SONAR Imaging using Compressive Sensing

2nd Place, Engineering Division - 2011 Denman Undergraduate Research ForumSonar imaging is a technique that can locate objects in a scene from their acoustic reflections. It utilizes multiple measurements from different angles to create a 2D image. Conventional imaging algorithms such as backprojection (BP) require a large number of measurements to produce accurate images. Theory indicates that an alternative approach called compressive sensing (CS) can create accurate images using just a fraction of the measurements. CS requires that the scene is compressible and that each individual measurement provides information about the scene, not just a particular location. The sonar imaging experiment must be designed to meet these requirements. The purpose of this study is to show that the compressive sensing framework can be used in sonar imaging with results comparable to conventional methods. A sonar test stand was built that can measure acoustic reflections from a scene. A pseudo-random noise (PN) code was used for transmission. Four speakers and 16 microphones were mounted randomly perturbed from a circle surrounding the scene. Software was developed to create an image from this data through FBP. Also, a CS algorithm was developed to reconstruct an image from limited measurements. This algorithm uses a random subset of samples from each measurement. Initial results show that FBP can effectively be used to image a scene using acoustic waves. The CS algorithm yields a similar quality image using less than 10% of the measurements. These results show that CS can be used in sonar imaging to greatly reduce the number of measurements that need to be collected. The findings are directly portable to radar imaging, a field with a high level of research for both military and civilian uses.Air Force Research Lab - Dayton, OHNo embarg