CLAHE-Based Low-Light Image Enhancement for Robust Object Detection in Overhead Power Transmission System
Zhikang Yuan, Jin Zeng, Zixiang Wei, Lijun Jin, Shengjie Zhao, Xianhui Liu, Yingyao Zhang, Gangjie Zhou
Abstract
The lighting quality of the unmanned aerial vehicle (UAV) images is the key factor that affects the intelligent operation and maintenance efficiency in power system. When captured in low-light condition, the UAV images suffer from loss of details, leading to a decrease in object detection accuracy. However, the research is lacking with respect to low-light image enhancement for UAV-based overhead power transmission system images. In this letter, we propose a low-light image enhancement method based on Contrast Limited Adaptive Histogram Equalization (CLAHE) to robustify object detection against low-light condition. We first separate the luminance and chrominance channels of the UAV image via color space conversion so that CLAHE is used to enhance luminance contrast without introducing chrominance shift. Moreover, we optimize the contrast limit parameter in CLAHE to balance contrast enhancement and noise suppression, which further boosts object detection accuracy. Experimental results validate the effectiveness of the proposed scheme in detail enhancement without noise over-amplification, leading to a magnificent increase of insulator detection accuracy from 58.0% to 82.0%.