Simple psychovisual approach for multiple target detection and tracking in passive sonar imaging systems

摘要

Abstract: Real-time detection and tracking of multiple targets inpassive underwater sonar systems is essential for fastand correct response generation. In this paper, weexploit a model of the human visual early processing ofsensory information in designing a smart system for thedetection of moving targets and reliable estimation oftheir velocities. Psychophysical studies indicated thatmotion information are extracted by a system thatresponds to oriented spatiotemporal energy. Motionenergy detection is modeled by a linear high- passtemporal filter (simple frame to frame difference) anda spatial band- pass Gaussian filtering stage(Laplacian of the Gaussian image) followed by a squarerand summation. The energy image is searched formultiple maxima- depending on the number of targets ofinterest, and a square window highlights each detectedtarget. Detected targets are labeled according to theirmotion energy level, this labeling adds a confidencelevel to the detected target and helps reduce falsealarms. A reliable estimate of the labeled targetvelocity magnitude and direction is produced throughtracking of the energy maxima location from one-frameto another. As in the case of the human visual system,there is always a trade-off between accuracy andreliability of the velocity estimate. The algorithm wastested on simulated data. The synthetic data consistsof image data in rectangular coordinates, spanning afixed area, and with a fixed frame rate. The imagevalue was generated as the number of detections overthe frame interval at equivalent pixel locations. Theintensity was modeled as a Poisson process. Theintensity of the Poisson process was varied spatiallyto simulate non-uniform probability of detection.!12