With the deepening of machine vision research and increasingly mature application in the commercial field, researchers have begun to pay attention to the practical applications of various visual methods in the industrial field. In order to improve the real-time performance and low cost of the algorithm, an optimized multi-scale template matching method for detecting automotive floor solder joints is proposed based on traditional image processing and multi template matching algorithms. This article uses digital analog information to optimize distorted images, and uses the coordinates and quantity information of solder joints to obtain image distortion coefficients. In order to improve the accuracy of small object detection, the image is segmented into small blocks and enhanced.By optimizing and iterating multiple times to obtain accurate coordinate transformation relationships, the final detection results are filtered, improving the detection accuracy and effectiveness. This method can be better applied to industrial production, with practicality and feasibility.