System and method for detecting and describing key points of high-speed train image

Abstract

The invention relates to the technical field of image processing, and provides a high-speed train image key point detection and description system and method. The system for detecting and describing the key points of the high-speed train image comprises a trunk feature extraction module, an up-sampling module, a key point detector and a key point descriptor. According to the deep learning algorithm based on self-supervised training, a classic image registration process including key point detection, key point description, feature matching and geometric transformation estimation is adopted, and the defects that a registration algorithm based on gray scale is large in calculation amount and long in consumed time are overcome; in order to overcome the defect that a traditional feature-based image registration algorithm is not high in robustness for a more complex scene, multi-level features of a to-be-registered image are extracted in a network forward pass to perform joint key point detection and description, and robust key point descriptors are generated while the positioning precision of key points is improved to a great extent.

Description

technical field [0001] The present application relates to the technical field of image processing, and in particular, to a system and method for detecting and describing key points of high-speed train images. Background technique [0002] Image registration can be regarded as a basic task in the high-speed train detection process. For example, the train chassis is an important part of the connection between the train and the track, and its safety detection is an important link to ensure the safety of high-speed trains. [0003] At present, most of the detection schemes of high-speed trains in my country are improved and developed on the basis of the Trouble of moving EMU Detection System (TEDS). The method is to first use the computer to register the image collected by the line scan camera installed near the railway track with the stored historical image, and then use the automatic identification technology or manual identification of the fault or foreign object. If an abn...

AI Technical Summary

Problems solved by technology

Grayscale-based image registration can be regarded as an iterative optimization process that directly compares all content of two images. Although this method is simple to implement, it is computationally intensive and time-consuming, and it is difficult to meet the requirements of dynamic measurement in practice.

According to the characteristics of high-speed train images, many researchers have improved popular key point detection and description algorithms such as SIFT, SURF, and ORB. The feature-based image registration method only needs to complete the calculation of feature descriptors for key points, not for The entire image content, so the amount of calculation is small and the calculation speed is fast, but traditional feature extraction algorithms can only extract shallow features, and it is difficult to obtain deeper and more expressive features

For situations where the image appearance changes too much due to lighting, weather changes, etc., these methods have problems such as low registration accuracy and lack of generalization ability.

[0005] To sum up, in order to provide a method that can overcome the shortcomings of grayscale-based registration algorithms, which are computationally intensive and time-consuming, and the traditional feature-based image registration algorithms are not robust to more complex scenes. The high-speed train image registration method has become a technical problem to be solved urgently by those skilled in the art

Images