Abstract:

Rapid, timely and accurate detection of wheat diseases plays an important role in improving wheat yield. Three kinds of diseased leaves suffering from wheat powdery mildew, stripe rust and leaf rust respectively were taken as the research objects, and a recognition model of wheat leaf diseases based on LM neural network was proposed. Firstly, the K-means algorithm was used to segment the wheat leaf disease area. The color features and texture features of wheat leaf disease area were extracted to construct data sets. Then, the identification model of wheat leaf disease was constructed by LM neural network and the data for identification was input. The recognition rate of wheat leaf diseases based on color and texture features was 95.3%. In the case of small samples, LM neural network algorithm can be used to identify wheat diseased leaves quickly and accurately.

Key words: wheat diseased leaves, lesion segmentation, feature extraction, LM neural network