Intelligent Classification Technology And Device Implementation Of Kidney Bean Seeds Based On Machine Visio

Snap bean was one of the common vegetables on the dining table.Improper harvesting of snap bean seeds could easily result in varietal mixture,and different varieties of snap beans required different cultivation and field management practices.Therefore,achieving varietal classification of snap bean seeds and improving their purity before planting was of great significance for enhancing planting efficiency and reducing workload.At that time,most of the classification and storage of snap bean seeds were carried out during cultivation and harvesting,with little substantial varietal classification.This approach easily resulted in varietal mixture.The few instances of varietal classification relied on manual visual identification,which was not only time-consuming and labor-intensive but also suffered from issues such as inconsistent standards and visual fatigue.Machine vision(MV),due to its high efficiency,accuracy,automation,multidimensional data analysis,and flexibility in detection,had been widely used in crop cultivation management,seed inspection,plant protection,drone agriculture,and food quality control.In particular,it had shown good results in the classification and identification of corn and rice seeds.This study utilized Machine Vision(MV)combined with Machine Learning(ML)techniques to investigate the classification method of snap bean seeds and designed an intelligent classification device for snap bean seeds based on this method.Firstly,this paper proposed the extraction of Region of Interest(ROI)using an improved K-Means algorithm based on Canny,which resulted in clear and distinct ROIs.The necessity of ROI extraction was also validated in subsequent analyses.Based on a detailed analysis of the fundamental principles of Convolutional Neural Networks(CNNs),a novel CNN called Snapbean-Net was developed for the varietal classification of snap bean seeds,incorporating convolutional operations and attention mechanisms.The research also compared the prediction speed and accuracy of Snapbean-Net with three basic CNNs,namely VGG-16,AlexNet,and ResNet18.The results showed that Snapbean-Net achieved a minimum prediction speed of 1.7 seconds and an identification accuracy of 98.91%,which exhibited significant improvements over traditional CNNs.Finally,the intelligent classification device for vegetable bean seeds was researched and developed.The device was divided into power,transportation,image acquisition,variety classification and device control system for detailed introduction and description,and the circuit design and component selection of the classification device were described and introduced in detail,and finally the device was tested,and the test results showed that the device is fast,stable,accurate classification,and can be used for intelligent classification of vegetable bean seeds.