多模板匹配性能

如何解决多模板匹配性能

对于网络钓鱼检测，我使用模板与 OpenCV 匹配将数千个徽标与潜在网络钓鱼页面的屏幕截图进行比较：

def doShot(file_path,ref_path,results):

colorread = 1

shot = cv2.imread(file_path,colorread)

scale_percent = 40 # percent of original size
width = int(shot.shape[1] * scale_percent / 100)
height = int(shot.shape[0] * scale_percent / 100)
dim = (width,height)

scaled_shot = cv2.resize(shot,dim,interpolation = cv2.INTER_AREA)

matches = []

dir_list = sorted(os.listdir(ref_path))
for f in dir_list:
        
    fpath = os.path.join(ref_path,f)
    
    template = cv2.imread(fpath,colorread)

    # Scale the template in varIoUs sizes
    for i in [0.9,1,1.1]:

        width = int(template.shape[1] * (scale_percent * i) / 100)
        height = int(template.shape[0] * (scale_percent * i) / 100)
        dim = (width,height)
    
        scaled_template = cv2.resize(template,interpolation = cv2.INTER_AREA)
        
        res = cv2.matchTemplate(scaled_shot,scaled_template,cv2.TM_CCOEFF_norMED)
        min_val,max_val,min_loc,max_loc1 = cv2.minMaxLoc(res)
        
        if max_val > 0.92:
            return True
            
return False

为了获得性能，我缩小了图像，但是比较所有徽标需要几分钟时间。如何让算法更快？