Road texture picture enhancement method coupling traditional method and WGAN-GP

Abstract

The invention discloses a road texture picture enhancement method coupling a traditional method and a WGAN-GP, and data is enhanced in combination with the traditional method and the WGAN-GP so as to achieve the effects of data amplification and balancing of various samples. The method comprises the following steps: firstly, converting road texture data collected by a commercial handheld three-dimensional laser scanner into two-dimensional visual texture pictures, and then classifying and preprocessing the road texture pictures and making a data set by using a manual method and a batch processing means. Based on an original small-sample-size and unbalanced pavement texture data set, a WGAN-GP network is adopted for further expansion on the basis of enhancement of a traditional method. The textural features of the newly generated picture are relatively more obvious and are easier to identify by a machine, so that the generalization ability of the model is improved, and the over-fitting phenomenon is reduced. In addition, the method not only can save labor and time cost, but also plays a crucial role in road condition analysis and automatic driving.

Description

technical field [0001] The invention belongs to the field of image processing, and relates to a road texture picture enhancement method coupled with a traditional method and WGAN-GP. The invention is applied to the enhancement of road texture picture data with small sample size and unbalanced distribution. Background technique [0002] Pavement is a very important civil infrastructure. Its structure and mechanical properties directly affect the comfort and safety of drivers and passengers. For this reason, road engineers have conducted in-depth research on its basic mechanism. Road texture is a very important indicator and plays a vital role in automatic driving. On the one hand, the road surface is in direct contact with vehicle tires, and the texture of the road surface surface will directly affect vehicle braking, tire noise, and vehicle bumps and vibrations; on the other hand, with the increase in the number of vehicles per capita, the road surface bears more and more l...

AI Technical Summary

Problems solved by technology

In addition, the present invention proposes a gradient penalty (gradient penalty), that is, by setting an additional gradient penalty item, the gradient of the discriminator does not exceed K, so that the weight distribution is uniform, and the learning ability of the neural network is fully utilized. To a certain extent, problems such as gradient disappearance, training difficulties, and model collapse have been solved.

Images