Research On Acceleration Technology Of Pulse Radar Signal Sorthing Algorithm

Radar detection as eye of modern information warfare plays an important role in electronic countermeasures.With radar signal environment has become increasingly complex,are some limitations in the pulse radar signal sorting algorithm there.There are still some difficulties in the sorting of pulse radar signals with a wide range of multiple frequencies and a wide range of parameters.The sorting processing speed of pulse radar signals with high density and large data volume still needs to be improved.In the future,the signal sorting of radar countermeasures will face more severe challenges.this paper,a new pulse radar sorting algorithm is designed,which is density-based fuzzy C-means multi-center reclustering radar signal sorting algorithm,DBSCAN.The algorithm combines the advantages of DBSCAN density clustering algorithm and FCM fuzzy C-means clustering algorithm.Not only sorting of pulse radar signals with unknown radar numbers,but it also can describe the relative membership of each sample point and each cluster similarly to the FCM algorithm.Compared with kmeans algorithm,DBSCAN algorithm and FCM algorithm,DFCMRC algorithm has a better sorting effect.The accuracy of sample test is more than 94%,and its relative membership is also more reasonable than FCM algorithm's.The structure of DFCMRC algorithm is optimized by changing the original algorithm of randomly generating initial center point to calculate ideal initial center point through density peak clustering algorithm.Because the ideal initial center point is very close to the target center point,the optimized algorithm can reduce the number of iterations and greatly shorten the calculation time of the algorithm.In order to further improve the separation of real-time,parallel processing is made for the algorithm,this paper adopts CPU + GPU heterogeneous platform to load the radar sorting algorithm,using the GPU's powerful parallel processing ability,effectively shorten the calculation time of the algorithm.In order to give full play to the computing performance of the hardware,this paper also optimizes the parallel algorithm several times,avoiding warp divergence and bank conflict,and solving a large number of data summation global synchronization problems.After testing,the speed ratio of the algorithm after parallel optimization is more than 80 times.When the two optimization methods are combined,the theoretical acceleration ratio can exceed 160 times.