A method and system for intelligent disease identification based on CNN and transfer learning

Abstract

The present invention provides a disease intelligent identification method and system based on CNN and transfer learning, which can reduce the interference of picture background, can achieve higher identification accuracy with a limited number of samples, and support multi-classification of training samples with higher operating efficiency , the disease image recognition method includes the following steps: image preprocessing, normalizing the size of the image, and then using Faster-RCNN to quickly locate the disease area to remove background interference; image feature extraction, using the triplet similarity measurement model to extract image features, and then using SIFT features are used as compensation features for weighted fusion; for disease classification and recognition, the deep convolutional neural network is used to learn the first image features of normal plant images, and then transfer learning is used to learn the second image features of diseased plant images, and finally combined with the first image features Classify and identify with the second image feature.

Description

technical field [0001] The invention relates to the field of intelligent identification, in particular to a method and system for intelligent identification of diseases based on CNN and transfer learning. Background technique [0002] At present, there are many technologies that can be used for recognition in the field of computer vision, and the two most commonly used in the field of agriculture are the OpenCV vision algorithm library based on image morphology and the SVM support vector machine. [0003] The OpenCV vision algorithm starts with the color and shape features of the disease graph, extracts the disease features in the visible light range, and then classifies and processes the images according to the features to realize the diagnosis. This type of method has extremely high requirements on the quality of disease images, and different lighting and backgrounds will affect the recognition results. This method has poor recognition accuracy, weak anti-noise ability, an...

AI Technical Summary

Problems solved by technology

There are two disadvantages in this method: (1) This method uses quadratic programming to solve the support vector. When the number of samples is large, the solution process of the matrix will take up a lot of running memory and computing time, and the computing performance is poor.

(2) The classic support vector machine algorithm only gives the algorithm of two-class classification, but in the application of disease identification in the agricultural field, it is generally necessary to solve multi-class classification problems. The classic SVM algorithm needs to be combined with other algorithms to achieve multi-class classification. Classification problems, increasing model complexity and development costs

Images