摘要
把控火灾的蔓延趋势是森林火灾预防的重要环节。目前针对火灾蔓延趋势的预测大多是关于蔓延模型对火线预测以及利用仿真软件对火线进行长时间离线模拟,但无论哪种预测方式都难以满足实际的火灾预防需求。卷积长短期神经网络(Convlstm)在短期的图像预测中表现出显著的优势,在此基础上加入灰狼优化算法(GWO)来优化神经网络模型的超参数,从而提高卷积长短期神经网络的整体预测效果。无人机由于机动性强,可以凭借搭载的传感器完成对真实火场信息的采集,无人机采集的红外图像数据可以通过透视变换来消除三维视觉上的偏差,最后将透视变换后的红外图像通过结合灰狼优化算法的卷积长短期神经网络完成对林火蔓延的实时预测。
Controlling the trend of fire spread is an important part of forest fire prevention.Most of the current predictions for fire spread trends are about spread models for fire line prediction and long time offline simulation of fire lines using simulation software,but either prediction method is difficult to meet the actual fire prevention needs.Convolutional long short-term memory neural network(Convlstm)shows significant advantages in short-term image prediction,based on which the gray wolf optimization algorithm(GWO)was added to optimize the hyperparameters of the neural network model,thus improving the overall prediction of convolutional long short-term memory neural network.The infrared image data collected by the UAV could eliminate 3D visual bias with a perspective transformation method,and the perspective-transformed infrared images were used tocomplete the real-time prediction of the spread of forest fires in through convolutional long and short-term neural network combined with the gray wolf optimization algorithm.
作者
王新权
李兴东
WANG Xin-quan;LI Xing-dong(School of Mechanical and Electrical Engineering,Northeast Forestry University,Harbin Heilongjiang 150040,China)
出处
《林业机械与木工设备》
2022年第5期34-40,共7页
Forestry Machinery & Woodworking Equipment
基金
中国自然科学基金(LH2020C042)
国家重点研发计划项目(2020YFC1511603)
中央高校基本科研业务费(2572019CP20)。
关键词
无人机
林火蔓延
透视变换
卷积长短期神经网络
灰狼优化算法
UAV
forest fire spread
perspective transformation
convolutional long and short-term neural network
gray wolf optimization algorithm
