Unmanned Boat Target Detection Based on Improved SSD
Yang Yin, Fan Gui, Shuai Chen · 2019
Currently, vision sensors are one of the most equipped sensors in unmanned boats, and they provide an intuitive description of the environment. Deep learning based convolutional networks can identify and classify targets well. Considering the need for real-time obstacle avoidance by unmanned boats, this section selects the lightweight model Single Shot Detector (SSD) as the recognition framework, but the SSD will appear discontinuous in the target tracking. To this end, the recognition framework of SSD&CAMShift is proposed, and CAMShift is used to obtain the pixel statistical probability of the target to achieve auxiliary tracking. The simulation results show that the designed tracker can automatically detect and track, and the data correlation between the upper and lower frames is greatly improved.