Terahertz human body security check image target detection model training data augmentation method

Abstract

The invention discloses a terahertz human body security check image target detection model training data augmentation method, and belongs to the technical field of image processing. The method comprises the following steps: in a data preparation stage, cutting a suspicious object region of interest image from target detection image annotation data to form a suspicious object image augmentation subset, and acquiring a human body foreground region by utilizing an image segmentation model or a manual labeling mode; in the training process, generating an area randomly in a human body foreground area, randomly selecting pictures from a suspicious object image augmentation subset to be covered, and forming augmentation training data. According to the method, a suspicious object image augmentation subset is cut from a small number of existing images with suspicious object labels, a segmentation algorithm or a manual labeling mode is utilized to extract a human body foreground area, a coverage area is randomly generated and suspicious object augmentation image coverage is randomly selected in the training process, and a data label is regenerated, so the training convergence process of the target detection model can be accelerated under the condition of unbalanced samples, and the model detection accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of terahertz human security inspection image processing, in particular to a training data augmentation method for a target detection model of a terahertz human security inspection image. Background technique [0002] Object detection can classify and locate objects of interest from images, which is a common task in computer vision, and deep learning is currently the main solution. [0003] The detection of suspicious objects in terahertz human security images belongs to a specific application field. The task goal is to find and locate suspicious objects hidden under human clothing through human security images. However, in the actual model training, there are scarce samples of suspicious objects, and the cost of data acquisition is high. The overall sample distribution is very unbalanced and other problems, which will lead to poor training effect of the target detection model and low detection accuracy. The...

AI Technical Summary

Problems solved by technology

[0004] The technical problem to be solved by the present invention is: how to overcome the problem of less label data and unbalanced samples of suspicious objects in terahertz security images, and provides a training data augmentation method for target detection models in terahertz human body security images. The proportion of the number of suspicious object tags, complete directional data augmentation, and improve the detection accuracy of the target detection model

Images