engleski

CUDA-based parallel underwater target reconstruction

This paper addresses underwater machinery reconstruction based on sonar system simulation. Sonar beams are considered as vectors and underwater object is represented with a large set of small triangles. Intersection points are calculated within sonar range by using a well known algorithm for intersection point calculation between a line, i.e. sonar beam, and a plane, i.e. object's triangle. GPU based parallel implementation is presented in this work. Execution time is compared with sequential CPU implementation. Depending on the number of rays and the number of triangles, experimental results show speedup ranging from 100x to 200x. While CPU- based underwater reconstruction simulation took few hours, GPU does the simulation in only few minutes.

underwater machinery reconstruction; sonar; cuda; gpu

nije evidentirano