -1], we only need the first 19 elements assert (len(BODY_PARTS) == out.shape[1]) points = [] for i in range(len(BODY_PARTS)): # Slice heatmap of corresponging body's part. heatMap = out[0, i, :, :] # Originally, we try to find all the local maximums. To simplify a sample # we just find a global one. However only a single pose at the same time # could be detected this way. _, conf, _, point = cv.minMaxLoc(heatMap) x = (frameWidth * point[0]) / out.shape[3] y = (frameHeight * point[1]) / out.shape[2] # Add a point if it's confidence is higher than threshold. points.append((int(x), int(y)) if conf > thr else None) return points def show_points(frame, points): for pair in POSE_PAIRS: partFrom = pair[0] partTo = pair[1] assert (partFrom in BODY_PARTS) assert (partTo in BODY_PARTS) idFrom = BODY_PARTS[partFrom] idTo = BODY_PARTS[partTo] if points[idFrom] and points[idTo]: cv.line(frame, points[idFrom], points[idTo], (0, 255, 0), 3) cv.ellipse(frame, points[idFrom], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED) cv.ellipse(frame, points[idTo], (3, 3), 0, 0, 360, (0, 0, 255), cv.FILLED) t, _ = net.getPerfProfile() freq = cv.getTickFrequency() / 1000 cv.putText(frame, '%.2fms' % (t / freq), (10, 20), cv.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 0)) cv.imshow('OpenPose using OpenCV', frame) # 获取前50帧有右手腕坐标的坐标并保存到csv文件中 def get_rwrist_scope_data(output): data50 = [] points = [] while cv.waitKey(1) < 0: hasFrame, frame = cap.read() if not hasFrame: cv.waitKey() break points = get_points(frame) assert (len(points) == 19) if points[BODY_PARTS["RWrist"]]: # if points[BODY_PARTS["Nose"]]: data50.append(points) show_points(frame, points) if len(data50) == 50: df = pd.DataFrame(data50) df.to_csv(output, index=False, header=True) break # 得到右手的起止范围 返回上、下、左、右的边界 def get_rwrist_scope(csv): def get_tuple(string): return (int(string[1:-1].split(',')[0]), int(string[1:-1].split(',')[1])) def get_rwrist_points(): df = pd.read_csv(csv) wrist_points = [] for i in range(df.shape[0]): # wrist_points.append(get_tuple(df.iloc[i, 0])) wrist_points.append(get_tuple(df.iloc[i, 4])) return np.mean(wrist_points, axis=0) center = get_rwrist_points() return center[1] + 20, center[1] - 20, center[0] - 20, center[0] + 20 def getData(path, record_num): u, d, l, r = get_rwrist_scope('RWistScopo.csv') # return:0-右手腕坐标为空，1-右手腕在起止范围内，2-右手腕不在起止范围内 def judge_in_bound(points): if points[BODY_PARTS["RWrist"]]: x = points[BODY_PARTS["RWrist"]][0] y = points[BODY_PARTS["RWrist"]][1] # if points[0]: # x = points[0][0] # y = points[0][1] if x >= l and x <= r and y >= d and y <= u: return 1 else: return 2 else: return 0 n = 0 frames = [] collecting = False while cv.waitKey(1) < 0: print(collecting, len(frames)) hasFrame, frame = cap.read() if not hasFrame: cv.waitKey() break points = get_points(frame) show_points(frame, points) judge = judge_in_bound(points) if not collecting and judge == 2: collecting = True frames.append(frame) if collecting and judge != 1: frames.append(frame) if collecting and judge == 1: if len(frames) < 15: frames = [] collecting = False else: resultList = random.sample(range(0, len(frames)), 15) mov = [] for i in range(len(resultList)): cv.imwrite(path + str(n) + '-' + str(i) + '.png', cv.resize(cv.cvtColor(frames[resultList[i]], cv.COLOR_BGR2GRAY), (128, 128))) n += 1 frames = [] collecting = False if n == record_num: break # get_rwrist_scope_data('RWistScopo.csv') # getData(os.getcwd() + "\\Data\\mov1\\", 2) # getData(os.getcwd() + "\\Data\\mov2\\", 2) # getData(os.getcwd() + "\\Data\\mov3\\", 2) # getData(os.getcwd() + "\\Data\\mov4\\", 2) # getData(os.getcwd() + "\\Data\\mov5\\", 2) # getData(os.getcwd() + "\\Data\\mov6\\", 2) # getData(os.getcwd() + "\\Data\\mov7\\", 2) # getData(os.getcwd() + "\\Data\\mov8\\", 2)