Sonar FoV Segmentation for Motion Estimation Using DL Networks

Abstract

Accurate navigation of an autonomous underwater vehicle is important for its reliable operation. However, this task is challenging due to limitations of radio wave propagation and poor visibility in the aquatic environment. Underwater navigation techniques based on analysing sonar images facilitated by machine learning have shown promising results. However, previously proposed techniques are still complicated for real-time applications. This paper investigates low complexity techniques for the motion estimation based on the use of images obtained by a sonar looking down to the seafloor. The sonar can use multiple beams within a field of view (FoV). Various configurations of beams are considered according to portions of the FoV covered and two estimation approaches are investigated. In one approach, the sonar images are directly processed by a deep learning (DL) network, whereas in the other, the images are converted into (reduced size) vectors before applying them to a DL network. The vector approach shows a significantly lower computation time (about 10 times faster), which makes it suitable for realtime applications. Both the approaches show a similar estimation accuracy, about 10% of the maximum magnitude of the motion. The vector technique has been used to estimate a simulated trajectory and compare the estimate with the ground truth, which showed a good match. It has also been applied to estimate the trajectory of an imaging sonar from a real data set from a ship's hull inspection. The estimated trajectory has successfully been used to build a mosaic by merging the sonar images from the real data set.

Motion Estimation of Underwater Platforms Using Impulse Responses From the Seafloor

“…The last layers of each PoseNet are concatenated together and connected to a fully connected layer and finally to an output regression layer that generates the motion estimates for the two DoF. As in [22] and [23], the Mean Squared Error (MSE) is used as the loss function of the regression layer as follows:…”

“…Therefore, the maximum speed in a particular direction is approximately 2.8 m/s. The value τ r approximately matches to the ping period in a model of a DIDSON 300 sonar [39] which has been used for investigation of the motion estimation in works [22], [23]. Data sets of impulse responses are generated for various parameter sets (configurations) of the transmitter and receivers.…”

“…In [22], the method from [21] is improved by using deep learning (DL) networks. In [23], sonar images from multiple sensors are compressed into vectors and grouped together to be used as the input of the DL network, thus reducing the DL network complexity, while keeping the high motion estimation accuracy.…”

“…Finally, a CVL requires an array with a large number of receivers [27], [16] while the proposed technique can achieve fairly accurate motion estimation with only two receivers and high accuracy with four receivers. Compared to the techniques based on environment referencing, such as in [20], [21], [22], [23], where sonar images represent estimates of the channel impulse response magnitudes varying in time, and thus ignoring the phase information, the proposed method is based on the whole information of the channel impulse response and thus is capable of providing a better motion estimation performance. The proposed technique is based on acoustic impulse response estimates with a much higher time resolution and carrying more information than the classical sonar images.…”

“…Note that a large amount of labeled data is required when training a DL network for regression tasks [29]. To solve this problem, we use synthetic data generated by a computer simulator, similar to the approach used in [22], [23]. The simulator can provide large data sets with the ground truth required for supervised learning.…”