Research On Dual-spectrum Image Monitoring Method For Mine External Fire

Mine fire is one of the major accidents in coal mines,which directly affects the safe production of coal mines.However,the existing mine external fire monitoring methods have the problems that poor anti-interference and reliability,unable to realize long-distance monitoring,and difficulty in locating the underground high temperature source and fire source in a timely and accurate manner.Therefore,it is of great significance to carry out research on the monitoring technology of external fire in mines to improve the ability of mines to resist disasters,suppress the spread of fires,and formulate disaster relief measures.Based on the above-mentioned industry background,thesis carried out research on dual-spectrum image monitoring method for mine external fire,and the main research work is as follows:(1)A fuzzy enhancement algorithm for mine degraded images based on adaptive wavelet transform is proposed,which solves the problem of visible light image degradation in surveillance videos caused by complex lighting conditions in mines.The wavelet shrinkage threshold method of Bayesian estimation is used to adaptively adjust the wavelet thresholds of high-frequency sub-images at different scales.Using the constructed adaptive threshold function of wavelet,the shrinkage threshold filtering and nonlinear enhancement of high-frequency sub-images at different scales are realized.The brightness component adjustment of the wavelet reconstructed image is realized by using the improved membership function and the fuzzy enhancement operator,and then the visible light enhanced image is obtained.In addition,a mine infrared image enhancement method using dual-domain decomposition and ILo G-CLAHE is proposed,which solves the problem of infrared image degradation in surveillance videos caused by complex mine environment.A bilateral filter is used to decompose the infrared image into base and detail sub-images.The overall brightness and contrast adjustment of the base sub-image is achieved through the CLAHE algorithm.The noise suppression and edge sharpening of the detail sub-image are realized by using the ILo G operator.The processed basicand detail sub-images are adaptively reconstructed,and the brightness of the reconstructed image is adjusted by using the Gamma transform function which introduces an adaptive adjustment factor,and then the infrared enhanced image is obtained.The experimental results show that the proposed method has strong robustness and can overcome the limitations of traditional image enhancement algorithms in the complex environment of mines.(2)A long-distance accurate temperature measurement method for mines using infrared thermal imaging is proposed,which improves the long-distance temperature measurement accuracy of underground infrared thermal imaging and realizes long-distance monitoring of high temperature sources and fire sources in coal mines.By studying the principle of infrared radiation temperature measurement,the functional relationship between single-wave radiation illuminance and image gray level was established,and the radiation temperature measurement formula of thermal imager was obtained.The concentration and infrared absorption intensity of main components in mine air were analyzed,and the environmental parameters affecting the accuracy of long-distance temperature measurement by infrared thermal imaging in mines were established.The absorption of mine environmental parameters and the scattering of dust are analyzed,and the attenuation model of the total air transmittance of the mine and the distance of radiation temperature measurement is constructed.The proposed method is validated in several different experimental environments.The results show that the object temperature measured by the proposed method is closer to the real object temperature than the radiation temperature.At the temperature measurement point of 35 m,the average relative error value measured by the precise temperature measurement method is 2.85%,which further shows that the proposed method can improve the monitoring range of mine fire and the accuracy of fire alarm.(3)A mine fire detection method using single spectral visual features is proposed,which overcomes the problems of high false alarm rate and missing alarm rate in traditional fire detection algorithms.The improved RSG algorithm is used to achieve the segmentation of fire suspected targets in visible light enhanced images.A flame sharp corner detection model is designed,and the sharp corner feature extraction of fire suspected targets is realized.By selecting the dynamic and static features of the fire in the visible light image,the feature vector of the suspected fire target is constructed.A fire sample dataset of video images of visible light is established,and early mine fire detection based on BP neural network is realized.The improved LCM algorithm is used to obtain the saliency map of the infrared weak and small targets,and the segmentation of the suspected fire area based on the saliency map is realized.By selecting the dynamic and static characteristics of the fire in the infrared image,the feature vector of the suspected fire area is constructed.A fire sample dataset of infrared video images is established,and early mine fire detection based on SVM classifier is realized.The experimental results show that when the mine fire is in the early stage or long-distance monitoring,the proposed method can solve the problems that the visual characteristics of the fire are less,and the geometric characteristics of the fire source and the interference source tend to be consistent,and the existing fire detection algorithms have high false alarm rate and missing alarm rate.(4)A mine high temperature source and early fire detection method based on dual-spectrum fusion images is proposed,which solves the problem of light source and heat source interference in complex mine environment,and improves the reliability and anti-interference of mine external fire monitoring method.A mine dual-spectrum image fusion algorithm combining Re NLU and VGG-16 in MS-ADo G domain is designed,which realizes high-quality and high-efficiency fusion of underground dual-spectrum images,suppresses noise in images,and eliminates the interference of artificial light sources.Using the constructed MS-ADo G decomposition model,the visible light and infrared enhanced images are decomposed into baseand detail images.Using the constructed Re NLU function,the weight of the infrared basic image is automatically adjusted with the grayscale of the visible light basic image.By adopting the basic image fusion strategy of “weighted average”,the fusion basic image that eliminates the light source interference is obtained.Using the pre-trained VGG-16 neural network model to obtain the depth features of the detail images,and obtain the fusion detail images through the “maximum value selection” strategy.Combined with the visual features of the fused images,high temperature source and early fire detection based on dual-spectrum fusion images are realized.The proposed algorithmis experimentally verified by selecting visible and infrared video images from different surveillance scenes.The results show that the proposed method not only has the advantages of anti-interference and high reliability,but also improves the monitoring ability of high temperature source and early fire in complex mine environment.(5)A mine high temperature source and fire source detection and location method using binocular vision is proposed,which solves the problem that the existing mine external fire monitoring method is difficult to locate mine high temperature source and fire source,and the infrared image cannot be directly used for binocular vision ranging.The infrared checkerboard calibration board with different gray body radiation was designed and fabricated,and the accurate calibration of the infrared thermal imager was realized by Zhang’s calibration method.Using infrared image enhancement method and fire detection model was proposed,the rapid extraction of high temperature sources and fire source regions in far-infrared images is realized,and lays a foundation for distortion correction and stereo correction for high temperature source and early fire source area using internal and external parameters and distortion parameters.A binocular ranging model based on infrared images is constructed,and the realization process of stereo correction,stereo matching and binocular vision ranging of the infrared binocular ranging device is determined.Using the stereo matching algorithm based on the saliency target,the fine disparity map of the high temperature source or the fire source area was obtained,and then the high temperature source and the fire source were located by the fine disparity map and the binocular ranging model.