Bird and UAVs Recognition Detection and Tracking Based on Improved YOLOv9-DeepSORT

Abstract

At present, the protection of birds, especially endangered birds, faces major challenges. In the process of protection, birds are often mixed with various drones, and it is difficult to accurately count the number of endangered birds, especially at night, which brings great difficulties to bird protection work. So tracking and identifying birds and drones is essential to ensure the accuracy and efficiency of bird conservation efforts. To solve these problems, this paper proposes a new multi-target tracking (MOT) model based on the combination of YOLOv9 detection algorithm and DeepSORT tracking algorithm. Firstly, the original RepNSCPELAN4 module is replaced by CAM context feature enhancement module in Backbone to improve the model’s ability to extract small target features. Following this, the AFF channel attention mechanism has been integrated with RepNSCPELAN4 in the Head section to create the RepNSCPELAN4-AFF module, which aims to better address semantic and scale inconsistencies. Finally, a new RepNSCPELAN4-AKConv module has been developed using the AKConv dynamic Convolution module to replace the RepNSCPELAN4 module in the original Head section, enabling the model to more effectively capture detailed and contextual information. In the bird-UAV visible light comprehensive dataset proposed in this study, the mAP0.50 and F1 Score of all categories are 81.3% and 71.9% respectively by the improved YOLOv9-DeepSORT model. The mAP0.50 and F1 scores of individual birds are 89.1% and 82.4%, respectively. Compared to the Basic YOLOv9 model, the former improves by 7.9% and 5.3%, while the latter improves by 23.9% and 17.0%. On infrared datasets, compared to the original model, the mAP0.50 and F1 scores of the improved model improved by 3.2% and 1.4% across all categories compared to the original model. The average accuracy of identifying individual birds and similarly shaped fixed-wing drones also improved by 2.2% and 7.5% respectively. Moreover, on the mixed visible light and infrared data sets, the model get mAP0.50 of 81.8% higher 0.9% than that of the YOLOv9. These experiments demonstrate the improved YOLOv9-DeepSORT method can expand the multiscene application range of bird recognition and tracking models, effectively promoting the extraction of video frame features in multi-target tracking.