Plant Disease Detection Method Based on Computer Vision Technology

Abstract

Aiming at the basic problems that plague crop growth in traditional agriculture, a method of identifying weeds using machine vision and applying chemical agents with selective variables was proposed. Collected visible light images of plant diseases, pre-processed the images, segmented the images using an instruction value composed of R, G, and B color components as a threshold, and wrote an algorithm to misjudge the background image after segmentation as a background. Pixels were used for information recovery; according to the analysis of the change in color characteristics after the occurrence of lesions, sample lesions were extracted using the two color features of G / R and G / B; the results of the damage degree of diseased leaves measured using image processing technology The analysis was performed and compared with the results of the plant disease degree determined by the paper card method in the traditional classification standard. The experiments show that the selected 7 characteristic parameters are used as the input of the neural network, and the number of types of cucumber leaf diseases that need to be identified is used as the output to build a BP neural network model. By adjusting various parameters in the BP neural network, the parameters with the best recognition effect are selected to train the network. The trained network is used to identify the plant disease image. As a result, the disease can be identified well, and the recognition rate is 93.5%.

Plant phenotyping relevance

画像処理とニューラルネットワークにより植物病害の病斑・被害度を推定する手法が研究の中心であり、従来の葉病害評価法との比較検証も行っているため。

Code and data availability

The article describes cucumber/tomato leaf disease images collected from a self-run experimental farm and MATLAB-based BP neural network analysis, but provides no public dataset, image repository, code, or model deposit. No availability statement or authors' public URL for assets appears in any block; the only URLs are

No evidence-backed public reproduction asset is currently recorded.