Research On Forest Fire Image Stabilization And Recognition Positioning Algorithm

With the rapid development of information technology and digital technology, forest fire prevention techniques have greatly improved. Analog monitoring equipment was gradually replaced by digital monitoring equipment. Some development of techniques, such as wireless spread spectrum communication technology, broadband network video surveillance technology, GIS (Geographic Information Systems) and digital PTZ technology, have laid the foundation for automatic identification and location of forest fires. Forest fire identification and positioning technologies is moving in the direction of intelligent when Traditional forestry was entering into the era of digital forestry and smart forestry. Some theoretical basis is provided to locate forest fires through applications of wireless sensor network, machine vision, networking and so on. Combined with video surveillance, cruise shooting and satellite remote, the new monitoring strategie can be all-round monitoring of forest fires.According to national nonprofit industry-specific project "Monitoring technology based on Things and its application in forestry"(201104037), the research topics of this paper was determined. In this paper. Algorithms of forest fires image stabilization, identification and localization based on TRIZ theory and gray projection was Object of this study. Then these algorithms will be applied to forestry video surveillance systems. The main contents and innovations of this paper included the following aspects:(1) Through the understanding of the subject content, necessary theory and algorithms are selected to study and analyze. TRIZ theory and its major systems were understood, and through simulation and experiment demonstrated the algorithm, such as Based on gray projection algorithm for electronic image stabilization, BP neural network algorithm and Edge detection algorithm. In this process, a key step in the algorithm is realized, so that ensure the algorithm could be utilized and improved in the latter study.(2) It will affect the trend of gray projection when moving objects appear in the image. For this reason, presented a gray projection image stabilization algorithm based on cutting and compensation. First, the problem was analyzed by functional component analysis on TRIZ. Innovative ideas come through the principle of technical contradiction and innovation principle, and find a suitable solution according to the principles of innovation. We used image object movement area retrieval algorithm to determine moving object range of pixels and separated images based on the results obtained. Extract corresponding portions of the original image compensation to separated images, in order to achieve no moving object effect relative to the original. This will remove the impact of moving objects on Gray Projection. Experiments show that this method could be a good solution to "Wrong stable" problem in gray projection image stabilization algorithm.(3) After the research and improvement gray projection image stabilization algorithm, we applied the algorithm to the forest fire monitoring system, and found that there was still "wrong steady" phenomenon. At this point we still used the analytical tools of TRIZ to analyz the problem, and by solving tool to find the appropriate innovation principles. Select the BP neural network and HSI color model according the principle of innovation for image classification. According to type of images reassigned, as this way, gray projection curve can reflect the main content of the image. Experiments show that this method had better treatment effect and improved the accuracy of image stabilization algorithm.(4) On the basis of the stabilization transmission of the forest fire image, proposed a forest fire smoke intelligent recognition algorithm based on bit-plane. Firstly, the problem is transformed into the image display problems with the use of the physical contradiction separation principle of TRIZ. Secondly, further analysis of the problem based on the principles of innovation, innovative ideas of classification of image detail is given. Accordance with the the concept of gray bit plane, the images were separated. The results show that there were relationship between the smoke and the background on different levels, real smoke was successfully simulated. Experiments show that this method is faster than conventional HSI model algorithm.(5) On the basis of stable transmission image and recognition of forest fire smoke, forest fire positioning algorithm based on monocular video was studied. The Problem what unable to locate the fire position in the invisible region by this method was analyzed, and proposed forest fire fire point positioning method based on DEM and forest fire smoke model. With phylogenetic rule of TRIZ, a system improvement program has been drawn. By improved system, smoke model and smoke spread analysis, the location of the fire point which in invisible region can be estimated. Experiments show that this method is feasible.Finally, according to the text of the content of the study were summarized, and the lack of paper was given. Also the further research directions and suggestions were proposed.