Forest Fire Detection Algorithm Based On Digital Image Processing

As the largest terrestrial ecosystem, forest is abundant in natural resources, which can provide a variety of daily necessities for human beings, and create enormous economic benefits. More importantly, the forest is the basis of ecological diversity, and the family of a large number of animals and plants. In addition, the forest also have the function of reduce soil erosion, protect the environment, and the prevention of natural disasters, etc.. So, it’s often be called the "lungs of the earth". China’s total forest area ranked fifth in the world, but per capita consumption is only1/5of the world’s average. At the same time, the frequency of forest fires still further erosion of valuable forest resources. Therefore, we should prevent the forest fire effectively, at the same time early detection of forest fire and early warning will greatly reduce the losses caused by forest fires. In this paper, digital graphics technology on early forest fire is depth studied.Firstly, the paper summarizes the development status of various fire detection technology at home and abroad, and also the neural network’s, then image fire detection technology combined with neural networks is been put forward. The paper also describes the color spaces and their conversion and the video image acquisition, image grayscale technology based on RGB component. Then several filtering operators, and image enhancement method of histogram equalization are studied based on experimental results.Secondly, the paper explains the background difference method, and the image segmentation method for a single space domain. Focus on the poor law of the color component, threshold segmentation and border detection algorithm. Then, the paper introduces the morphological processing technology.Thirdly, the principle of flame characteristics extraction and the fire image recognition is studied in depth. Discusses the diffusion characteristics of the flame image area, which to be the fire image recognition pre-judgment conditions, can improve the efficiency of the system. Then, discrimination laws and detection algorithms for dynamic and static visual features of the flame image, is given in terms of color, texture, shape, etc.. At last, learning and training the flame image features as the input signal of BP neural network.The experiments show that the fire detection system combined with BP neural network can detects the forest fire effectively. This approach has a higher immunity and bright actual application prospect.