The Detection Of Pests In Bean-pod Based On Machine Vision Technology

Peas,green beans,vegetable soybeans and other leguminous vegetable are much in demand these years,however,worms in leguminous vegetable are a problem for the quality and safety of leguminous vegetable.There has been much interest in depending on manual inspection to solve the problem in our country,but conventional inspection method by manual inspection is with high work load,low efficiency and easy to error check or lacking check.With the superiority of low cost,Machine version has been a well-established technology with numerous applications in appearance inspection of agricultural products.Based on the machine version technology,an efficient approach about inspection of worms in vegetable soybeans was developed.More details about these research conclusions as follows:(1)Three schemes,i.e.inspection scheme based on machine version,inspection scheme based on the conventional technology of transmission radiography and inspection scheme based on near infrared transmission technique,were compared with each other.Results show that inspection based on machine version is the first-rank.(2)Based on complementary metal oxide semiconductor(CMOS)industrial camera and halogen tungsten lamp,an image acquisition platform included scatter image acquisition and transmission image acquisition was developed.To enhance the accuracy and efficiency of the platform,four main parameters,i.e.transmission angle,material of micro slide,transmission distance and image acquisition environment,were worked out.(3)For the vegetable soybean containing worm hole,the key is to analyze its diffuse image,using G component of indicated value as a threshold to image segmentation,via the gray value range of 69 ~ 230 as the target area,combining with the analysis the difference of domain and area to get interested area,and establish Pod integrity-assisted judgment algorithm that based on local threshold segmentation to ensure the integrity of the interest area.Via the gray value range of 0 ~ 40 as the target area,combining with methods such as feature histogram analysis to finish the recognition and identification on the surface of a vegetable soybean.Model identification accuracy rate is 91%(4)For the vegetable soybean containing internal worm insect body,the key is to analyze its transmission image,adopting S component of indicated value as a threshold segmentation,and give a further study of the image segmentation,morphology processing and denoising filtering image preprocessing method,via a comparative analysis combining with the connected domain analysis and feature histogram analysis to complete the extraction of interested area.According to the vector characteristics of interesting region in transmission image in different types internal worm insect of vegetable soybean,respectively via the algorithms that contain area difference,roundness difference to finish the establishing of internal worm recognition and identification model with the identification accuracy rate of 72%(5)Using Visual Studio 2013 integrated development environment,we developed a software system for non-destructive testing of soybean internal worms,embedded an integrated identification model,with the identification accuracy rate of 84%,of pests inside soybean meal.By comprehensively identifying the results of the identification of the wormhole on the surface of the soybean meal and the internal body of the worm body,the final determination of whether the whole pod is a pest or not is made.According to the experimental results,the internal pest recognition platform built in this paper has a strong ability to detect vegetable soybeans and has a high recognition accuracy.The method and model for the identification of insect pests inside bean pods based on machine vision technology were preliminarily established,which provided reference for the subsequent development of a quick detection device for the internal pests of soybean meal.