| Nowadays, forest fires occur frequently all over the world, which not only causes directly damage to forest ecosystem, but also does great harms to our society and country. It urgently becomes a significant problem at present how to realize forest fire auto-recognition with monitoring systems. The thesis applies theories of digital image processing, computer vision, pattern recognition and artificial intelligence. And, thoroughly analyzing vision features about forest fire image, the algorithm for fire image recognition based on the BP neural network is realized. The aim is to design and realize the whole image auto-recognition system for forest fire. the main content and conclusions of this paper are listed as follows:(1) Algorithms for fire image processing are carefully researched. According to color information of color image, algorithm for flame color image segmentation based on color feature clustering in HSI color space is realized.(2) Image recognition algorithm for fire flame is ameliorated. the color and shape features of fire flame are extracted and integrated into flame feature vector. then, the algorithm for fire image recognition based on the BP neural network is realized.(3) Image auto-recognition system for forest fire is designed and realized. The design of hardware and software framework of the whole image auto-recognition system for forest fire is completed, and the software modules of the system are programmed using VC++6.0 development tool. Finally, the experiment result shows that the system remains with higher accuracy, lower rate of false alarm, quicker recognition speed and stonger anti-jamming ability, which has broad application prospects. |
PDF Full Text Request