为了满足输电线路山火易发地区的低漏检、高精度、大范围、高时效性火点近实时监测需求,本文以地球同步轨道卫星影像为基础,提出了一种基于MC-CNN的山火检测算法.通过结合大津算法(OTSU)和上下文算法来增加潜在火点,从而在一定程度上降低火点检测的漏检率;引入PCA算法对输入特征进行优化,构建多通道网络结构,并利用联合概率和PSO参数寻优算法获取不同通道火点识别权重,在加权平均的基础上最终判定火点;同时,采用固定高温热源和太阳耀斑对虚假火点进行去除,以降低误报率.为了验证所提算法的有效性,本文随机选取了 2019年至2022年期间输电线路附近历史卫星监测山火案例,并利用已知火点样本对火点反演结果进行验证.计算结果显示,该算法的火点检测精度达到了 89.4%.
In order to meet the near real-time monitoring requirements for wildfires in high-risk areas along power transmission lines,which demand low false negatives,high accuracy,wide coverage,and high timeliness,this paper proposes a wildfire detection algorithm based on multi-channel convolutional neural network(MC-CNN)using geosynchronous orbit satellite imagery as the foundation.To some extent,the algorithm reduces the false negative rate of wildfire detection by combining the OTSU algorithm and context algorithm to enhance potential fire points.The principal component analysis(PCA)algorithm is introduced to optimize input features,constructing a multi-channel network structure.Weighted average of different channel fire point recognition weights is determined using a combination of probability and particle swarm optimization(PSO)parameter optimization algorithm,ultimately identifying the fire points.Furthermore,fixed high-temperature heat sources and solar flares are utilized to eliminate false fire points,reducing the false alarm rate.To validate the effectiveness of the proposed algorithm,historical satellite-monitored wildfire cases near power transmission lines between 2019 and 2022 are randomly selected.Known fire point samples are used to validate the results of fire point inversion.The calculation results show that the accuracy of this algorithm for wildfire detection reaches 89.4%.