python opencv sift 手动区域光流跟踪

import cv2 import numpy as np #特征点匹配 def FeatureMatch(des1,des2): ratio = 0.05 # K近邻算法求取在空间中距离最近的K个数据点，并将这些数据点归为一类 matcher = cv2.BFMatcher() raw_matches = matcher.knnMatch(des1, des2, k = 2) good_matches = [] for m1, m2 in raw_matches: # 如果最接近和次接近的比值大于一个既定的值，那么我们保留这个最接近的值，认为它和其匹配的点为good_match if m1.distance < ratio * m2.distance: good_matches.append(m1) return good_matches def OpticalFlow(): # 打开usb摄像头，或者读取视频文件，一样 cap = cv2.VideoCapture("1.mp4") # sift特征检测器 sift = cv2.SIFT_create(nfeatures=500, contrastThreshold=0.1, edgeThreshold=50, sigma=1.0) lk_params = dict(winSize=(13, 13), # lucas kanade参数：搜索窗口、金字塔层数 maxLevel=3, criteria = (cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 0.03)) color = np.random.randint(0, 255, (1000, 3)) # 随机颜色条 # 拿到第一帧图像并灰度化作为前一帧图片 ret, old_frame = cap.read() old_gray = cv2.cvtColor(old_frame, cv2.COLOR_BGR2GRAY) # 利用opencv进行提取roi区域 x,y,w,h = cv2.selectROI("roi",old_frame) roi = old_frame[y:y+h,x:x+w].copy() # 查找roi区域特征 kp1, des1 = sift.detectAndCompute(roi, None) # 检测全局特征点 kp3, des3 = sift.detectAndCompute(old_frame, None) # 特征匹配 good_matches = FeatureMatch(des1,des3) # 解析匹配点 p0 = [] for mc in good_matches: p0.append( [[kp3[mc.trainIdx].pt[0], kp3[mc.trainIdx].pt[1]]] ) p0 = np.array(p0).astype(np.float32) # 创建一个mask, 用于进行横线的绘制 mask = np.zeros_like(old_frame) # 保存视频 fourcc = cv2.VideoWriter_fourcc(*"mp4v") out = cv2.VideoWriter('output.mp4',fourcc,20.0,(old_frame.shape[1] ,old_frame.shape[0]),True) # 黑白为false，彩色为True while True: # 读取图片灰度化作为后一张图片的输入 ret, frame = cap.read() if ret == False: break frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) # 进行金字塔LK光流检测需要输入前一帧和当前图像及前一帧检测到的特征点 pl, st, err = cv2.calcOpticalFlowPyrLK(old_gray, frame_gray, p0, None, **lk_params) if len(pl[st == 1]) <= 1: break # 清除错误率太高的点 if max(err)>10: plb = [] for p,e in zip(pl,err): if e<10: plb.append(p) pl = np.array(plb) if len(pl) <= 1: break # 读取运动了的角点st == 1表示检测到的运动物体，即v和u表示为0 good_new = pl # 绘制轨迹 for i, (new, old) in enumerate(zip(good_new, p0)): a, b = new.ravel().astype(int) c, d = old.ravel().astype(int) mask = cv2.line(mask, (a, b), (c, d), color[0].tolist(), 2) frame = cv2.circle(frame, (a, b), 5, color[1].tolist(), -1) #质心 center = np.mean(p0, axis=0).astype(np.int32)[0] frame = cv2.circle(frame, (center[0], center[1]), 5, [255,0,0], -1) # 外接矩形 x,y,w,h = cv2.boundingRect(p0) # 画出运动物体的轮廓 frame = cv2.rectangle(frame, (x, y), (x + w, y + h), [0,0,255], 2) # 叠加mask img = cv2.add(frame, mask) cv2.imshow('frame', img) out.write(img) k = cv2.waitKey(50) & 0xff if k == 27: break # 第十步：更新前一帧图片和角点的位置 old_gray = frame_gray.copy() p0 = good_new.reshape(-1, 1, 2) cv2.destroyAllWindows() cap.release() out.release() if __name__ == '__main__': OpticalFlow()