Research on Image Damage Assessment Testing Method Based on Deep Learning

The key technologies of unmanned intelligent equipment include target detection technology, target decision-making technology, high-precision aiming control technology, and target damage assessment technology, among which target detection and damage assessment technology are the core. It is urgent to conduct research on target detection and damage assessment technology to provide the most direct data support for unmanned intelligent equipment to strike again or turn to fire through target detection and damage assessment. A method for image damage assessment that integrates attention mechanism and multi-scale features is proposed, and pixel level segmentation and quantitative analysis of target damage areas are achieved by improving the Mask R-CNN architecture. Experiments have shown that in explosive shock scenarios, this method improves mAP by 21.3% and reduces evaluation time to 0.8 seconds per frame compared to traditional methods, providing a new paradigm for dynamic damage assessment.