The Realization Of Infrared Dim Small Target’s Track-before-detect Algorithm Based On Particle Filter Using DSP

With the development of the infrared imaging technology, the infrared search and tracksystem (IRST) is widely used in military and civil fields. As the key techniques in IRST, IRsmall target detection and tracking has become a research focus at home and abroad.Especially in the modern battlefield, the actual weapon system requires infrared small targetdetection and tracking with high probability of detection and tracking accuracy. It not onlyputs forward the high requirements for the algorithm, but also requests that the algorithm canbe realized using hardware,. With the digital signal processor (DSP) fast development, thedata processing and storage capacity of the DSP have greatly been improved. It not onlymakes the algorithm which is complex and needs large data implemented on the hardwareplatform become possible, but also provides the hardware foundation for the realization of IRdim small target’s track-before-detect(TBD) algorithms based on particle filter using DSP.In this pape, it has respectively introducted about the background suppression based onmathematical morphology filter, the particle filter algorithm, the IR dim small target’s TBDalgorithms based on particle filter and the commonly performance evaluation index which isused in the target’s detection and tracking. At the same time it has done the simulations for thealgorithms whose results indicate that the algorithm has very good detection performance forIR dim small target’s detection and tracking.In this paper, a scheme that IR small target detection and tracking system has beendesigned which is regarded DSP as the core is proposed. The target’s TBD algorithms basedon particle filter has been used in a real IR small target detection and tracking system in thescheme. The IR small target detection and tracking system has been designed innovative usingthe DSP/BIOS. In this paper hardware design and algorithm are introduced in details. In thehardware design, system is consisted of ICETEK-DM642-PCI development board andmonitor, the ICETEK-DM642-PCI development board is as the system data processingplatform, TMS320DM642(DSP) is used as the core of the platform. Compared with otherdata processing equipment, DM642has a good processing speed, programmable, andflexibility, which is very suitable for infrared image processing. The system manages thehardware resources by using the DSP/BIOS, and uses DSP/BIOS providing the chip supportlibrary to achieve the operation of the hardware modules. In the algorithm realization, the system uses a infrared small target detection before tracking based on particle filter. In thealgorithm realization, the system is divided into image pre-processing algorithm, the particleinitialization, the right value calculating, re-sampling, and particle detection module updates,the modules are realized by using c language. The system calls the task of image processingalgorithm by establishing the DSP/BIOS real-time thread.After debugging the hardware and software together of the system and experiment inpractice, it has a conclusion that the IR dim small target can be exactly detected and tracked bythe system under the specified case which has been proved. Through the analysis of the data whichgets form the actual operation, the result indicates that：when the particle number is600and theSNR is between3.5and4, the detection probability is100%, the time of system running is236.64ms and the mean of tracking’s RMSE is less than2pixels.