Steel rail defect image generation method based on 3D model and point cloud processing

Abstract

The invention provides a steel rail defect image generation method based on a 3D model and point cloud processing. The method comprises the following steps: firstly, constructing a normal 3D model of the steel rail and initial 3D models of the defective steel rail; then, increasing the number of the initial 3D models of the defective steel rail and the types of the defects by a point cloud processing method; smoothing the transition difference between the defect and the background based on curvature and inverse curvature operation; finally, achieving automatic labeling by a label mapping method, while improving the diversity and complexity of the rail simulation data background by a texture replacement method. According to the method provided by the invention, infinite high-quality marked steel rail defect simulation data can be generated. The data can be used for later neural network training or auxiliary training, and the influence of insufficient data samples on steel rail defect detection tasks is effectively improved.

Description

technical field [0001] The invention relates to the technical field of data set generation, in particular to a method for generating rail defect images based on 3D models and point cloud processing. Background technique [0002] my country is a big country with railways, and the total length of railways has exceeded 131,000 kilometers. As the length of the railway increases, the maintenance of the condition of the railway becomes more and more important. In railway condition maintenance, the automatic location and identification of rail surface defects is an important part. [0003] As an important technical means for the service state detection of rail transit infrastructure, the rail surface defect detection technology based on deep learning technology trains the neural network model through a large amount of defect image data, so as to realize the identification and positioning of various defects on the rail surface . In actual scenarios, good maintenance makes the num...

AI Technical Summary

Problems solved by technology

[0005] (1) The training process of the generator requires the participation of a large amount of real data. If the real data is less, a high-quality simulation data set cannot be obtained;

[0006] (2) It cannot obtain its label information synchronously when generating data, and it needs to consume a lot of manpower for labeling work in the later stage;

[0007] (3) The generated defect types are random, and the generated results cannot be controlled

[0008] In addition to the above common defects, since the rail defect size information in the rail data set is relatively small compared to the scene where it is located, that is, it accounts for a small proportion in the entire image, it is difficult to use the generative adversarial network to generate virtual data. Defect information, so although the overall effect of the generated rail is more realistic, it lacks key defect information

Images