SAR-BASED COASTLINE DETECTION AND MONITORING

The coastal environment is among the most fragile regions on our planet. Its efficient monitoring is crucial to properly manage human and natural resources located in this environment where a large portion of our population lives. The objective of this contribution is to design and develop a new set of methods suitable for detecting and tracking the coastline. Synthetic aperture radar (SAR) technology is chosen because of the characteristic response from water and the acquisition consistency allowed by constant illumination, day-and-night, and all-weather functioning. The proposed iterative detection method is based on superpixel segmentation. The resulting superpixels are filtered and then partitioned in land and water classes based on their median backscattering with Otsu’s algorithm. The rationale is that the segmentation can follow the coastline before the filtering can degrade the spatial resolution. A quantitative assessment of the results measures the distance to a manually-detected shoreline for the Lizard Island case study; the average distance is 12.63 m, with 80% of the sampled points within 20 m. The innovative coastline monitoring process exploits the consistency of SAR by analyzing a long time series. After a season-wise grouping, the land-water index is introduced to erase the time oscillation of water backscattering caused by different sea states. The proposed index is modeled in time on a pixel basis. A visualization technique that exploits the HSV codification of the color space highlights where and when changes happened. A case study for this technique is carried out over the Reentrancias Maranhenses natural area. A quality assessment shows good accordance with optical data that depicts the region’s dynamic.

Deep learning-aided high-precision data detection for massive MU-MIMO systems

The data detector for future wireless system needs to achieve high throughput and low bit error rate (BER) with low computational complexity. In this paper, we propose a deep neural networks (DNNs) learning aided iterative detection algorithm. We first propose a convex optimization-based method for calculating the efficient detection of iterative soft output data, and then propose a method for adjusting the iteration parameters using the powerful data driven by DNNs, which achieves fast convergence and strong robustness. The results show that the proposed method can achieve the same performance as the known algorithm at a lower computation complexity cost.