Novel Animal Detection System: Cascaded YOLOv8 With Adaptive Preprocessing and Feature Extraction

Leveraging advanced technologies, such as the cascaded YOLOv8-based approach, this research aims to detect wild animals, thereby preventing Wild animal intrusion in residential areas and sudden road crossings. A reliable wildlife animal detection system is essential for monitoring biodiversity, understanding animal behaviour, and supporting global conservation efforts. This paper uses datasets to introduce a cascaded YOLOv8-based approach for wildlife animal detection. Initially, the input dataset undergoes adaptive histogram equalisation for contrast enhancement, followed by super-pixel-based Fast Fuzzy C-Means (FCM) for segmentation. Features are then extracted using ResNet50, DarkNet19, and Local Binary Pattern, and finally, the optimal cascaded YOLOv8 detects the wildlife animals based on these features. The proposed MATLAB-based technique for detecting wildlife animals performs at its best, achieving 97% accuracy along with excellent metrics for kappa, precision, sensitivity, specificity, and F measures. This research contributes to advancing wildlife conservation efforts by providing a robust and efficient method for monitoring and preserving biodiversity. Future research endeavours may explore integrating advanced deep learning models and incorporating diverse datasets to refine further and enhance wildlife animal detection capabilities, ultimately facilitating more effective conservation strategies in natural ecosystems.