Mapping and visualizing ancient water storage systems with an ROV — An approach based on fusing stationary scans within a particle filter

Dense, Sonar-based Reconstruction of Underwater Scenes

“…INTRODUCTION comparable to, and often slower than, the vehicle dynamics. To avoid the motion-induced artifacts that result from this limitation, some of the techniques will either assume-or keep-the vehicle in-place while a full scan is assembled [61]. Platforms equipped with dead-reckoning sensors relax this operational limitation by relying on the pose estimate that these sensors provide to compensate for vehicle motion while the scan is assembled [58,59].…”

“…One of the segmentation techniques most commonly found in the literature is that of (fixed) thresholding [58,61,82]. In this type of approaches, the intensity of each pixel is compared with an intensity threshold , and a scatterer is assumed to be present at the corresponding site (x ij = 1) if y ij ≥ ; otherwise the site is considered empty (x ij = 0):…”

“…As we have seen in chapter 3, multibeam sonars provide data on echo strength over range, from which range measurements must be determined. This is often accomplished by first using standard image processing operations as a pre-processing step to improve measurement accuracy, and then taking the first or strongest return above a threshold [13,58,61].…”

“…Unfortunately, the combination of the geometry of multibeam sonars with platform motion frequently precludes the use of this family of algorithms-as platforms move perpendicular to the scanning plane to maximize coverage rate, overlap between consecutive images is eliminated. In cases where there is overlap due to in-plane motion, ICP can only provide partial (in-plane) constraints [61], and these tend to be poorly informative in the case of small fields of view. Thus, out-of-plane motion generally leads to the use of submap-based techniques, where sensor measurements and odometry are accumulated over a short time frame to produce a "virtual" sensor measurement that can be registered with previous ones and produce a relative motion estimate [82,104,116].…”

“…To address these limitations, some techniques complement the threshold operation with additional image processing steps, such as morphological operations such as dilation and erosion[13]. Other approaches have also added some other form of post-segmentation outlier rejection, usually by looking at the agreement between the range measurements of neighboring beams to improve the accuracy of the range estimates[58,61,82]. Recognizing the difficulty in obtaining an accurate and robust segmentation using threshold-type methods, other approaches have instead addressed these limitations later in the pipeline[115].…”