基于图像背景建模的电火花检测-《计算机技术与发展》

文章信息/Info

Author(s):: WANG Peng-fei; YANG Yu-wang; XIA Ji-an; School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094,China

Keywords:: electric leakage; electric spark; background generation; object detection; color feature

摘要:: 电器设备发生漏电会使设备工作受到影响,甚至引发火灾造成经济损失、威胁人身安全。电器漏电产生的电火花不同于一般火焰,其闪动时间较短、目标较小,导致普通传感器很难识别。为此,提出一种基于图像背景建模的电火花检测技术。该方法使用了均值、中值背景建模,Codebook 算法背景建模以及高斯混合模型建模这四种不同的背景建模算法,去除电火花检测时的背景干扰,提取前景并对前景区域进行大小、颜色特征的分割,从而识别电火花。实验结果表明,在简单背景环境下均值、中值背景建模方法更适合进行电火花检测,而复杂背景环境下 Codebook 算法和高斯混合模型建模的结果远优于前两种方法,且高斯混合模型的实验结果略优于 Codebook 算法。实际应用时应根据现场环境选择最合适的方法。

Abstract:: The occurrence of electric leakage will affect the work of equipment,even cause fire,which poses a threat to property security and life safety. Electric spark is quite different from ordinary flame. The brief flashes and small areas make it difficult for common sensor to recognize. To solve this problem,we present a method of electric spark detection based on background generation. Mean background modeling,median background modeling,codebook background modeling and MOG (mixture of Gaussian) background modeling are used in this method to remove background interference. Then the area and color feature in foreground regions is used to judge whether the foreground region is electric spark. Experiments show that mean background modeling and median background modeling are more suitable for detection of electric spark in simple environment,and Codebook background modeling and MOG background modeling are superior to the first two in complex environment. Moreover,MOG background modeling is slightly better than Codebook background modeling. Which method will be chosen is defined by practical application scenarios.