X-ray image target detection method, device and equipment

Abstract

The invention discloses an X-ray image target detection method, device and equipment, mainly based on a multi-classification and multi-label structure, by acquiring multiple candidate labels for undetermined frames in X-ray images, assigning multiple labels to undetermined frames, and based on The multiple candidate labels determine the category of the object in the frame to be determined, which means that the finally determined categories of the object in the frame to be determined are not limited to one or more types. The present invention can strengthen the regularization of the target detection tool, effectively increase the gain of the target detection tool, and then improve the target recognition rate. It has been verified that the present invention can improve the generalization ability of the target detection tool during specific implementation, and facilitate the rapid and accurate classification of contraband. While assisting security inspectors to work, the detection rate of contraband is greatly improved, and the focus of the present invention is on the classification task, which has nothing to do with the regression task, has little impact on the entire system, has improved independence, and is convenient for those skilled in the art to implement. Realize the invention.

Description

technical field [0001] The invention relates to the technical field of machine vision, in particular to a method, device and equipment for detecting an X-ray image target. Background technique [0002] The role of X-ray security inspection machines in security systems is crucial, especially in crowded traffic nodes such as subway stations, railway stations, and airports. Security inspection machines are an important security tool to ensure the safety of passengers' personal property and even the stability of the country and society. . The traditional security inspection method is that the security inspector observes the X-ray imaging of the luggage in the security inspection machine with naked eyes to identify whether there are contraband. But as we all know, X-ray images have nothing to do with the appearance of the item, but are related to the composition, density, angle and position of the item itself, which is quite different from the natural image of the item, and ther...

AI Technical Summary

Problems solved by technology

Although the natural image target detection technology can better identify a category for the foreground target (set a label for a clear candidate area), it obviously cannot solve the practical problems faced in the application of X-ray image target detection.

[0004] Therefore, when the target detection tool suitable for natural image processing is directly applied to X-ray image training, it may not be able to achieve equivalent results. Model tools have poor performance and obvious defects

Images