Optimization And Improvement Of Binocular Image Fire Detector

Fire is a serious threat to the safety of human life and property.The real-time detection and early warning of fire have great significance.When a fire occurs,it can timely locate the fire-point in space and guide the fire extinguishing equipment to put out it automatically,which can greatly reduce the loss caused by the fire.The normal binocular image fire detectors generally use a single fixed detection mode to detect and locate the flame,which has the problems of high false alarm rate and low positioning accuracy.This thesis mainly focuses on the optimization and improvement of the binocular image fire detector.The specific research works are as follows:(1)The hardware system of controlled spectrum switching based on binocular is designed.Firstly,the image of two channels is collected,preprocessed and transmitted by two "Hisilicon 3516A" chips at the same time;then,the real-time and accurate flame detection in the monitoring area is realized by image processing;after the flame is detected,the specific spatial coordinates of the fire point are calculated accurately by binocular stereo positioning;finally,the coordinates of multiple detectors are unified to achieve cooperative work among multiple detectors in large space.(2)The existing algorithm is optimized to improve the accuracy of flame detection and the accuracy of fire location.Flame detection:firstly,complete motion regions are obtained by dynamic object extraction and regional growth method.Secondly,detecting the features of flame circularity,flame sharp cusp and area change rate in dynamic object region.At the same time,the flame feature of Fourier descriptor is added in this thesis,it can describe the flame's dynamic characteristic in shape by calculating the Euclidean distance of Fourier descriptor.Finally,combining all the detection results of flame features to make a comprehensive judgement whether where exists a fire.Fire point stereo location:in this thesis,on the basis of the existing regional matching algorithm,the region optimal matching,ROI based local matching,region correlation checking and other processes are added.In order to reduce the probability of mismatching and the complexity of the algorithm,the algorithm calculates the exact region of the points to be matched.Meanwhile,the regional correlation verification is based on the correlation of pixels in the region,which further improves the accuracy of matching.(3)Qt UI program is designed to implement the visual operation and data display.A large number of experiments are carried out on the experimental platform to test the optimized and improved binocular image detector.The experimental results show that the optimized and improved binocular image fire detector has low false alarm rate,high fire point positioning accuracy and great improvement in real-time performance.