Adaptive Gamma Correction and CLAHE for Low Light Conditions on Pothole Detection at Different Vehicle Speeds

Abstract

Potholes represent a significant issue in road infrastructure in Indonesia. It has the potential to cause vehicular damage and elevate the risk of traffic accidents. Detecting potholes at night faces significant challenges due to low light conditions that can adversely affect image quality. This study proposed image enhancement method utilizing Adaptive Gamma Correction (AGC) and Contrast Limited Adaptive Histogram Equalization (CLAHE) to address the low light issue, alongside YOLOv11 as the object detection algorithm. Experiments were conducted across four speed scenarios (30 km per hour, 40 km per hour, 50 km per hour, and 60 km per hour) to evaluate the efficacy of the proposed method. The results showed a significant improvement in detection performance with an average recall increase of approximately 0.1 at all speeds compared to the baseline YOLOv11 model. Furthermore, improvements were observed in precision and mAP at 50 values. At a speed of 30 km/h, precision reached 0.966 with mAP at 50 of 0.958. Upon increasing the speed to 40 km per hour, precision decreased to 0.949, and mAP at 50 of 0.925. At 50 km per hour, precision further declines to 0.934 and mAP at 50 of 0.882. At 60 km per hour, precision was reduced to 0.917 with an mAP at 50 of 0.831. Despite higher vehicle speeds leading to a reduction in object detection performance, integrating the image enhancement technique with YOLOv11 effectively mitigates the low light challenge for pothole detection.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint