Temporally-illuminated Region of Interest Detection for Autonomous Underwater Vehicle Navigation
Autonomous Underwater Vehicles (AUVs) undertake scientific research tasks with few human interventions. A docking station (DS) can enhance long-term underwater missions by providing a location for AUVs to recharge batteries and update control instructions. There are three main solutions for recovering AUVs with docking stations: acoustic sensors, vision-based machine-learning algorithms, and magnetic field analysis. In this report, we present a vision-based method utilizing blinking light LED markers and an adaptive landmark filter to maximize visibility in varying water conditions and distances. We developed a system for the docking process. On the docking station, we place a blinking light in fixed frequency as a signal source. On the AUV, we create a buffer to store series frames and segment each image into rectangular regions. After getting the intensities of each region in each frame, we apply the Butterworth filter and band pass filter on them to get filtered intensities data. The preliminary result showed this system could be used in long-distance navigation.