python+opencv移动侦测(帧差法)

1.帧差法原理

移动侦测即是根据视频每帧或者几帧之间像素的差异，对差异值设置阈值，筛选大于阈值的像素点，做掩模图即可选出视频中存在变化的桢。帧差法较为简单的视频中物体移动侦测，帧差法分为：单帧差、两桢差、和三桢差。随着帧数的增加是防止检测结果的重影。

2.算法思路

文章以截取视频为例进行单帧差法移动侦测

3.python 实现代码

def threh(video,save_video,thres1,area_threh):
    cam = cv2.VideoCapture(video)#打开一个视频
    input_fps = cam.get(cv2.CAP_PROP_FPS)
    ret_val, input_image = cam.read()
    index=[]
    images=[]
    images.append(input_image)
    video_length = int(cam.get(cv2.CAP_PROP_FRAME_COUNT))
    input_image=cv2.resize(input_image,(512,512))
    ending_frame = video_length
    fourcc = cv2.VideoWriter_fourcc(*'XVID')
    out = cv2.VideoWriter(save_video,fourcc, input_fps, (512, 512))
    gray_lwpCV = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY)
    gray_lwpCV = cv2.GaussianBlur(gray_lwpCV, (21, 21), 0)
    background=gray_lwpCV

# es = cv2.getStructuringElement(cv2.MORPH_ELLIPSE, (9, 4))

    i = 0 # default is 0
    outt=[]
    while(cam.isOpened()) and ret_val == True and i <2999:
        ## if i % 2==1:
        ret_val, input_image = cam.read()
        input_image=cv2.resize(input_image,(512,512))
        gray_lwpCV = cv2.cvtColor(input_image, cv2.COLOR_BGR2GRAY)
        gray_lwpCV = cv2.GaussianBlur(gray_lwpCV, (21, 21), 0)
        diff = cv2.absdiff(background, gray_lwpCV)
        outt.append(diff)
        #跟着图像变换背景
        tem_diff=diff.flatten()
        tem_ds=pd.Series(tem_diff)
        tem_per=1-len(tem_ds[tem_ds==0])/len(tem_ds)
        if (tem_per <0.2 )| (tem_per>0.75):
            background=gray_lwpCV
        else:
            diff = cv2.threshold(diff, thres1, 255, cv2.THRESH_BINARY)[1]
            ret,thresh = cv2.threshold(diff.copy(),150,255,0)
            contours, hierarchy = cv2.findContours(thresh,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)
        #     contours, hierarchy = cv2.findContours(diff.copy(), cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
            for c in contours:
                if (cv2.contourArea(c) < area_threh) | (cv2.contourArea(c) >int(512*512*0.3) ) :      # 对于矩形区域，只显示大于给定阈值的轮廓（去除微小的变化等噪点）
                    continue
                (x, y, w, h) = cv2.boundingRect(c) # 该函数计算矩形的边界框
                cv2.rectangle(input_image, (x, y), (x+w, y+h), (0, 255, 0), 2) 
                index.append(i)
        #     cv2.imshow('contours', input_image)
        #     cv2.imshow('dis', diff)
        out.write(input_image)
        images.append(input_image)
        i = i+1
    out.release()
    cam.release()
    return outt,index,images```
##调取函数
outt=threh('new_video.mp4','test6.mp4',25,3000)