Red Fuji Apple Based On Computer Vision Technology Quality Inspection And Classification

With the continuous development of global trade,the circulation of commodities between countries is getting closer and closer.Among international trade commodities,apples are favored by people in many countries because of their rich nutrition and high beneficial elements to human body.Our country as the world's largest producer of apple,accounts for more than 50% of the total world output,is very important for economic growth in China's foreign trade exports,but due to the low level of China's apple quality,apple grading method still rely on artificial classified and with the aid of mechanical AIDS are classified according to the size and weight for apple,caused apple quality classification accuracy is not high,and in the grading process will cause serious damage to apple,at the same time rely on artificial classification will also be additional labor cost,make apple export prices increase,reduce the international market competitiveness.Taking Fuji Apple as the research object,this study designed and built a Fuji data acquisition system and collected data.2D images of Fuji apple from different angles were obtained.On the basis of image preprocessing,different algorithms were used to extract the characteristic parameters such as diameter,red proportion and defects.According to the national standard of Fuji grade,the decision tree algorithm was used to grade Fuji apples.Finally,the Fuji Apple automatic detection and grading system is designed and developed.The main research results and conclusions are as follows:(1)Build a hardware system for automatic image acquisition of Fuji Apple,set the rotation speed of the image acquisition rotation platform and the shooting time of the camera,and calibrate the Fuji Apple image acquisition camera.The camera is controlled by the lower machine controller to collect images of Fuji Apple every 120 degrees.(2)by comparing the artificial detection method and image feature extraction method of red Fuji apple fruit diameter size,proportion and the accuracy of the extraction of defect features,image feature extraction method of fruit diameter size detection accuracy is 96%,the maximum relative error is 3.94%,the detection accuracy of red accounted for 98%,the maximum error is 2.1%,for defect detection accuracy is 97%,the fault rate is below 5%,high precision automatic external morphological characteristics parameters extracted the apple.(3)According to the extraction results of the external morphological characteristic parameters of Fuji apple,the defect was taken as the first weight,the fruit diameter and color proportion as the second weight to construct the automatic grading model of Fuji apple.Through the quality detection and grading of 100 Fuji apples,the maximum error rate was 6% and the grading accuracy was 96%.(4)Integrating the Hongfuji Apple quality detection grading model with the hardware platform,an automated grading platform integrating the multi-dimensional feature image acquisition platform,automatic detection and grading platform and human-computer interaction platform of Hongfuji Apple was built.The single fruit grading rate was 1.5s,meeting the requirements of automatic grading level.