基于计算机视觉的森林火灾识别算法设计

Design of Forest Fire Identification Algorithm Based on Computer Vision

为解决森林火灾监测系统中由于林火行为监测样本数据量大、维数多导致监测误报率高、实时性差等问题,提出一种基于计算机视觉的支持向量机算法进行森林火灾监测,提升识别精度,实现全天候林火自动监测预警。首先,获取图像进行预处理,并初步判别图像中是否存在烟火区域;然后,进行林火特征提取,训练样本生成特征向量,采用基于径向基核函数与多项式核函数的SVM算法进行烟火识别;最后,选取多功能森林防火机动巡查灭火装备为试验平台应用该算法进行试验验证。结果表明:所提出算法具有理想的森林火灾识别效果,识别准确率高达97.82%,并且可以与多功能森林防火机动巡查灭火装备通讯进行精确扑救,为森林防火装备智能化探索提出新思路。

In order to solve the problems of forest fire monitoring system,such as large amount of data, high dimension and poor real time, a support vector machine based on computer vision is proposed to monitor forest fire, improve the recognition accuracy and realize automatic monitoring and warning of all weather forest fire. First, the image is acquired for preprocessing, and it is initially determined whether there is smoke region or flame region in the image. Then, the characteristics of forest fire are extracted, and the sample is trained to generate feature vectors, and the SVM algorithm based on radial basis kernel function and polynomial kernel function is used for pyrotechnic identification. FinaUy,a multifunctional fore对fire-fighting equipment is selected as an experimental platform and the algorithm is used for experimental verification. The results show that the proposed algorithm has ideal forest fire identification effect, and the recognition accuracy rate is as high as 97.82%. It can carry out precise save and rescue with the multifunctional forest fire prevention mobile inspection fire-fighting equipment communication, and put forward new ideas for intelligent exploration of forest fire prevention equipment.