| Material burning has strong radiation characteristics in the microwave frequency band,and microwaves are permeable to shallow coverings.Therefore,microwaves can be used to detect smoldering fire spots before and after disasters and fire spots under shallow coverings,thereby realizing full-time fire detection.Compared with infrared imaging and optical imaging,portable microwave radiation fire detector imaging has lower image resolution.Therefore,based on the principle of microwave radiation imaging,the use of image postprocessing algorithms to improve the resolution of microwave radiation images has important research value.Aiming at the problem of low imaging spatial resolution and small detection range of portable microwave radiation fire detectors,a method combining super-resolution reconstruction algorithm of convolutional neural network(SRCNN)and scale-invariant feature conversion algorithm(SIFT)is designed in this paper.,To achieve high-resolution large-scene fire detection.First of all,this paper designs a super-resolution reconstruction algorithm for microwave radiation images of fire scenes(SRCNN-MW)based on SRCNN,uses this algorithm to super-resolution images,and verifies the effectiveness of the algorithm from the following four aspects :(1)Intuitive evaluation of peak signal-to-noise ratio;(2)Comparison of high-resolution images reconstructed based on bilinear interpolation algorithm and SRCNN-MW algorithm;(3)Image feature point detection;(4)Image registration accuracy Evaluation.Among them,the SRCNN-MW algorithm not only improves the resolution of the microwave radiation image,but also improves the detailed information of the image.Then the SIFT-based image fusion method is used to fuse the image sequence,and the consistency of the relative position changes of the feature points of the fire points is analyzed to verify the effectiveness of the feature point detection method for fire detection.Based on the SIFT feature points,the registration of two microwave radiation images of the fire field is realized,and the overlapping and non-overlapping areas of the images are judged from the registration results,and finally the fusion of the images is realized.This paper uses the SRCNN-MW algorithm and the SIFT algorithm to process the microwave radiation image of the fire field,and realizes the high-resolution reconstruction of the fire field image of the large scene.The experimental results in this paper prove the effectiveness of the method and are of great significance to the imaging display research of portable microwave radiation fire detectors. |
