Algorithms Survey Of Dim Targets Track-Before-Detect Based On Dynamic Programming

Detection and tracking of dim targets represents the hotspot in the field of modern radar technology, and TBD is the crucial mean, which directly uses the original data of sensor without processing by low threshold; and make the best of useful information in the data. The ratio of signal increases through the accumulation of information in the time domain, meanwhile, detection and tracking of dim targets can be achieved.This paper investigates the application of algorithms based on DP-TBD into the detection of dim target of radar, which includes the following parts:In the first place, this paper introduces traditional dynamic programming algorithm(DPA), and it detects and tracks motion targets in the background of Gaussian. DPA can accumulate data according to the destination path and follows the principle that the accumulated energy is the largest. Therefore, this study can concludes that target track in the data field has an important effect on the observation and unrelated accumulation of the data through further observation. After a period of accumulation, value function will exceed threshold of Mo malicious, when target can be detected and tracked back.Secondly, it talks about the TBD algorithm, modeling and simulation based on particle filtering and HF transforming. By comparison to TBD algorithm based DP, this paper comes up with some part needing improvement: it adds the direction of weight to traditional Dynamic Programming Algorithm to reduce the proliferation of energy, and makes use of the effect on target track exerted by strong noise points in the weakened low Snr conditions of the finite-step forward-looking strategy, after which, it reduces the amount of algorithm computation through the combination with two-level threshold detection method.Finally, this paper raises the algorithm of track-before-detect for multiple targets. It introduces the extreme value method based on DP. Due to the fact that extreme value method cannot be applied to two targets with the same measuring values at the same time, bubbling method based on DP is presented. It first detects maximum strength goal and then removes the goal from the observed value, after which, it repeats detecting the maximum and removing it again until all the goals are detected. As how it works, it transforms detection and tracking of multiple targets into those of single one.