Image Database Construction And Intelligent Recognition Of Common Spiders In Rice Paddy Fields

The rapid identification and characterization of arthropod species is the basis of biodiversity investigation.The traditional classification method mainly relies on manual observation of the external morphological characteristics of specimens,comparison of each characteristic to the retrieval table one by one,and completing the species classification and identification.This kind of manual method is cumbersome and time-consuming,which is inefficient to investigate.In recent years,with the rapid development of artificial intelligence technology,the use of computers to complete the recognition and identification of arthropod species has attracted more and more attention.In this paper,common spiders in rice paddy fields were taken as the research object,and the construction of a image database of their morphological characteristics,the AutoML model training,and the development and application of intelligent recognition tools for spider species based on WeChat applets were conducted.The following main results are obtained:1.A total of 22 common spider species were collected from rice fields in Zhejiang Province.Their primary color pictures and classification characteristic pictures captured and collected using electronic microscope imaging system,etc.,with text description information,were uploaded to the cloud,and made into a image database of spiders in rice paddy fields.2.A total of 2,209 photographs of spiders in simple backgrounds were taken.599 pictures of 18 arthropods were collected through internet search engines.These data were then expanded by rotation method to obtain a total of 11,232 images,which were used as the image dataset for model training.After the uniform of image resolution,the enhancement of the model effect by high resolution and training set expansion is verified.The model was also improved by adding three different data enhancement operators,Cutout,Sharpness and Brightness,respectively.3.Using WeChat developer tools,the collected and processed primary color pictures and text descriptions of spiders in rice paddy fields as well as the trained model interface were integrated into the WeChat applets.Based on this WeChat applets,the intelligence system of rice paddy spiders was then built,including spider picture display,the introduction of morphological characteristics and life habit,the identification of rice paddy spiders,and other functions.After the program was completed,an experiential version was put online to carry out the identification of spider species in rice paddy fields.These results provide a quick and easy identification tool for the investigation of spider species diversity in rice paddy fields,as well as an aid for rapid and accurate indoor identification of spider species.