Hyphantria cunea larva net curtain image data set enhancement method

Abstract

The invention discloses a hyphantria cunea larva net curtain image data set enhancement method, which relates to the technical field of image processing, and comprises the following steps: (1) preprocessing a hyphantria cunea larva net curtain image collected in advance; (2) constructing a network framework based on the improved DCGAN, wherein the network framework mainly comprises the steps of constructing a classification task module and an adversarial task module, defining a DCGAN loss function and setting hyper-parameters in the network; (3) training the DCGAN network; (4) collecting and sorting the expanded data set obtained in the step (3), removing unqualified and repeated images, and selecting qualified images to obtain the expanded data set; and (5) verifying the reliability and quality of the expanded data set. The method can solve the problem that data sets obtained by manually shooting the fall webworm screen image are insufficient in number and category, ensures the diversity of the generated samples, and has the beneficial effect of high training speed.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for enhancing an image data set, in particular to a method for enhancing a screen image data set of American white moth larvae. Background technique [0002] The American white moth poses a serious threat to the growth of garden plants. The damage is mainly caused by larvae feeding on leaves, resulting in low stress resistance of trees, and even death of the whole plant in severe cases. The larval stage has obvious net screens, which is the best period for control. The common control method is to spray drugs on a large scale manually, but this method is inefficient and causes serious environmental pollution. An intelligent target spraying technology is urgently needed to achieve accurate automatic spraying. Medicine work. Accurate target recognition is the premise to achieve target spraying. In recent years, with the development of neural networks, many method...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the deficiencies in the prior art, the present invention provides a method for enhancing the screen image data set of white moth larvae, which can solve the problem of insufficient original data set, expand the data set and enhance the diversity of sample types, and aim at the Moth net screen images are not suitable for training. Corresponding improvements have been made to training set images and generative confrontation networks.

Images