In the domain of vehicle surveillance under adverse weather conditions, this paper presents a groundbreaking adaptive model that transcends traditional object detection algorithms through strategic preprocessing and advanced feature integration. Leveraging CGANet for the effective removal of rain artifacts, our model ensures high-fidelity image input, crucial for accurate object detection. We introduce a sophisticated blend of convolutional down sampling and dense connection techniques to bolster feature extraction, while spatial pyramid pooling (SPP) is employed to accommodate variable input sizes, enhancing object recognition flexibility. To mitigate bounding box redundancy, our model incorporates an innovative non-maximum suppression mechanism, tailored for precision. A comprehensive evaluation against established benchmarks demonstrates our model’s superiority, achieving a significant uplift in average, and other critical metrics. This advancement not only sets a new standard for vehicle surveillance systems, especially in challenging environmental conditions but also paves the way for real-world applications demanding high accuracy and reliability in object detection.