Research On The Recognition Of Forest Fire Recognition Based On Video

Forest fire is the worldwide disaster, which has characteristics, such as it spreads fast, and it is difficult to put out the fire and save life and property. How to prevent forest fires, and achieving automatic monitoring of forest fires are the current important issues need to be resolved. The video-based forest fire monitoring technology is free from limitations of space and environment, which is suitable for the forest fire monitoring. It is a new and effective fire monitoring technology. Combined with digital image processing, computer vision, pattern recognition, artificial intelligence, and other technologies, this thesis carries out the in-depth analysis of the characteristics of the video images of forest fires, and designs an identification method of forest fires based on support vector machine. Finally, it designs and achieves a complete video-based forest fire monitoring system. The main contents and results of the thesis are as follows:Firstly, it introduces the main characteristics and phenomena of the development process of forest fires. Based on this, it discusses the characteristics of flame and smoke of forest fires. Through the study of these characteristics, it finds some rules and features of fire images, so as to provide the important basis for the monitoring and identification of fires.Secondly, it introduces the concept of image processing and the basic method of image preprocessing of forest fires. Aiming at the deficiencies of the traditional image median filtering algorithm and the shortcoming of low operational speed, it improves the traditional image median filtering algorithm and puts forward an improved median filtering algorithm. Moreover, it uses the global optimization algorithm to conduct the self-adaptive search of optimal window layout, and these algorithms can more accurately select the flame target in the processing of fire image features. It studies and discusses the smoke and fire identification method in the videos of forest fires, and gives a detailed analysis of distinctions of the flame and smoke of fires with colors, textures, shapes and dynamic features of other interference phenomena. It also proposes the corresponding video image recognition algorithm for each distinction.Finally, it introduces the concept of SVM. It attempts to carry out relevant experiment by taking the support vector machine as the carrier and the above results as the input, and combining with characteristics of fire images. Currently, these studies have no successful application instances. Experimental results show that support vector machine can better reduce the impacts of environmental disturbance on the fire image recognition, and ensure the reliability of identification with the flame recognition accuracy rate of 95%.The thesis has three aspects of innovation.(1)Through the experiment, this kind of forest fire identification method which uses image integration features can effectively identify forest fires(,2) the design of video-based forest fire monitoring system has better sensitivity and anti-jamming capability, so that the remote video monitoring system can rapidly and effectively identify forest fires and give alarms.(3)these studies have no successful application instances. Experimental results show that support vector machine can better reduce the impacts of environmental disturbance on the fire image recognition.