Cloud Service for Sonar Signal Processing

Abstract

Nowadays, a lot of autonomous unmanned underwater vehicles (AUV) are created in the Institute of Marine Technology Problem FEB RAS. Most of all include side-scan sonar (SSS) EdgeTechSonar with 32-bit resolution. Robots solve different problems. There are bottom map construction, monitoring of underwater channels, tubes, mineral exploration. Signal processing requires complex numerical resources. Unfortunately, AUV doesn’t solve these problems on the board independently. Real signal requires additional processing. Hence, sonar data is processed in the laboratory. Currently researchers have large databases of AUV. Data processing in the laboratory is inconvenient. If engineer adds additional processing algorithm he changes all local software. In this paper authors use cloud approach for solving this problem. The data from SSS EdgeTechSonar upload directly to the cloud service. Afterwards, user chooses any filter for sonar image processing. The cloud service includes a database of sonar images with its geolocation. Moreover, user uses different filters for image processing. Filters in cloud service are executed with parallel technology for fastest processing. There are 512 parallel processors with mpi-technology and several GPU-processors. Choice of technology depends on filter properties. Authors research some filters for sonar image processing and design special parallel algorithms for each of them.