Wheat drought recognition method based on image depth learning

Abstract

The invention relates to the technical field of wheat abiotic stress identification, in particular to a wheat drought identification method. A method for identifying the drought of wheat comprises thefollow steps of: a) carrying out a drought control experiment of wheat in a greenhouse by adopting a pot experiment method, wherein the drought grade are suitable, light drought, medium drought, severe drought and special drought; b) continuously acquiring wheat images of different drought grades by using a single-lens reflex camera, labeling the wheat images, and randomly dividing the data setsinto a training set and a test set; c) extracting image features using convolution operations of Inception_V3, Resnet_50, Resnet_152, which are pre-trained on ImageNet, and training the final classification layer; d) testing the trained depth learning network with the test set, and representing the accuracy of drought stress identification and classification with the accuracy of drought stress identification DSI and drought stress classification DSC respectively. As the wheat phenotypic characteristics are extracted by a depth learn method through acquiring a wheat digital image, the drought stress is identified and classified base on a disaster-bearing body, the drought stress is timely, accurate, lossless, and the wheat phenotypic characteristics are high in accuracy.

Description

technical field [0001] The invention relates to the technical field of wheat abiotic stress recognition, in particular to a wheat drought recognition method. Background technique [0002] Drought stress is the main factor affecting wheat yield. Drought in any growth stage will have a serious impact on wheat yield. Timely and accurate monitoring and early warning of drought stress in wheat and improving the accuracy of irrigation play an important role in ensuring high and stable yield of wheat. Although the traditional drought monitoring based on soil moisture sensors can reflect the soil moisture status to a certain extent, it has the characteristics of large error, small coverage, and difficulty in field application. Wheat plants have different phenotypic characteristics under different drought levels, such as changes in leaf color and texture, leaf area reduction, leaf curling, dryness and death, etc., by acquiring digital images based on wheat phenotypic characteristics ...

AI Technical Summary

Problems solved by technology

[0005] The technical problem to be solved by the present invention is: in order to solve the existing deficiencies of low drought monitoring accuracy, small coverage, time-consuming and laborious, etc., the present invention provides a wheat drought identification method, which uses a deep learning method to obtain wheat images through a single-lens reflex camera Extract wheat phenotypic characteristics to identify and classify different drought stresses of wheat in a timely, accurate and non-destructive manner, transfer and apply the pre-trained deep learning model to the wheat drought image data set, and non-destructively identify and classify wheat drought, This method uses phenotypic characteristics to identify and classify drought stress based on the hazard-bearing body, so it is timely, accurate, and non-destructive, and obtains a high accuracy rate

Images