System Of Quantity Statistics And Size Measurement Of Shoe Upper Based On Machine Vision

In the shoe production process,quantity statistics and size measurement of shoe uppers are very important procedures,which still detect using traditional manual methods.This manual detection method has many problems such as slow speed,high error,and high cost.This subject takes the upper as the research object,and propose a machine vision based shoe quantity statistics and size measurement system with high detection speed and high stability.The main work is as follows:(1)The current research status of machine vision detection technology and shoe upper detection at home and abroad is analyzed.Based on the current statistical analysis of the quantity and size measurement of shoe uppers and their disadvantages in the footwear industry,the statistics and size of the shoe upper based on machine vision are designed,and the hardware part of the system is selected.(2)The template matching algorithm is used to locate the upper image and complete the statistics of the upper.For the gray-based and distance-based template matching algorithms can not adapt to the complex production,a template matching algorithm based on the edge direction is proposed.Firstly,selecting the Sobel operator with high stability and accurate positioning to extract the image edge direction;secondly,calculating the similarity measure,obtaining the best matching position,and adding the pyramid search strategy to reduce the complexity of the algorithm.The method can adapt to the changes of illumination,noise,and occlusion existing in the production environment of the shoe upper,and realizes the positioning of the shoe upper well,thereby completing the statistics of the number of shoe uppers.(3)The feature point matching algorithm is introduced into upper image matching,and a matching algorithm based on SURB combined with random sample consensus is proposed.Firstly,the SURF algorithm extracts the image feature points;secondly,the binary robust independent elementary features descriptor is used to describe it;finally,the random sample consensus algorithm is used to eliminate the mis-match points.Experimental results show that SURB and random sample consensus feature point matching algorithm can adapt to the image scale changes,lighting changes and noise interference.(4)The upper size is measured using the minimum enclosing rectangle algorithm.Firstly,the camera is calibrated to eliminate the effects of the lens distortion problem.At the same time,camera parameters are obtained and the relationship between image distance and physical distance is obtained;secondly,the length and width of the upper image are measured using the minimum enclosing rectangle algorithm,and the actual sizeof the upper is obtained through conversion of the proportional relationship;finally,the size is compared with the manual measurement result,which meet the actual production requirements.(5)Using the Python's own Tkinter graphics toolkit to write the system's human-computer interaction interface,combining with OpenCV image processing open source library to achieve the system's count and measurement functions,using the Sqlite database to save the data in the system,and finally organic integration,to achieve a fully functional,user-friendly system software.