An innovative unsupervised gait recognition based tracking system for safeguarding large-scale nature reserves in complex terrain

Abstract

Abnormal human activities play a significant role in triggering emergencies within vast nature reserves. The vast area, complex terrain, and insufficient electricity and high-bandwidth network infrastructure present significant challenges in effectively supervising nature reserves. Fortunately, intricate terrains often boast restricted access points, typically confined to just a few narrow pathways and the gait recognition technique utilizes only a small amount of binary-processed low-quality gait data and seamlessly integrates with low-resolution and low-power-consumption cameras making it particularly suitable for human activities supervision in nature reserves. However, extensive existing supervised along with a limited number of unsupervised methods are unable to be implemented in real-world application due to the reliance on the pre-labeled training set and the insufficient retrieval accuracies. Here, we present an electronic tracking system for safeguarding large-scale nature reserves in complex terrain based on the unsupervised gait recognition technique for the first time. 1) The proposed method doesn't require any known matching relationships in the training set. 2) It consistently achieves 100% top-1 retrieval accuracies, with a distinct gap between the distances of top-1 and top-2 retrievals. This distinction allows us to detect abnormal behaviors, such as individuals who enter without exiting, exit without entering, or venture into restricted areas. It effectively mitigates the impact of human activities on the protected area at low cost offering an application case of gait recognition technology (GRT) in the field of nature conservation.