25_光流估计
作者:互联网
# 光流估计 # 1. 光流估计 import numpy as np import cv2 cap = cv2.VideoCapture('D:/pycharm/pycharm-cope/opencv/resource/videos/02_Foreground.avi') # 角点检测所需参数 # 如果不限制角点最大数量,速度就会有些慢,达不到实时的效果 # 品质因子会筛选角点,品质因子设置的越大,得到的角点越少 feature_params = dict(maxCorners=100, qualityLevel=0.3, minDistance=7) # lucas-kanada参数 # winSize:窗口大小 maxLevel:金字塔层数 lk_params = dict(winSize=(15, 15), maxLevel=2) # 随即颜色条 color = np.random.randint(0, 255, (100, 3)) # 拿到第一帧图像 ret, old_frame = cap.read() old_gray = cv2.cvtColor(old_frame, cv2.COLOR_BGR2GRAY) # cv2.goodFeaturesToTrack函数返回所有检测特征点,需要输入:图像,角点最大数量(效率),品质因子(特征值越大的越好来筛选) # 距离相当于这区间有比这个角点强的,就不要这个弱的了 # **变量 作为传入参数,是用来传入不定长的变量 p0 = cv2.goodFeaturesToTrack(old_gray, mask=None, **feature_params) # 拿到第一帧的角点,后面视频中是对第一帧的角点进行追踪 # 创建一个 mask mask = np.zeros_like(old_frame) while (True): ret, frame = cap.read() frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY) # 需要传入前一帧和当前图像以及前一帧检测到的角点 p1, st, err = cv2.calcOpticalFlowPyrLK(old_gray, frame_gray, p0, None, **lk_params) # st=1 表示 good_new = p1[st == 1] # st==1 表示找到的特征点,没找到的特征点就不要了 good_old = p0[st == 1] for i, (new, old) in enumerate(zip(good_new, good_old)): a, b = new.ravel() c, d = old.ravel() mask = cv2.line(mask, (int(a), int(b)), (int(c), int(d)), color[i].tolist(), 2) frame = cv2.circle(frame, (int(a), int(b)), 5, color[i].tolist(), -1) img = cv2.add(frame, mask) cv2.imshow('frame', img) k = cv2.waitKey(150) & 0xff if k == 27: break # 更新 old_gray = frame_gray.copy() p0 = good_new.reshape(-1, 1, 2) cv2.destroyAllWindows() cap.release()
标签:25,old,gray,int,frame,cv2,角点,估计,光流 来源: https://www.cnblogs.com/tuyin/p/16546406.html