Design Of Infrared Target Real-time Detection System

With the rapid development of infrared sensor and processor technology,infrared detection technology has become more and more widely used,and the detection equipment has become smaller and smaller.In the course of the infrared inspection and guidance process,the reliability and real-time performance of the detection determine whether there is success.Therefore,the research and design of the infrared target real-time detection system will have important practical significance.This text mainly carries on the improvement of the detection algorithm and the design of the hardware platform,including researching the accurate and high-efficient detection algorithm,designing the high-performance hardware platform.Ultimately,real-time detection of infrared targets is achieved.First,infrared processing technology and infrared image display enhancement technology are introduced.Then study the single frame-based background prediction algorithm and multi-step gradient algorithm.According to the advantages and disadvantages of the algorithm and the amount of computation,a new algorithm was designed that based on multi-directional gradient infrared target detection algorithm in multi-scale.According to the omnidirectional gradient features of infrared targets,the gradients in eight directions are calculated using the improved neighborhood gradient calculation method to achieve infrared target detection.The Gaussian Image Pyramid is constructed,and the multi-scale calculation of the infrared target is achieved through the layer-by-layer detection of the image.After single-frame detection,the use of multi-frame image information to determine the target reduces the false detection rate of the algorithm.Segmentation of the preselected target is performed using the region growing method,and then the mean gray value,mean square error of the gray scale,size,centroid,and Hu moment invariant are calculated.Using this information,the target's trajectory will be obtained by using the improved Manhattan distance to achieve target matching.Ultimately,the prior knowledge is used to filter out the target trajectory of interest and achieve accurate detection of the target.Finally,according to the characteristics of high computational complexity and high degree of parallelism,the hardware processing platform of ARM+FPGA is designed.In this platform,ARM is the core processor,and FPGA is used for algorithm acceleration and interface expansion.It also uses common interfaces such as Camera Link and SDI.So the device is easier to expand.In order to improve the overall efficiency of the system,a reasonable and efficient algorithm structure will be designed based on the characteristics of the algorithm and the data flow,in which the FPGA completes the calculation of a large amount of data.After actual testing,it can realize effective detection of 50 frames per second at a resolution of 640×512 resolution with 2×2 to 21×21 pixels,and has the advantages of low power consumption and small size.The realization of this real-time detection system has initially solved the need for real-time detection of portable infrared targets,and also provided a reference for the design of other detection systems.