A garbage image automatic generation method for garbage classification

Abstract

The invention discloses a junk image automatic generation method for garbage classification, which comprises the following steps of: artificially collecting a plurality of junk images, and performingdata enhancement and normalization processing on the obtained images to obtain an original training sample set; Converting the training sample data of the original training sample set into a TFRecoreformat; Constructing a generation network by adopting a deep convolutional neural network; Constructing a discrimination network by adopting a sparse self-coding deep convolutional neural network; Training the generation network and the discrimination network in an alternative iterative optimization mode until the model converges; And establishing a garbage image generator to generate a garbage image. According to the invention, the balance parameters used for balancing the generation network and judging the network are introduced; According to the method, the learning rate is updated by usingthe balance parameters to achieve the convergence condition, a large number of high-fidelity junk image samples can be automatically generated only by collecting a small number of junk image samples,and the problems that the labor is high, the operability is poor and the cost is high in the manual image sample collection process are solved.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a method for automatically generating garbage images that can be used for garbage classification. Background technique [0002] Garbage classification is a scientific management method for effective disposal of garbage. Facing the ever-increasing waste production and deteriorating environmental conditions, how to maximize the utilization of waste resources, reduce the amount of waste disposal, and improve the quality of living environment through waste classification management is one of the urgent issues of common concern to all countries in the world. [0003] At present, garbage sorting is mainly done manually, which is labor-intensive and inefficient, and the types of garbage that can be sorted and processed are very limited; in fact, in the face of huge garbage output, only a very small amount has been effectively processed. With the increasing amount of garbage, how to real...

AI Technical Summary

Problems solved by technology

[0006] However, unlike ordinary objects, junk images are poorly standardized and contain complex scene information

The test shows that when the existing image generation technology is used to process garbage images, the image generation process is unstable and the generated image has a high degree of distortion

When generating pictures for training garbage classifiers, there are problems of low classification accuracy and high false detection rate, which cannot be applied in practice

Images