一种自适应的工业热异常检测方法

An Adaptive Industrial Thermal Anomaly Detection Method

工业作为影响城市热环境的重要机制之一,准确检测出引起热异常的工厂,可为分析工业热异常对城市热环境的贡献率提供准确信息,对科学规划工业建设、改善城市热环境具有重要意义。现有的工业热异常检测方法因固定临界值参数的局限性,易造成检测结果多提或漏提等问题。因此,本文提出一种自适应工业热异常检测方法(adaptive thermal anomaly detection,Adaptive-TAD)。该方法首先利用谷歌地球选取工厂训练样本;然后根据先验知识选取训练标准差倍数;最后基于二分法思想,结合工厂训练样本和标准差倍数训练样本,训练出最佳临界值,从而实现多时相多空间的工业热异常检测。实验结果表明:Adaptive-TAD方法能够高效、高精度地完成多空间多时相的工业热异常检测,且检测结果的正确率优于经典的3倍方法和应用广泛的1.645倍方法。

As one of the important mechanisms affecting the urban thermal environment,the accurate detection of the factory causing thermal anomaly can provide accurate information for the analysis of the contribution rate of industrial thermal anomaly to the urban thermal environment,which is of great significance to the scientific planning of industrial construction and the improvement of the urban thermal environment.The existing industrial thermal anomaly detection method is prone to cause the detection result to be raised or missed due to the limitation of the fixed threshold parameter.Therefore,an adaptive thermal anomaly detection method(Adaptive-TAD)was proposed.The method first uses Google Earth to select the factory training samples;then selects the training standard deviation multiple based on the prior knowledge;finally,based on the dichotomy idea,combined with the factory training samples and the standard deviation multiple training samples,the optimal threshold is trained to achieve the multi-temporal phase and multi-space industrial thermal anomaly detection.The experimental results show that the Adaptive-TAD method can perform multi-space and multi-temporal industrial thermal anomaly detection with high efficiency and high precision,and the accuracy of the detection results is better than the classical 3 times method and the widely used 1.645 times method.