Research On Image-based Fire Risk Recognition In Complex Large Space

In recent years,with the development of economy and city,large space and high buildings are continuing to emerge.They bring practicality and beauty,permeability and good lighting,and new fire risk simultaneously.To fast and accurately recognize fire,and prevent the occurrence and development of fire danger,it is necessary to reach fire recognition technology and evaluate fire risk in complex large space to provide the decision basis for building safety management.Image-based fire recognition technology is not affected by the space height,air velocity and so on.It can solve the problem of fire recognition in large space building.The features from flame and smoke of images which are taken by video surveillance system are analyzed to recognize fire.And when the fire happens,fire risk evaluation is researched in order to let the fire command comprehend the disaster situation and improve decision-making ability in the fire rescue management in order to avoid mistakes in commanding and minimize the loss of the accident.Aiming at the problems existing research,the new method for early fire recognition and risk evaluation in complex large space environment(such as color similar objects,illumination change,object movement,changing movement pattern,etc.)is proposed.Specific studies are as follows.(1)Research on the extracting algorithm of new image features for flame and smoke.The typical flame and smoke image feature extraction methods are simulated in order to analysis advantages and disadvantages of those methods by using video image database.The feature extraction algorithms of the projected number ratio of the upper and lower edge for flame,and the smoke motion vector information and the correlation coefficient based on frequency domain are proposed.(2)Research on flame image recognition algorithm.Using flame color characteristics and target tracking technology,the flame suspicious target segmentation algorithm based on multi information fusion is designed.In order to improve the generality of the algorithm,the flame image feature set optimization model based on genetic algorithm and rough set attribute reduction is proposed.The cost of false classification would be embedded into the classification model.The compositive classification method based on SVM and Adaboost is presented,which includes the dynamic weight updates and number cutting of sample.Compared with the existing algorithms,the proposed algorithm improves the accuracy and speed of the flame identification.(3)Research on smoke image recognition algorithm.Because of the smoke turbulence phenomenon,the smoke segmentation algorithm based on blocking differential threshold and adaptive background updating is designed.It can extract gray(white)smoke and black smoke at the same time.The optical flow motion vector information and correlation coefficient are used as the basis of classification.The smoke identification model based on One-Class Support Vector Machine is established.The simulation results show that the algorithm has a certain ability to eliminate water vapor.The above researchs provide an important theoretical basis for the safety management of large space buildings.(4)Evaluation on fire risk.By the analysis for fire disaster factors,the fire risk evaluation index system are established by the images captured from the video monitoring system.The evaluation indexes’ weight is designed by the degree of deviation between expert evaluation and comprehensive evaluation,and fuzzy consistent judgment matrix.The fire risk evaluation model based on attribute recognition is constructed to predict the development scale and trend of the fire and provide scientific guidance for selecting appropriate fire fighting and rescue plan.(5)Research on of remote video fire safety management system.On the basis of the above theoretical study,using the advantages of DSP small volume,low power consumption,high reliability,and fast speed and combining the demand of fire safety management,the intelligent fire safety management system is design to improve the level of automation management which buildings prevent and control fire.