• G. Satish
• Dr. M. Sailaja

Night-time animal movement on highways can create serious collision risks because reduced illumination makes animals difficult to detect at an early stage. This paper presents a deep learning-based low-light animal detection and collision prevention system for night-time highways. The proposed system uses YOLOv5 for detecting animals in low-light road scenes and extends the detection capability to six classes: Bear, Boar, Deer, Fly, Leopard, and Tiger. The trained detection model achieved a precision of 94.5%, a recall of 95.2%, and a mean Average Precision (mAP) of 96.6% at an Intersection over Union (IoU) threshold of 0.5. To support collision prevention, an application layer is incorporated after detection for the wildlife classes Bear, Boar, Deer, Leopard, and Tiger. Distance estimation is performed using the detected bounding-box dimensions and predefined real-world animal heights. Based on the estimated distance, a driver warning mechanism is used to indicate potentially close animals and support timely driver awareness. The proposed system therefore combines multi-class animal detection with distance estimation and driver warning for a practical night-time highway safety application.

Low-light animal detection, YOLOv5, deep learning, night-time highways, wildlife detection, distance estimation, driver warning, collision prevention.

"Deep Learning Technique-Based Low Light Animal Detection and Collision Prevention System for Night-time Highways", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.13, Issue 9, page no.c767-c773, September-2026, Available :http://www.jetir.org/papers/JETIR2609283.pdf

"Deep Learning Technique-Based Low Light Animal Detection and Collision Prevention System for Night-time Highways", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.13, Issue 9, page no. ppc767-c773, September-2026, Available at : http://www.jetir.org/papers/JETIR2609283.pdf