Adaptive Image Restoration and Object Detection and Tracking using Semantic Segmentation and Cross-Correlation based trackers for light-weight embedded devices

We introduce an innovative strategy for object detection and tracking based on semantic segmentation masks for detection and cross-correlation trackers specifically designed for light weight embedded-devices, enabling real-time operations even in environments with limited resources, including open-source robotics initiatives. An adaptive image enhancement algorithm has also been developed to tackle the dynamically changing underwater conditions. The outcomes highlight the efficacy of our tracking method, allowing Autonomous Underwater Vehicles (AUVs) to pursue objects swiftly on light-weight embedded devices, achieving comparatively high frame rates.