| The damage caused by fire to the forest is difficult to estimate.It is of great practical significance to detect the forest fire in real time and find out as early as possible whether there is a fire.In terms of forest safety protection,many scholars at home and abroad have conducted research and carried out practical application of forest fire monitoring.Due to the limited monitoring scope of the monitoring equipment and the impact of the forest environment,the real-time performance of the monitoring system is poor and the coverage is limited.For these problems,a real-time monitoring system for forest fire security based on infrared monitoring technology is studied.The main research contents of this thesis include:(1)Analyze the principle of infrared thermal imaging,compare its characteristics of visible light in forest fire images,study the characteristics of fire spread,and propose a forest fire monitoring and identification scheme based on infrared video images;(2)Designed an FPGA-based infrared image and video hardware acquisition system,which mainly includes image acquisition module,data buffer module and VGA display module,which meets the requirements of real-time video display;(3)image pre-preparation of collected fire infrared video Processing: Firstly,the histogram equalization correction is performed on the infrared video,then the fire suspected area is detected by the frame difference method,and the suspected area edge detection is performed by the Canny operator,and the suspected area is segmented from the background,and finally the mathematical form is performed.The closed operation of the learning obtains the complete segmentation part;(4)extracts the static,texture and dynamic features of the flame from the segmented suspected region,and performs data normalization processing to analyze the extracted features to exclude Those non-fire factors interfere with;(5)model the fire Identification: The support vector machine classifier is used to identify the extracted suspected region features,the radial basis kernel function is selected as the kernel function of the vector,and the cross-validation method is used to determine the value of the penalty factor c and the kernel function parameter g,and then with the BP neural network.The fire identification performance of the network is compared.The former has better recognition effect,and the accuracy rate on the test sample is as high as 93.8333%.Finally,the real-time monitoring platform of forest fire security is realized by means of MATLAB and Visual C++ software.The research results of this thesis can be used as reference materials in the field offorest fire prevention monitoring,providing important fire information services for the research of forest fire behavior and fire scene command,which has certain reference value. |
