基于Meanshift的鼠标交互式运动目标跟踪与检测

运行环境Win 10、python 3.6.3、Opencv 3.4.2

import numpy as np
import cv2

# 用于描述矩形对角线坐标(ix,iy)、(tx,ty)
ix,iy=-1,-1 # 初始坐标
tx,ty=-1,-1 # 终点坐标
switch_ok = False # 矩形ROI提取完成标志
drawing=False # 绘图标志
video_src=0 # 提取摄像头
#video_src='images/slow_traffic_small.mp4' # 提取视频
sTep1 = True

# 创建回调函数
def check_onmouse(event,x,y,flags,param):
    global ix,iy,drawing,img2,frame,tx,ty,switch_ok
    
    # 当按下左键是返回起始位置坐标
    if event==cv2.EVENT_LBUTTONDOWN: #左键点击事件
        drawing = True
        ix,iy=x,y
    # 当鼠标左键按下并移动是绘制图形。 event 可以查看移动， flag 查看是否按下
    elif event==cv2.EVENT_MOUSEMOVE and flags==cv2.EVENT_FLAG_LBUTTON:  #判断滑动与左键拖拽
        if drawing==True:
                img2 = frame
                cv2.rectangle(img2,(ix,iy),(x,y),(0,255,0),-1)
    # 当鼠标松开停止绘画。
    elif event==cv2.EVENT_LBUTTONUP:
        drawing = False
        switch_ok = True
        tx,ty = x,y
    #右键点击事件：重新绘制ROI    
    elif event==cv2.EVENT_FLAG_RBUTTON: 
        switch_ok = False
        sTep1 = True

cap = cv2.VideoCapture(video_src)          # 开启摄像头
cv2.namedWindow('camshift')                      # 创建名为 camshift的窗口
term_crit = ( cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 1 )
while(1): 
    if drawing == False:
        ret,frame = cap.read()                       # 读取视频
        img2 = frame
    cv2.setMouseCallback('camshift', check_onmouse)  # 在窗口上增加回调函数
    if switch_ok == True:   #关键点信息选择完毕
        if sTep1 == True:
            #截获鼠标点击ROI点
            roi = frame[ix:tx,iy:ty]
            r,h,c,w = ix,tx-ix,iy,ty-iy
            track_window = (c,r,w,h) 
            
            # 获取标记区域图像特征
            b,g,r,_=np.uint8(cv2.mean(roi)) #取掩模标记区域颜色的平均值，并显示出来
            bgd = np.zeros(roi.shape,dtype=np.uint8)
            bgd[:,:,:] = [b,g,r]
            hsv_roi_Type =  cv2.cvtColor(bgd, cv2.COLOR_BGR2HSV)
            '''
            if sTep1 == True:
            方案1：只能提取人皮肤图像信息
            #截获鼠标点击ROI点
            roi = frame[ix:tx,iy:ty]
            r,h,c,w = ix,tx-ix,iy,ty-iy
            track_window = (c,r,w,h)    
            hsv_roi =  cv2.cvtColor(roi, cv2.COLOR_BGR2HSV)
            mask = cv2.inRange(hsv_roi, np.array((0., 60.,32.)), np.array((180.,255.,255.)))
            
            roi_hist = cv2.calcHist([hsv_roi],[0],mask,[180],[0,180])
            #roi_hist = cv2.calcHist([hsv_roi],[0],mask,[20],[0,20])
            cv2.normalize(roi_hist,roi_hist,0,255,cv2.NORM_MINMAX)
            print('c,r,w,h = ',c,r,w,h)
            print('ix,iy,tx,ty = ',ix,iy,tx,ty)
            sTep1 = False
            '''            
            hsv_roi =  cv2.cvtColor(roi, cv2.COLOR_BGR2HSV)
            mask = cv2.inRange(hsv_roi, np.array((hsv_roi_Type[0][0][0]-20, 60.,32.)), np.array((hsv_roi_Type[0][0][0]+20.,255.,255.)))        
            roi_hist = cv2.calcHist([hsv_roi],[0],mask,[180],[0,180])
            cv2.normalize(roi_hist,roi_hist,0,255,cv2.NORM_MINMAX)
            print('所选取的关键框边角坐标 = (',ix,',',iy,')、(',tx,',',ty,')')
            sTep1 = False
                        
        hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
        dst = cv2.calcBackProject([hsv],[0],roi_hist,[0,180],1)
        # apply meanshift to get the new location        
        # 依据直方图反投影信息，得到新的图像密度中心，利用meanshift计算前后两个密度中心的相对关系 apply meanshift to get the new location
        ret, track_window = cv2.meanShift(dst, track_window, term_crit)
        #ret, track_window = cv2.CamShift(dst, track_window, term_crit)  ---- 法二 使用cv2.CamShift，可以随图像变化动态改变选择框
        # Draw it on image
        x,y,w,h = track_window #得到新采样窗口位置
        img2 = cv2.rectangle(frame, (x,y), (x+w,y+h), 255,2)
        
    cv2.imshow('camshift',img2)   
    
    k = cv2.waitKey(60) & 0xff
    if k == 27 :
        break
    else:
        cv2.imwrite(chr(k)+".jpg",img2)
            
cv2.destroyAllWindows()
cap.release()