Research On Stitching And Recognition Of Coal Mine Monitoring Images

Coal mine safety assurance is of great important to guarantee lives of coal miners.Video surveillance is one of the important means of coal mine safety assurance. However, with artificial supervision, most monitoring models of existing coal mine video surveillance systems are too simple to effectively or timely capture abnormal states, which can not sufficiently meet the requirements of coal mine safety assurance. Therefore, it is quite necessarily introduce some technologies as core means of mine perception technologies in video surveillance systems, such as intelligent video sensing, digital image processing and so on. However, occlusions, poor illumination conditions, serious interference by mineral dustand other problems in monitoring environment could result in a serious decline in the quality of monitoringimages, posing a serious challenge to image stitching in video surveillance, moving targets detection as well as their behavior pattern analysis. Considering the special environment of coal mine, this paper puts forward solutions to the problems discussed above which revolve round image de-noising, image enhancement, image stitching of mine surveillance video, moving target tracking and recognition a+s well as its behavioral analysis under complicated circumstances. Financial supports for this work are provided by Scientific Research Project of Huaibei Mining Group(Research on Intelligent Video Sensor for Multi Object Mining), Jiang Su province “333 project”(The project name is key technology research on image stitching and recognition of coal face, No. BRA2014048), and Project of National Natural Science Foundation of China(Model And Method of Distributed Monitoring Information Source with High Reliability Compression Encoding for Coal Mine IOT, No.70533050). The main works are accomplished as follows:(1) A new borehole image de-noising method based on visualcharacteristics is proposed to cope with problems(e.g., weak illumination, high noise) of the underground image. The method uses a uniform CIELab color space, to dynamically and adaptively determine the filter weights, thereby reducing the edge information and other details damages of image contour, consequently makes the denoised image clearer.(2) According to the requirement of image quality improvement, an adaptive image enhancement method based on immune genetic algorithm is proposed. By applying the adaptive immune genetic algorithm to the image enhancement processing, this proposed algorithm is an enhanced non-linear adaptive algorithm, which can automatically adjust the image quality according to the variation of image gray-scale. With the new vaccines choosing strategy and immunization operation, this algorithm allows for automatically seeking for the optimum transformation parameters of nonlinear image enhancement function, to achieve the image enhancement.(3) According to the requirement of large scene video surveillance, a method of automatic and quick stitching of mine monitoring image is proposed. This method is based on Harris feature point extraction algorithm and SIFT feature extraction algorithm analysis, by means of the improved RANSAC algorithm filters to match points and calculate the transformation matrix, to achieve great improvement of the ahti-scaling algorithm changes and noise immunity. This method also implements the LSH algorithm to improve image stitching success rate and real-time performance.(4) Intelligent visual localization algorithm based on PCA-SIFT features is proposed according to the actual needs of intelligent video surveillance in the complex environment of coal mine. The SIFT features are integrated into the Mean Shift tracking methods to improve the accuracy and real-time performance of location perception.(5) A fast template matching algorithm based on the fast Fourier transform is proposed according to the characteristics of underground moving target recognition analysis. By using the SVM classification training method to build the database of targets models and behaviors, this algorithm conducts the moving targets behavioral analysis.The above research results enrich the research contents of intelligent video monitoring field, and to solve a series of key technical problems of the construction of intelligent video surveillance system, and having an important supporting role for building coal mine intelligent monitoring system.