Research And Implementation Of Soybean Leaf Parameters And Insect Moth Detection Based On Image Recognition

Leaves are the most important organs of plants,through which plants carry out photosynthesis,transpiration and nutrient transformation.The parameters of the leaves,such as area and circumference,can be used to analyze the growth status of plants,predict the growth model of plants,and provide scientific guidance and management for crop cultivation.Based on the demand of agricultural informatization,image processing and pattern recognition technology are more and more applied to the field of agricultural engineering.Compared with traditional information acquisition technology,image processing technology has good reproducibility,fast processing speed and higher accuracy.In this paper,soybean leaves are taken as the research object,and the collected images are segmented through the pre-collection of leaf images,and the edge detection algorithm is analyzed,and the appropriate image processing algorithm is selected to achieve the accurate segmentation of leaf contour;the shape,texture and HOG features of soybean leaves are extracted,on this basis,S VM classifier is selected to classify soybean leaves.The calculation method of the wormhole rate is discussed according to whether the edge of the wormhole is complete.Special software was developed to accomplish the core tasks of the project:the measurement of soybean leaf parameters and the determination of insect borer rate.The main research results of this paper include the following:(1)Soybean leaf sample collection and foreground segmentation.Three kinds of soybean leaf images were collected and preprocessed.OTSU segmentation algorithm was selected to segment the soybean leaf images and the middle leaf images were extracted by using the outer rectangle.Finally,Sobel operator was used to extract and detect edge contour.(2)Image feature extraction.The feature extraction and calculation of the segmented leaf images,including geometric features,shape features,invariant distance and texture features,were carried out.The HOG features were extracted and analyzed emphatically,which laid a foundation for the classification and recognition of soybean varieties.(3)Classification and identification of soybean leaves.Three extracted leaf HOG features were combined with SVM to classify and identify 300 soybean leaf images,and the recognition rate reached 94.67%,while the recognition rate of BP neural network was only 88.67%.The paper also verifies the selection of SVM kernel function.The experiment shows that the radial basis kernel function can achieve the best recognition effect,and the recognition rate is 95.33%.(4)Estimation of leaf parameters and insect moth rate.According to the blade parameters and the moth-eaten rate determination algorithm,30 leaves of three kinds(10 leaves of each kind)were selected for parameter determination,then 10 leaves with complete edge contour and 10 leaves with incomplete edge contour were selected for the moth-eaten rate determination.Three blades tests showed that the leaf length measurement accuracy of 96.4%on average,blade width measurement accuracy of 91.23%on average,circumference measurement accuracy of 92.44%,leaf area measurement accuracy of 96.13%,rate of blade at the edge of the complete bug eat by moth measuring the average relative error is 0.062,the edge victims to the remaining part on the edge of the blade measuring the average relative error is 0.589.Compared with the traditional measurement method,the measurement system in this paper has the characteristics of accuracy,simple operation and high efficiency.(5)This paper completes the development of geometric parameter determination and variety identification algorithm for soybean leaves.This software was developed on the Microsoft Visual Studio2013 platform,which used MFC package and called OpenCV library to shorten the development time.