An Improved Hybrid Model for Target Detection

Target Detection has entered into various practical implementations in various fields, including healthcare, military and defense, autonomous driving, pedestrian detection, remote sensing, and numerous others. In this paper, we primarily focus on advancing a refined model dedicated to detecting multiple objects. The proposed model achieves this by training fused images on the latest iteration of the YOLO(v5 and v8) and R-CNN models. In this paper, fusion was performed on thermal and visible images using three distinct methods: Multiscale Fusion, Channel-Based Fusion, and Blind Source Separation (BSS). These fusion techniques will be comprehensively elucidated in the subsequent sections of the paper, highlighting their respective contributions to the research. This approach aims to address object detection challenges in hazardous and time-critical environments. Extensive experimental analyses are conducted on the publicly accessible OSU Database. This research not only improves target detection but also elevates the overall understanding of intricate settings. By combining multiple image modalities, our model demonstrates a remarkable capability to comprehend complex environments, unlocking new target detection and recognition frontiers. The proposed scheme sets a new benchmark for reliability and resilience, making substantial strides toward advancing safety and security in critical applications such as autonomous driving, surveillance, and search and rescue operations. This research paves the way for a safer and more secure environment where advanced target detection systems playa pivotal role in ensuring the well-being of individuals and communities.