Infrared Duplicate Video Cleaning Method Based on Improved YOLOv7
Yiwen Hu, Yutong Hu, Yong Zhang, Tongyu Liu, Hua Gong · 2024
Infrared video data is widely applied in military, transportation, security monitoring, and law enforcement. However, the equipment used for collecting and processing data often generates a significant amount of duplicate data that can negatively impact the quality of the infrared data. To address this issue and improve the quality of infrared data, we propose an improved YOLOv7 network in this paper. Our network, named AGT-ASPP-SimAM-YOLOv7, utilizes the adaptive grayscale transform (AGT) technique to enhance the contrast and brightness of infrared images. Additionally, we introduce the atrous spatial pyramid pooling (ASPP) module to fuse multi-scale feature information and the simple parameter-free attention module (SimAM) to dynamically assign feature weight. We also construct the SYLU infrared duplicate video dataset (SYLUIDVD), which includes complex backgrounds such as mountains, sky, and buildings. Experimental results show that the AGTASPP-SimAM-YOLOv7 network has the highest cleaning accuracy compared to VGG19, ResNet18, VGG16, and YOLOv7.