A Study On Real-time Monitoring System Based On Image Processing For Low-density Insect Infection Of Grain Storage

Grain production and stock in China maintain at a high level for many years.Although the loss rate of the storage of grain is low,the loss rate of grain is still high due to the large base of grain stock.Insect is one of the main causes for grain loss.It is important to be fully acquainted with the information of insects in granary before taking preventive measurements.Therefore,it has a great practical value to develop a real-time monitoring system for detecting infection of low-density insects during grain storage using image processing.It can help grain managers to take preventive measurements at the beginning stage of low-density insect infection.There are few researches on this kind of system.The existing monitoring system has some problems,such as the complexity of the system,the lack of mobility and pertinence.On the basis of the cloud service,web technology and the image processing,a real-time monitoring system for low-density insect infection situation of grain storage was established.Then,the system operation was carried out by the studies of insect trapping,image acquisition,image processing,insects counting,cloud server building and web application development.It improved the pertinence and mobility of the monitoring system and reduced the complexity of it.The research contents and results were showed as follows:(1)The framework design of the real-time monitoring system for low-density insect infection during grain storage: According to the performance and functional requirements of the monitoring system,consisting the framework by hardware,cloud server and web client.The function of the system was divided into four parts: image acquisition,image processing and insect counting,cloud server and web client.The hardware terminal was responsible for image acquisition,image processing and insect counting.The data was transferred from the hardware terminal to the cloud server by wireless network.Cloud server exchanged data with the hardware terminal and responded to the request of web client.It used cloud storage and database to storage date.Web client enabled users to supervise granary and setting parameters after logging in.(2)The design and implementation of hardware of real-time monitoring system for low-density infection in grain storage: The hardware of the system was composed of Raspberry Pi 3B+ and the second generation trap,which was used for insect trapping,image acquisition,image processing and insect counting.Through the optimization of the first generation trap in camera,dimension,the position of light and background color,the second generation could take high-quality images of clear insect outlines and uniform brightness without grain background.According to the framework of system,the Raspberry Pi 3B+ with dual-band wireless network card(2.4 GHz/5.0 GHz)and the highest CUP frequency was selected by comparing parameters of different types of Raspberry Pi to achieve goals of image acquisition and processing and data transmission.(3)Software design and implementation of real-time monitoring system for low-density infection in grain storage: Based on the hardware of the system,the software achieved the functions of image acquisition,image processing and insect counting,cloud server and web client.Raspberry Pi used Fswebcam and Python to capture images.Good visual images were acquired by adjusting the camera's parameters of brightness,contrast,gain,gamma and saturation.By calling the OpenCV function through Python,the effects of different grayscale,filtering and threshold segmentation algorithms were compared.The image processing flow consisted of clipping,G-component grayscale,median filtering,histogram double-peak threshold segmentation and closed operation.Insect counting was conducted using the eccentricity difference of insects,awns and grass seeds.The EC2,S3 and MongoDB were used to build cloud server,store.images and other data,respectively.JavaScript running environment was built by Node.js,and JavaScript was used as server-side scripting language.Web client was bulit by using Angular architecture.Through setting components,services and modules,combined HTML and CSS,the display and function of each page in web client was realized.The data exchange between web client and server was achieved by JSON.(4)Testing results and analysis of real-time monitoring system for low-densityinfection in grain storage: The monitoring effect of the system for low-density insect infection in grain storage was evaluated by sensitivity,trapping rate and counting accuracy.The longest time of trapping the first pest only accounted for 3.25% of the total monitoring time in the stored grain with different low density insect infestation concentration(46.9 min).The trapping rate at different density of red flour beetle was between61.98% and 71.53%,and the trapping rate at different density of insect was between 42.71%and 47.92%.There was a significant linear relationship between the number of insects trapped and the total number of insect in grain.The accuracy of image counting in the system reached by 92.48%.The accuracy of counting is higher when the number of insects was smaller.With the increase of the number of insects,the system counting accuracy decreased.The above results showed that the system had a high sensitivity at the low pest density pest,and there was no significant difference in trapping rates.The total number of insects in grain could be estimated by the number of insects in the trap,and the counting accuracy was high.