Research On Ship Distance Information Perception Based On Video

In recent years,with the development of artificial intelligence and unmanned driving technology,the core technology of unmanned ships has gradually become a research hotspot in the maritime field,and the economic value of unmanned ships has gradually been recognized by the industry.A core technology of unmanned ships is situational awareness.Only on the basis of accurate and reliable situational awareness can safe and effective navigation decisions be made.The marine environment is complex and changeable.At present,there are two main ways for marine vessels to acquire target information through situational awareness:one is marine radar;the other is ship automatic identification system AIS.But these two methods still have deficiencies.Marine radar is unable or difficult to find small targets such as floating objects in the sea in time.Small ships such as pirate ships or fishing vessels will not install or turn on AIS equipment.Therefore,the reasonable application of visual devices such as visible light cameras to perceive the surroundirig environment can be a good supplement to the traditional way.In this paper,through the processing of video images,the mobile target model based on the depth separable convolution is used to detect ship targets.The binocular vision is used to obtain the depth information of the ship images,and then the relative position between the ships is automatically determined and the distance information is output.The paper compares and analyzes the distance information obtained,and the relative error is within the acceptable range.Experiments show that video-based binocular vision combined with ship detection to obtain the target distance is feasible for autonomous cargo transportation ships.