Subway clamp appearance abnormity detection method

Abstract

The invention discloses a subway clamp appearance anomaly detection method. The method comprises the steps of: firstly, using high-speed cameras erected on the two sides of a train for collecting clamp images; positioning a rod piece part through a Faster RCNN, then detecting the rod piece part, and judging whether the rod piece is abnormal or not; enabling the Faster R-CNN to position and intercept the spring part in the whole clamp image, if the spring part is not intercepted, judging that the spring is lost, otherwise, enabling the intercepted spring image to be subjected to height comparison with the corresponding normal spring image, and judging whether the spring is partially lost or not; and if the spring does not have partial missing abnormality, sending the spring image to an OC-CNN network, and judging whether the spring has cracks or not. According to the method, the Faster R-CNN algorithm and the OC-CNN algorithm are used for carrying out anomaly diagnosis on the train clamp part, influences caused by weather, illumination and other reasons can be effectively avoided, and the robustness of the algorithm is improved.

Description

technical field [0001] The invention relates to the technical field of detection, in particular to a method for detecting abnormal appearance of subway clamps. Background technique [0002] In recent years, with the development of science and technology and the progress of the times, my country's railway transportation industry has ushered in rapid development, and the speed of train operation has reached an unprecedented level. When the train is running at high speed, any minor fault may cause a major accident, which makes the fault inspection and regular maintenance of the train's exterior parts particularly important. At present, the fault inspection of trains is mostly manual mode. On the one hand, manual inspection will consume a lot of manpower and material resources and affect the profit of the enterprise. The safe operation of the system poses serious risks. For these reasons, manual inspection can no longer meet the needs of the rapid development of the railway in...

AI Technical Summary

Problems solved by technology

(2) Machine learning: The traditional machine learning method includes two steps of feature extraction and classification. The feature extraction step needs to formulate different feature extraction methods in different application scenarios, and then perform a classification algorithm to judge the fault with the extracted features. , due to the diversity of feature extraction brought about by the complexity of the actual work scene, the accuracy and robustness of the machine learning model are not high in actual work.

(2) Reasoning time: Due to the complexity of the machine vision model itself and the huge amount of calculations, the model usually consumes more time to reason about a high-pixel image, which is obviously unacceptable for actual industrial scenarios

Images