Contrast Stretching and Contrast Limited Adaptive Histogram Equalization for Recognizing Vehicles Based on Yolo Models

Rapid urbanization has escalated the volume of vehicular traffic in urban corridors, presenting challenges such as increased congestion and the necessity for comprehensive vehicular data. This research addresses the deficit of ClosedCircuit Television (CCTV) datasets in Semarang by amassing visual recordings from municipal surveillance systems at traffic intersections, crafting a dataset that encapsulates diverse traffic scenarios across diurnal and nocturnal settings. Utilizing the YOLOv5 algorithm, this study quantitatively assesses traffic density in Semarang to facilitate the detection of vehicles under suboptimal lighting conditions. The research juxtaposes the efficacy of Contrast Stretching (CS) and Contrast Limited Adaptive Histogram Equalization (CLAHE) techniques on nocturnal imagery with baseline low-illumination environments. The outcomes corroborate the effectiveness of YOLOv5 in discerning and cataloging vehicular entities on thoroughfares, notwithstanding the challenges of image clarity. The dataset generated herein is bespoke to the traffic milieu of Semarang and substantiates the applicability of YOLOv5 in improving nocturnal vehicle detection. This advancement holds significant promise for the enhancement of traffic surveillance and vehicular management systems.