Design Of Automated Intelligent Analysis System For Pest Images

In order to solve the problem of detecting and identifying insect pests in the food factory,the article has designed an intelligent analysis system for automated insect pests.The system has achieved real-time testing of pest targets,which can be widely used in agricultural production,food processing,and pest monitoring.Therefore,the design and realization of the intelligent analysis system of automated insect pests have important application significance.The article analyzes the actual use environment of the equipment,and has designed a dedicated automated insect image analysis system for ultraviolet insect catcher equipment.Based on STM32 and Raspberry Pi,the system has completed the design of wireless cameras and motor drive modules.On the basis of not changing the existing structure of the factory,the corresponding mechanical structure is designed to facilitate the installation and replacement of the equipment.Special software is designed on the computer to operate the system and visualize the results.The software can control the wireless camera structure in real time,identify the pest pictures taken by the camera and store information.Can view the pest condition through the software and analyze and predict the insect sentiment.The designed system was tested and the results were displayed to prove the effectiveness and robustness of the system.The article analyzes the characteristics of insect pest data sets,and waits for available datasets through image processing and data increase.A new convolutional neural network model is designed for this dataset.This model is based on the Faster-RCNN network structure,combining network structures such as residual networks,characteristic pyramid structures and hybrid attention mechanisms.Improve the network of the network to make full use of the characteristics of the image to improve the detection effect of convolution neural networks on insect images.The comparison experiments and ablation experiments were conducted on the proposed algorithm.The detection accuracy of the network after joining the algorithm was increased by about 5%,and the recall rate was increased by about 10%.Through the experiment,the improvement algorithm could effectively improve the network performance.In response to the existing agricultural and insect pests,the article built a pest data set,proposed a practical target detection network model,and designed a embedded system that can be practically used.The feasibility of the design is proved through the algorithm comparison experiment and the system testing experiment,indicating that the design system can complete the detection and identification of insect pests in real time,and the accuracy of the test meets the requirements.