Human body security check image detection method and system based on improved YOLOv5s

Abstract

The invention provides a human body security check image detection method based on improved YOLOv5s. The human body security check image detection method comprises the following steps: collecting a real-time terahertz human body security check image; preprocessing the terahertz human body security check image by using a median filtering algorithm and a logarithmic nonlinear transformation algorithm; marking the preprocessed terahertz human body security check image, and acquiring a data set based on the marked terahertz human body security check image; an improved YOLOv5s network model is constructed; training the improved YOLOv5s network model based on the data set; and obtaining a terahertz human body security check image detection result based on the trained improved YOLOv5s network model. The method has good accuracy and real-time performance, and meets the requirement of a terahertz human body security check system for suspicious target detection.

Description

technical field [0001] The invention belongs to the technical field of terahertz imaging, and in particular relates to a human body security inspection image detection method and system based on improved YOLOv5s. Background technique [0002] Terahertz waves are electromagnetic waves with a frequency range between 0.1THz and 10THz between infrared rays and microwaves. Terahertz imaging technology can achieve higher imaging accuracy and is widely used in various fields such as biomedicine, wireless communication, environmental monitoring, and security inspection. In the field of security inspection, terahertz imaging systems have the unique advantage of being able to penetrate materials such as clothing, plastics, and paper, and detect materials such as metals, liquids, ceramics, and powders. They can be used for the visualization of hidden targets without involving privacy issues. Compared with traditional X-rays, the photon energy of terahertz radiation is low, and the pas...

AI Technical Summary

Problems solved by technology

At present, the application of deep learning algorithms to the recognition and detection tasks of terahertz human security images started relatively late, and most of them are target detection algorithms based on optical images. For example, Xiao H et al. proposed a layered and cropped Faster R-CNN Detection framework, and adding denoising and enhancement modules to improve the quality of terahertz images; Hou Bingji et al. proposed to use Online Hard Example Mining (Online Hard Example Mining) to optimize the loss function of Faster R-CNN to solve the problem of imbalance between positive and negative samples in terahertz images ; Pang L et al. used the YOLOv3 algorithm for real-time detection of human metal contraband in millimeter-wave images, and proposed a data expansion method to solve the problem of less image data

The above methods have greatly improved the performance of the detection algorithm, and have shown good results in the field of terahertz security detection. It is difficult to achieve a good balance between

Images