Proceedings Article | 31 January 2023
KEYWORDS: 3D acquisition, Target detection, Detection and tracking algorithms, Error analysis, Mathematical modeling, Data modeling
Through the passive observation and algorithm processing of the two-station infrared camera, the three-dimensional coordinates of the landing point of the shell are located, and the target is automatically reported throughout the day for the range training, which is an important way for training evaluation. The traditional photoelectric imaging target reporting method directly locates the double station based on the detection of the explosion point of the image shell, locates the three-dimensional position of the shell explosion point, and outputs the target reporting results. This method faces two difficulties. Firstly, because the ground in the image cannot provide enough geometric prior knowledge, the detection deviation of the shell explosion point in the image is large, resulting in the target reporting result based on the double station positioning of the shell explosion point cannot meet the 1m accuracy requirement. Secondly, due to the interference of the mushroom cloud at the pre-sequence shell blast point, the explosion point of the group shell is difficult to be effectively detected, so the method does not support the continuous target reporting of multiple bullets. Based on this, this paper proposes a range prediction target reporting algorithm, which is based on the detection of the shell's aerial trajectory for double-station positioning, first locates the three-dimensional track points of the shell target when it moves in the air, then fits out the three-dimensional trajectory of the shell in the air, and finally predicts the three-dimensional position of the shell explosion point based on the bulls eye GPS information, and outputs the target reporting results. The algorithm avoids the detection of image explosion points, which can improve the positioning accuracy of explosion points, and can avoid the interference of mushroom clouds at the explosion points of pre-sequence shells and realize the automatic target reporting of continuous shells. In addition, in view of the problem of matching between group shell stations, this paper adopts the method of multi-target matching target based on track direction estimation. Firstly, it performs single-frame multi-target preliminary matching based on the elevation difference of dual-station direction finding rays and obtains all three-dimensional track points in the air of each shell target. Then, based on the two-dimensional histogram of the direction or the Mean Shift algorithm, the three-dimensional track direction of each shell target is estimated, and the true and false track points are checked based on the three-dimensional track direction of the target, and the false matching points are eliminated. Finally, the three-dimensional trajectory fitting and the position prediction of the explosion point of the shell target are carried out, and the target reporting results are output. Experiments verify that the positioning error of a single shell in this algorithm is 0.57m, and the positioning error of the traditional target reporting algorithm is 1.17m, indicating that even when the shape of the explosion point is regular, the positioning error of the predicted target can be doubled; For the one-shot two-bomb test that can detect the target, the positioning error of the algorithm in this paper is better than 1m, which meets the overall application requirements. In view of the target training of the Yellow Flag Sea test of ten bombs and fourteen bombs in a row, the algorithm fits the multi-target three-dimensional track parallel to each other and distributes it uniformly, which can intuitively and qualitatively judge the correctness of the matching target. In addition, from a quantitative point of view, the test found that the algorithm can not only eliminate the mismatch point, but also suppress the point with large error, thereby reducing the multi-track fit error and improving the multi-target reporting accuracy: the maximum fit error of the track < 0.5m, and the average fit error <0.3m. In summary, this paper proposes a multi-target prediction target reporting algorithm based on three-dimensional trajectory direction estimation, which avoids the detection of image shell explosion points, based on the image shell aerial track detection, conducts three-dimensional trajectory estimation of each shell target, predicts the three-dimensional position of each shell target explosion point, outputs high-precision target reporting information of each shell target, the target accuracy is better than 1m@1km, and supports a multi-bomb continuous target reporting, which can provide scientific and effective data support for training evaluation.
