基于多算法协同的电力设备红外图像增强方法与性能验证

Multi-Algorithm Collaborative Infrared Image Enhancement Method for Power Equipment and Performance Validation

摘要: 电力设备红外热成像检测是预防性维护的核心手段，但原始红外图像常因低对比度、噪声干扰及缺陷区域与背景灰度差小，导致过热接头、老化绝缘子等关键缺陷难以精准识别。针对这一问题，本文提出一种多算法协同的红外图像增强框架，通过高斯滤波去噪预处理、全局直方图均衡化提升整体对比度、自适应直方图均衡化(CLAHE)优化局部细节的三步协同策略，实现噪声抑制与细节增强的平衡。实验结果表明：该框架使红外图像信息熵从6.2提升至7.8 (相对提升25.8%)，缺陷区域与背景的灰度差扩大2~3倍，信噪比(SNR)从28 dB提升至35 dB。本框架有效解决了单算法增强中噪声过度放大、局部细节丢失的问题，为电力设备缺陷自动检测提供了高质量的图像输入支撑。

Abstract: Infrared thermal imaging detection is a core means of preventive maintenance for power equipment. However, original infrared images often suffer from low contrast, noise interference, and small gray-level differences between defect regions and backgrounds, making it difficult to accurately identify key defects such as overheating joints and aging insulators. To address this issue, this paper proposes a multi-algorithm collaborative infrared image enhancement framework, which adopts a three-step collaborative strategy of Gaussian filtering for noise reduction preprocessing, global histogram equalization for overall contrast improvement, and Contrast Limited Adaptive Histogram Equalization (CLAHE) for local detail optimization, achieving a balance between noise suppression and detail enhancement. Experimental results show that the framework increases the information entropy of infrared images from 6.2 to 7.8 (a relative increase of 25.8%), expands the gray-level difference between defect regions and backgrounds by 2~3 times, and raises the signal-to-noise ratio (SNR) from 28 dB to 35 dB. This framework effectively solves the problems of excessive noise amplification and loss of local details in single-algorithm enhancement, providing high-quality image input for power equipment defect automatic detection.