节点文献

图像通信与处理技术在林火监测中的应用研究

Study on Image Communication & Processing Technique in Forest Fire Monitoring

【作者】 沈海峰

【导师】 李东升; 盖志武;

【作者基本信息】 东北林业大学 , 机械设计及理论, 2003, 硕士

【摘要】 林火监测技术的发展越来越离不开图像通信与处理技术的支持,利用监测系统,如何实现远程实时监测和火灾自动监测是急需解决的问题。 本文在方案设计、硬件设计以及网络模型设计进行系统的分析与研究。为满足全方位监测要求,在森林区域安装多套客户机监测系统,对CCD拍摄的视频数据压缩成Mpeg-1格式,在Client/Sever网络模型基础上与服务器保持通信,实现远程传输视频。 本文为达到林火自动监测,探讨小波分析理论、神经元网络理论在图像处理中的应用。森林火灾图像经小波分解后,在小波域中予以去噪预处理,并借助阈值化图像分割技术、WSA与Kohonen神经元网络自动分割图像。最后提出最大后验概率(MAP)准则作为判决依据,借助感知器和BP神经元网络对图像予以识别,为防止森林火灾的发生、减轻森林火灾带来的危害奠定基础。

【Abstract】 Forest fire monitoring technique depends heavily on the provision of image communication & processing technique. It urgently becomes a significant problem how to realize real-time & long-distance monitoring and fire auto-recognition with monitoring systems.The paper takes a systematic research in project and network model design. For the sake of complete monitoring and alleviating system and network burdens, installing many Client systems with monitoring devices can communicate with Sever in forest area, and compressing video taken by CCD camera into Mpeg-1 format is adapt to internet with TCP/IP protocol.In research of forest fire auto-recognition, the paper discusses digital image processing with wavelet theory and NN. Image de-noise processes in wavelet domain by wavelet transform and automatically segments by Otsu method , WSA and Kohonen NNs. Then the paper brings forward MAP rule as recognition judgment criterion, and uses perception and BP NN to process image, it is very useful for preventing forest fire and alleviating harms to our society and country.

  • 【分类号】TN919.8
  • 【被引频次】21
  • 【下载频次】328
节点文献中: