OpenCV错误位置关闭关键点

Question

I do not know what's wrong here, the points drawn from the function(cv2.drawMatchesKnn) are in a good position but points drawn from the loop are in a bad position. In the picture you can see it well. I know that there may be a small shift but it completely misses them.

if len(self.frame_kps) > 0:
        if self.t_matcher == "flann":
            matches = self.matcher.knnMatch(self.base_desc,self.frame_desc,k=2)

        goodMatch=[]
        good_without_list = []
        for m,n in matches:
            if(m.distance<matcher_dist*n.distance):
                goodMatch.append([m])
                good_without_list.append(n)

        if(len(goodMatch)>=self.MIN_MATCH):
            self.matches = cv2.drawMatchesKnn(self.base_gray,self.base_kps,gray,self.frame_kps,goodMatch,None,flags=2)

            for g in good_without_list:
                keyPoint = self.frame_kps[g.queryIdx]
                (x1,y1) = keyPoint.pt        
                cv2.circle(self.matches, (int(x1),int(y1)), 4, (255, 0, 0), 1)  

What is wrong here?

https://i.stack.imgur.com/FIziK.png

Answer 1

You will need to add the width of base_gray width to the x coordinates of frame_kps to get it right in the final display.

This is because your the draw matches combines 2 images together which changes the final position of the points in display.

 h, w = numpy.shape(self.base_gray)
 for g in goodMatch:
                keyPoint = self.frame_kps[g.trainIdx]
                (x1,y1) = keyPoint.pt   
                x1 = x1 + w
                cv2.circle(self.matches, (int(x1),int(y1)), 4, (255, 0, 0), 1)