python+opencv目标跟踪

test1.py

import cv2

"""
背景和需要跟踪的物体差异很大
"""

# 获取视频
video = cv2.VideoCapture('../opencv/20210423_164452.mp4')

# 生成椭圆结构元素
es = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (9, 4))
# 设置背景帧
background = None

while True:
    # 读取视频每一帧
    ret, frame = video.read()
    print(ret)

    # 获取背景帧
    if background is None:
        # 将视频的第一帧图像转为灰度图
        background = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
        # 对灰度图进行高斯模糊,平滑图像
        background = cv2.GaussianBlur(background, (21, 21), 0)
        continue
    if ret:
        # 将视频的每一帧图像转为灰度图
        gray_frame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

        # 对灰度图进行高斯模糊,平滑图像
        gray_frame = cv2.GaussianBlur(gray_frame, (21, 21), 0)

        # 获取当前帧与背景帧之间的图像差异,得到差分图
        diff = cv2.absdiff(background, gray_frame)

        # 利用像素点值进行阈值分割,得到一副黑白图像
        diff = cv2.threshold(diff, 25, 255, cv2.THRESH_BINARY)[1]

        # 膨胀图像,减少错误
        diff = cv2.dilate(diff, es, iterations=2)

        # 得到图像中的目标轮廓
        cnts, hierarchy = cv2.findContours(diff.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

        for c in cnts:
            if cv2.contourArea(c) < 1500:
                continue
            # 绘制目标矩形框
            (x, y, w, h) = cv2.boundingRect(c)
            cv2.rectangle(frame, (x+2, y+2), (x+w, y+h), (0, 255, 0), 2)

        # 显示检测视频
        cv2.namedWindow('contours', 0)
        cv2.resizeWindow('contours', 600, 400)
        cv2.imshow('contours', frame)

        # 显示差异视频
        cv2.namedWindow('diff', 0)
        cv2.resizeWindow('diff', 600, 400)
        cv2.imshow('diff', diff)
        if cv2.waitKey(1) & 0xff == ord('q'):
            break
    else:
        break
# 结束
cv2.destroyAllWindows()
video.release()

test2.py

背景分割器

import cv2

# 获取视频
video = cv2.VideoCapture('../opencv/20210423_164452.mp4')
# KNN背景分割器,设置阴影检测
bs = cv2.createBackgroundSubtractorKNN(detectShadows=True)

while True:
    # 读取视频每一帧
    ret, frame = video.read()
    # 计算视频的前景掩码
    if ret:
        fgmask = bs.apply(frame)
        # 图像阈值化
        th = cv2.threshold(fgmask.copy(), 244, 255, cv2.THRESH_BINARY)[1]
        # 膨胀图像,减少错误
        dilated = cv2.dilate(th, cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (3, 3)), iterations=2)

        # 得到图像中的目标轮廓
        contours, hier = cv2.findContours(dilated, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

        for c in contours:
            if cv2.contourArea(c) > 1600:
                # 绘制目标矩形框
                (x, y, w, h) = cv2.boundingRect(c)
                cv2.rectangle(frame, (x, y), (x+w, y+h), (255, 255, 0), 2)

        # 显示差异视频
        cv2.imshow('mog', fgmask)
        # cv2.imshow('thresh', th)
        # 显示检测视频
        cv2.imshow('detection', frame)
        if cv2.waitKey(1) & 0xff == ord('q'):
            break
    else:
        break

video.release()
cv2.destroyAllWindows()