BT-IRVIF-Net: An Effective Fusion of Infrared and Visible Light Belt Tear Images

In the field of belt tear detection on belt conveyors, the current mainstream methods mostly rely on the recognition of damage through infrared image detection of the heat generated by a tear. However, this method has some limitations and drawbacks, especially when the tear size is small and the heat dissipates rapidly, limiting the detection effectiveness. To address this issue, this study proposes an unsupervised belt tear infrared and visible image fusion network based on RFN-Nest, named the Belt Tear Infrared Visible Image Fusion Network (BT-IRVIF-Net). This network effectively integrates the thermal targets in the infrared images with the detailed information in the visible light images to enhance the accuracy of belt tear detection on belt conveyors. Considering the particularity of the infrared images in the belt operation environment, we utilized Dynamic Data Enhancement (DDE) technology to process the infrared images, preserving the high-temperature areas by removing the background. Additionally, to address the shortcomings of the RFN-Nest network in the decoding stage, we introduced upsampling operations in the decoding part to enhance the image reconstruction capability. This study validates the high adaptability and accuracy of BT-IRVIF-Net in the field of belt tear detection on belt conveyors through comparative analysis with advanced models. Comprehensive experimental results indicate that the proposed method has potential significant practical implications and value in real-world applications.