Spectral-Spatial Feature Enhancement Algorithm for Nighttime Object Detection and Tracking

Object detection and tracking has always been one of the important research directions in computer vision. The purpose is to determine whether object contained input image enclose with a bounding box. However, most methods are applied daytime objects, processing nighttime objects imprecise. In this paper, spectral-spatial feature enhancement algorithm for proposed, which inspired by symmetrical neural networks. proposed method consists following steps. First, preprocessing performed on unlabeled images, including low-light enhancement, detection, dynamic programming. Second, features times extracted modulated domain-adaptive structure. Third, Siamese network can make full use features, trained as tracker above images. Fourth, test set subjected then obtain final results. step includes Gabor filtering. spatial-spectral target fully step. NAT2021 dataset used experiments. Six employed comparisons. Multiple judgment indicators were analyze experimental results show that achieves excellent performance.