Research On Forest Fire Smoke Detection Method Based On Video Analysis

Forests play a key role in the sustainable development of ecosystems,and the occurrence of forest fires seriously threatens the development of ecosystems.In the early stage of the forest fire,the flames were difficult to be detected due to the trees,but the smoke was easily captured by surveillance video.Compared with the traditional sensor technology,the forest fire detection technology based on computer vision has the advantages of wide monitoring range and strong real-time performance.It can provide early warning in the early stage of forest fire,which greatly reduces the damage caused by the fire.Therefore,it is of great significance to study the forest fire smoke detection method based on video analysis.According to the characteristics of forest fire smoke video,in order to accurately and quickly identify forest fire smoke images,this paper conducts indepth research on related issues such as moving target detection in smoke video,smoke image feature extraction,and design based on support vector machine classifier,and proposes based on Video analysis of forest fire smoke detection methods.The detection of moving objects in smoke video is very important for the recognition of forest fire smoke.It is the basis for the subsequent static and dynamic feature extraction of smoke and the calibration of smoke position.Among several common methods for moving target detection,the Vi Be algorithm has the characteristics of less memory consumption and low computational complexity,but it has the disadvantages of poor adaptability to light changes and sensitivity to noise.In this paper,the light invariance of the SILTP operator is used To improve the Vi Be algorithm,the experimental results show that the improved algorithm has strong robustness to illumination changes and noise.Furthermore,the freeman chain code is used to extract the edges of the moving target image,and the moving target image is obtained through an external rectangular frame.At the same time,the static and dynamic characteristics of smoke are studied.Static features include texture features,high frequency energy features,and irregular edge features;dynamic features include center of gravity movement features and motion direction features.The results of the study show that it is difficult to accurately distinguish smoke and non-smoke images with a single feature,and multiple feature combinations are required to effectively distinguish.To this end,this paper designed a smoke recognition classifier based on support vector machine to analyze and compare the factors that affect the classification accuracy,including feature data normalization,different kernel functions,different parameters and different feature combinations,according to the comparison results Optimize the classifier to obtain the smoke classifier with the best recognition effect.The combination of moving target detection,smoke image feature extraction and smoke classifier constitutes a forest fire smoke detection algorithm,and the performance test of the algorithm.The test results show that the smoke video image detection algorithm based on video analysis proposed in this paper is for smoke video images.The detection accuracy rate reached more than 93%,realizing accurate detection of forest fire smoke.