Research On Several Key Techniques Of Multi-Sensor Multi-Target Track Correlation And Data Combination

Multi-sensor data fusion is a comprehensive subject, of which the purpose is to perform the research on the theories, the techniques and the methods of processing the multi-sensor data. The generating, the forming and the developing of it are the outcome of the rapid development of modern scientific technology. It has wide application prospects in military and civil areas. By taking the "11th five-year" defense naval equipment research projects as the background, draught by the application requirements of the actual multi-sensor data fusion, this paper performs the research on the issues of track-to-track correlation and data combination in the multi-sensor data fusion, focusing on the multiple radar track-to-track correlation algorithm, the heterogeneous sensor track-to-track correlation algorithm, the multiple radar track data dynamic weighted mean combination algorithm as well as the one-dimension bearing only sensor measurement and radar track data combination algorithm.1. Summarizes the definition, the basic principles, and discusses the key techniques and processing flow of the distributed multi-sensor data fusion system.2. Presents a multiple radar track-to-track correlation algorithm based on transitive closure fuzzy clustering. For the multiple radar and multi-target correlation problems, this algorithm uses the method of transitive closure clustering based on the fuzzy statistical quantity to perform the clustering analysis of the track data from the multi-radar, implementing the multiple radar track-to-track correlation. Firstly, assemble the samples of all the track from the multiple radar compositions to be classified sample set; secondly, selects the target location information and the speed information in the track as the fuzzy factors; establishes the fuzzy similarity relationship between each track pair; then calculates the transitive closure of the fuzzy similarity relationship to get equivalence relation; and finally, uses the equivalence relation to determine the track-to-track correlation pair and implement the multiple radar and multi-target track correlation. This algorithm is insensitive to the radar track error, therefore, it is suitable for performing correlation determination under the conditions of maneuvering and cross motion being performed by the target.3. Presents a heterogeneous sensors track-to-track correlation algorithm based on B-mode gray correlative degree. For the problem of multi-target track-to-track correlation in the heterogeneous sensor fusion system composed by one-dimension bearing only sensor and radar, based on the gray correlative analysis method of the gray system theory, the algorithm takes the target track in the heterogeneous sensor data fusion system composed by the one-dimension bearing only sensor (e.g. infrared, passive sonar) and the radar as the time series of the target bearing information, and the bearing time series set is composed of all the target tracks, performs the B-mode gray correlative analysis on the elements of this set, establishes the track gray correlative matrix, and determines the track-to-track correlative pair according to this matrix, implementing bearing only track-to-track correlative of the heterogeneous sensor. This method is also suitable for the bearing only sensor track-to-track correlation on the same platform.4. Presents a heterogeneous sensors track-to-track correlation algorithm based on fuzzy numbers similarity degree. For the problem of multi-target track-to-track correlation in the heterogeneous sensor fusion system composed by the one-dimension bearing only sensor and the radar, this algorithm uses triangular membership function to fuzzy the track bearing information and then to get the fuzzy number of the bearing information, and next calculate the degree of similarity between the fuzzy numbers of the different track bearing information, forming the matrix of the degree of similarity of the fuzzy number of the bearing information, determine the track correlation pairs based on the similarity degree's matrix, and implementing bearing only track-to-track correlative of the heterogeneous sensor. This method is also suitable for the bearing only sensor track-to-track correlation on the same platform.5. Presents a multi-radar target track data weighted average algorithm with weighting factor dynamic allocation. For the weighted average combination method often used in engineering practice for multi-radar track data combination, firstly, analyzes the influence of weighting factor allocation upon the fusion precision; secondly, proves the optimal weighting factor allocation principle; finally, presents a weighting factor dynamic allocation method, which uses the track information of the same target tracking by the multiple radar to perform dynamic estimate for the precision of the track data outputted by any radar, and exploits the track data precision estimated dynamically to calculate the weighting factors according to the principle of optimal weighting factor allocation, implementing the dynamic allocation of the weighting factors in weighted combination of the multi-radar target track data, and solve the multi-radar track data weighted average combination problem under the condition that the information of radar track data precision cannot be known precisely, i.e. the optimal allocation of the weighting factor cannot be performed. The precision of the combined track data will be higher than that of any radar track data.6. Presents a algorithm for the combination of the one-dimension bearing only sensor target bearing measure and the radar target track data. This algorithm firstly solves the target bearing information of the target information detected by the radar and the bearing only sensor relative to the same coordinate system, and solves the bearing precision by the approximation method; secondly, combines the bearings by the optimal weighting factor allocation weighted average method; finally, exploits the combined bearing information to correct the target location information detected by the radar, and takes the corrected target location information as the combined result of the target location. The precision of the combined track data will be higher than that of the radar track data. This algorithm gives the precision of the combined track data while gets the combined result.