| Title |
DeepNetwork-based Segmentation Model for Low Detectable Underwater Target Tracking |
| Authors |
신원(Won Shin) ; 설호석(Hoseok Sul) ; 최지웅(Jee Woong Choi) ; 송택렬(Taek-Lyul Song) ; 김다솔(Da-Sol Kim) ; 고현석(Hyunsuk Ko) |
| DOI |
https://doi.org/10.5573/ieie.2023.60.1.27 |
| Keywords |
DeepNetwork-based segmentation model; Bearing-time records image; Low detectable underwater target tracking |
| Abstract |
In submarine combat systems, passive sonars are used to detect enemy targets without exposing the ship's location. Since bearing information can be obtained through passive sonar, and information about distance can be obtained by performing additional target maneuver analysis using the obtained bearing information, it is important to secure accurate bearing information. The signal received through the passive sonar can be represented by imaging the target's bearing information with respect to time, which is called BTR(Bearing-Time Records) data. In this paper, we propose a deep learning-based segmentation network to improve the target detection rate from BTR data. Since it is difficult to obtain the actual target bearing data as military information, a simulated BTR dataset was created and network learning and experiments were conducted through it. In particular, we propose a Spatial Convolutional Layer-based segmentation network to well extract target objects in BTR images with directionality. The proposed model shows the best target detection performance compared to the existing deep learning-based segmentation model in experimental datasets with various intensity noise environments and target detection probabilities. |