An internet of things field crop leaf disease detection method based on full convolution network

Abstract

The invention relates to an internet of things field crop leaf disease detection method based on the whole convolution network, which utilizes the internet of things to collect a plurality of images of the diseased leaves, labels the images, and then normalizes the scale of each image sample of the diseased leaves. The training set is expanded to 20 images by four preprocessing operations: translation, rotation, scaling and color shaking. All samples in the training sample set are averaged, and then each training sample in the training sample set is subtracted from the averaged value, and thenscrambled to form the averaged training set. The FCN is constructed by training the FCN with the samples from the training set after averaging. In the test, the full-size leaf image is used as the input to detect the disease on the FCN after training. The invention can learn the multi-level features from low to high, quickly realizes the disease detection with high precision, and is especially suitable for the leaf disease detection of crops based on the video leaf image of the Internet of Things.

Description

technical field [0001] The invention relates to the technical field of deep learning, in particular to a method for detecting leaf diseases of field crops in the Internet of Things based on a fully convolutional network. Background technique [0002] Crop diseases have seriously affected the yield, quality and sales of crops. Disease control is an important link in crop production management, and it is also an important expense. The premise of crop disease control is to detect the occurrence of diseases in time and identify the types of crop diseases. The occurrence of most diseases first causes symptoms on the leaves of crops. In fact, leaf symptoms are the main basis for crop disease detection. However, manual detection of crop leaf diseases is a time-consuming, expensive task that requires long-term training of professionals. The automation of crop leaf disease detection based on the Internet of Things is the basis for the development of intelligent agriculture. Detect...

AI Technical Summary

Problems solved by technology

Although this method achieves high detection accuracy, the disadvantages of this method are large storage overhead and low computational efficiency

Moreover, since the size of the pixel block is much smaller than the whole image, only some local features can be extracted, thus limiting the detection performance of the method.

Images