Study On Vegetable Pest Counting Algorithm Based On Visual Perception

Traditionally,manual counting methods was carried out on the number of pests.But due to large labor costs,heavy workload,subjective and other shortcomings,using machine vision technology to monitor vegetable pest has become a hotspot.But the vast majority of current methods are to be carried out under the condition of ideal laboratory,which can't be directly applied to pest monitoring in the field.In order to study the distribution of pests in the field environment,this author studied a new algorithm for counting the southern vegetable pest using yellow sticky trap,which was supported by the National Spark Program Key Project "Southern Vegetable Major Pest Monitoring and Early Warning Technology Integration and Application Demonstration".Compared to the classical algorithms of edge detection and threshold segmentation,this paper proposes some new algorithms include pest image segmentation sub-algorithm based on the structure of random forest,feature extraction sub-algorithm of irregular structure,background removal sub-algorithm,interference target removal sub-algorithm and detection model counting sub-algorithm.Those sub-algorithms were integrated to create a vegetable pest count algorithm based on visual perception(VPCA-VP)to classify and count multiple pests.This paper also uses C++ framework of MFC to develop a pest counting software based on visual perception,providing a more user-friendly interface for users.The software can calculate the number of pests for a single pest image,it also accumulates the number of pests by multiple images.There are three main innovations in this paper:(1)using structured random forest pest image segmentation algorithm to avoid the interference about shadow and background;(2)Combining the image fill algorithm and the object selection algorithm to remove the background image;(3)using the feature extraction algorithm of irregular structure to calculate the eigenvector of pest image.In this paper,83 images were taken and analyzed.The experimental results showed that the accuracy rate of the vegetable pest counting algorithm based on visual perception(VPCA-VP)was 94.89% and the recall rate was 96.29%.Among them,the accuracy rate of Thrips was 93.19% and the recall rate was 94.50%.The accuracy rate of Whitefly was 91% and the recall rate was 90%.The accuracy rate of Fruit Fly was 100% and the recall rate was 92%.The algorithm can be better to remove the interference about shadow,background and other types of pests,so it is conducive to the field environment pest counting.The algorithm has good performance and has a wide application prospect in farmland monitoring.