100 using a frequency-based technique, which finds the translation and rotation relating to each image in each window in order to bring them into the correct placement and alignment with each other, while keeping robust against noise and the different anomalies associated with acoustic images,” Hurtós explains. She describes this technique as an algorithm designed for estimating translational shifts between images, with numerous features for overcoming the different sources of noise that can interfere with FLS image registrations, associated with the physics limitations of FLS technology, and with qualities specific to sonar and UUV dynamics. “As each image is aligned with the next, the algorithm tracks and calculates a score of how good each alignment is. Once registration is finished, that allows the user to filter out images that haven’t aligned together above a minimum they set,” Hurtós notes. “The software will notify them of how many images will be used in the mosaic, based on their chosen threshold. There are several options they can choose for how the disconnected parts of the survey imagery can be reconnected in the finished mosaic.” Before generating the mosaic, they may first preview the trajectory that the software believes the UUV took during its survey, based on the approximation gleaned from one pairwise image registration to the next. The trajectory is constructed of directional lines between consecutive images, and, naturally, this should closely match the navigation path originally plotted in the user’s GCS software (potentially revealing deviations in the UUV’s path caused by flow and currents of water during the survey). “The trajectory can then be enhanced by including registrations between non-consecutive images: surveys in a back-and-forth pattern with close loops between passes will capture images with overlapping detail on consecutive passes, and we can therefore include registrations between images that exceed a minimum overlap – even up to 100% if, for instance, a UUV went down a tunnel and then back out the same way,” Hurtós says. That enables the mosaic to be formed with the best position estimates (and IQUA has some acquisition guidelines of its own that it recommends to the surveyor to increase the chance of getting a good mosaic out of the data). The last step is to render the mosaic. Final options include the ability to manually or automatically adjust for values such as cropping the specific region of each image to be used. A much clearer and more detailed mosaic can June/July 2024 | Uncrewed Systems Technology With precise and intuitive tuning, mosaics (this one using a Kongsberg Discovery M3 acquired dataset) can evidence cracks, biofouling and other features of concern needed to persuade government or harbour authorities of critical action Before generating the mosaic, they may first preview the trajectory that the software believes the UUV took during its survey
RkJQdWJsaXNoZXIy MjI2Mzk4