Tracking multiple fragmented objects with 2D imaging sonar
Kevin J. DeMarco, Noel E. Du Toit, Ayanna Howard · 2016
In this work we have developed a multiple object tracker that is capable of tracking objects that produce fragmented returns when ensonified. The fact that the types of objects that we are tracking (e.g., divers) produce multiple returns required us to develop an augmentation to the classical multi-hypothesis tracking approach. To account for the fragmented object returns, a novel adaptive Kalman filter R measurement matrix algorithm was developed. By distorting the R matrix and, thus, the measurement validation region, the object tracker was able to “catch” the fragmented blobs that belonged to it. Also, we demonstrated that the same object tracking algorithm can be used to track objects that do not produce fragmented returns, such as fish. We demonstrated the effectiveness of the multiple object tracker at providing long continuous tracks on data we collected at the Georgia Tech acoustic dive well and at NASA NEEMO Mission 20.