Research On Optical Image Detection Technology For The Scatterable Mine

As the mines are inexpensive and lethal,they have become an important weapon in military defense.At present,limited by the detection conditions,the single mine detection method has been unable to complete the daily mine detection work.Therefore,the development of composite mine detection technology combined with multi-channel and multi-technology has become a research trend.In addition,in view of the defect existed in multi target detection of the remote scatterable mines,which are low work efficiency,high false alarm rate and low detection rate,the target detection and recognition technology based on multisensor provides a new direction for the research of composite mine detection system using the space-time relativity of image and the complementarity of information.Therefore,the multi-sensor cooperative detection and recognition technology studied in this thesis has important theoretical and practical significance of the research on composite mine detection system.In this thesis,aiming at the detection of ground scatterable mine,a vehicle multi-sensor scatterable mine detection system is designed which can achieve identification and location of scatterable mine targets at a relatively long distance.The main research content includes: the design of mine detection and recognition algorithms for infrared and visible images,the research of mine recognition and location based on multi-sensor,the design of software interface and hardware platform,the hardware implementation of the algorithm and the optimization of the system.The specific work is as follows:(1)The visible light image and the characteristics of the target in the visible light image are analyzed.The Ada Boost algorithm based on Haar feature and the LBP algorithm are applied to the detection and recognition of the mine target,and the system of first detection and rerecognition is adopted to meet the requirements of high detection rate and detection speed.The scatterable mine images in different scenes is analyzed experimentally in order to verify the effectiveness of the algorithm.(2)The target detection and recognition algorithm in infrared mine scene is designed according to the characteristics of infrared images and the prior knowledge of mine targets.Experiments show that the proposed sub-segmentation algorithm based on maximum entropy can segment the mine target accurately,the appropriate morphological approach is selected to optimize the segmentation image,and the proposed multi-level discriminant algorithm for closed region is effective in detecting and identifying mine targets.(3)According to the system detection requirements,the appropriate infrared and visible light detector is selected.The calculation of the image coordinates and the geographic coordinates of the target is completed and the coordinate error is estimated,which paves the way for the fusion of coordinates.(4)According to the recognition results of infrared and visible light images,the fusion model of confidence and coordinates is proposed and the whole system of mine target detection and recognition algorithm is improved.Experiments show that the algorithm can fuse and locate the target recognition results of the multi-sensor in the same scene.(5)According to the functional requirements and performance indicators of the project,software interface and hardware platform design is completed.By analyzing the complexity and efficiency of the algorithm,the hardware task allocation is completed.The algorithm is transplanted to the hardware platform and the system is optimized in order to improve efficiency.(6)The multi-sensor scatterable mine detection system is debugged and the development of the system is completed.The system is used to achieve scatterable mine detection and positioning,which verifies the effectiveness of the system.