Research On Agricultural Information Processing Based On Gene Expression Programming

Agricultural information processing is an essential link to realize agricultural informatization and agricultural modernization.It will be beneficial to optimize the allocation of resources and achieve maximum benefits,if the advanced and efficient information processing technology is introduced to various aspects of agricultural production and management.The agricultural information analysis technology and application,which are based on data mining and image processing,are developed rapidly nowadays.However,commonly used methods do not perform well on certain properties,and they are difficult to meet actual needs.It turns out to be a significant research to find a more suitable solution or to improve the existing methods.Gene expression programming is a kind of algorithm with good performance in many aspects,such as function discovery,clustering,parameter optimization,etc.It is consistent with the needs of the agricultural information processing.This paper focuses on its performance in agricultural information processing,and applies it to four practical problems,namely,crop growth modeling,seed variety clustering,low illumination image enhancement and leaf spot segmentation.Its principal purpose is to provide a new and more effective solution for the processing of agricultural information.The main work and results in this paper include:1)Crop model mining method based on gene expression programming.The main environmental factors affecting the growth of tomato were set as independent variables,and the model of carbon dioxide exchange rate was established and predicted.The proposed method can effectively construct the model and realize a more accurate prediction from the experiment results.Compared with the commonly used multiple regression and BP neural network modeling methods,it turns out that the function model excavated by the gene expression programming method not only has better prediction accuracy,but also has stronger robustness.2)Clustering algorithm based on weights and gene expression programming is proposed.The similarity between data samples was set via the proposed attribute weights,and the clustering steps were guided by gene expression programming.The results of wheat varieties and UCI data sets shows that the proposed method get better clustering results in terms of the accuracy than other common used clustering methods.3)Image global contrast enhancement method optimized by gene expression programming.In this method,gene expression programming was used to optimize parameters for the best results.The low-illumination plant images were collected as experimental objects.The proposed method was compared with enhancement methods based on the Retinex theory.The results show that the image processed by the global contrast enhancement algorithm optimized by gene expression programming is more natural in visual perception,and is superior to the contrast method in evaluating the image quality index.4)Image segmentation based on gene expression programming and spatial fuzzy clustering.Gene expression programming algorithm was used to find the best clustering center.Experiments on noisy images and tea lesion segmentation show that the proposed method has better noise immunity,higher segmentation accuracy.The proposed algorithm optimized the performance of fuzzy clustering image segmentation.