Due to the problem of automatic detection of smart meter wiring hole si ze,a circle detection method mainly based on multi-threshold method combined with improved randomized Hough transform is proposed in this paper.Firstly,after clipping, the color image containing all smart meter wiring holes is obtained in real-time.The binary image with hole features is obtained by threshold segmentation after image gray processing and filter proces sing.Based on the connected-domain labeling algorithm,the image containing multiple wiring holes can be accurately cut into a number of sub-images,and each sub-image contains at most one true wiring hole.Secondly,for each sub-image,we use the continuous thresholds in a given range to extract the Canny boundary,and extract the circle information by the improved randomized Hou gh transform.Finally,according to the peak size and position,we determine the ef fective cir cle and its radius.The experimental results show that the proposed method can ac curately extract the wiring holes with different background colors and different sizes of wiring holes.The average error of the c ircle center coordinates is not more than 2pixels,and the average error of the circle radius is not more than 2.5mm.