Infrared and visible light image collaborative weak registration instance segmentation method

• Main Authors: ZHANG PENGYU, YU DABING, DAI XIN, LIU JIANGE, CAO LI, XU CHANG, LI QINGWU
• Format: Patent
• Language: Chinese; English
• Published: 19.04.2024
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; PHYSICS

The invention discloses an infrared and visible light image collaborative weak registration instance segmentation method, which comprises the following steps of: firstly, respectively sending a visible light-infrared image pair into a visible light image feature extraction module and an infrared image feature extraction module, and extracting multi-scale visible light image features with clear texture details and multi-level infrared image features with remarkable temperature information; then the multi-modal image features are sent into a multi-modal feature alignment module to correct infrared image features which may cause deviation; the visible light image features and the corrected infrared image features are sent to a multi-modal adaptive fusion module for full fusion; and finally, a fusion result is sent to a region candidate module and a feature screening module to generate a final high-precision instance segmentation result. According to the method, a multi-modal instance segmentation function can be