Distributed compressive sensing vs. dynamic compressive sensing: improving the compressive line sensing imaging system through their integration

Abstract

In recent years, a compressive sensing based underwater imaging system has been under investigation: the Compressive Line Sensing (CLS) imaging system. In the CLS system, each line segment is sensed independently; with regard to signal reconstruction, the correlation among the adjacent lines is exploited via the joint sparsity in the distributed compressive sensing model. Interestingly, the dynamic compressive sensing signal model is also capable of exploiting the correlated nature of the adjacent lines through a Bayesian framework. This paper proposes a new CLS reconstruction technique through the integration of these different models, and includes an evaluation of the proposed technique using the experiment dataset obtained from an underwater imaging test setup. BACKGROUNDThe laser line scanner (LLS) is state-of-art underwater active imaging technology [6]. The LLS system employs serial raster scan image formation and is compatible with the push-broom type image acquisition commonly used in many survey operations. While the LLS system provides high photon efficiency and narrow instantaneous field of view that are critical for operating in degraded visual environment, the non-adaptive serial raster scan image formation can be a concern in certain situations. For example, higher repetition laser will be required with increased platform speed to maintain the same image spatial resolution, which inevitably will increase the system cost and require more sophisticated noise mitigation. For unmanned platforms such as UAVs and AUVs where power is at a premium, another concern is that the LLS sensor may consume significant power acquiring redundant data that will be discarded during the image compression phase (Figure 1). Figure 1. Traditional image acquisition paradigm in a LLS system [3]The compressive line sensing (CLS) imaging system has been developed [1] to address these concerns. By adopting the concept of resource compression, this system aims to achieve faster signal formation, more flexible system configuration and higher energy efficiency. The technique also enables compact and robust system design. A series of test-tank experiments were conducted to validate the CLS concept [3], and while the results demonstrated the effectiveness of the underlying distributed compressive sensing (DCS) signal model was demonstrated in [3], there was still room for improvements in signal reconstruction process. In this paper we describe our attempt to improve the performance of the CLS system by integrating the "dynamic" compressive sensing concept into the signal reconstruction.We provide some theoretical foundations in Section 2 and then introduce the CLS concept in Section 3. In Section 4 we introduce the new signal reconstruction technique through the integration of distributed compressive sensing and Ocean Sensing and Monitoring VII, edited by Weilin W. Hou, Robert A. Arnone, Proc. of SPIE Vol. 9459 94590D · © 2015 SPIE · CCC code: 0277-786X/15/$18 ·