Research Of Vessel Target Image Recognition Technology

Nowdays, with the social economic growth and the rapid development of inlandnavigation, the amount of inland vessels and their average carrying capacitycontinued to grow, so that the vessel navigation safety problems have becomeincreasingly prominent, therefore, the vessel navigation monitoring has become thefocus area of Maritime Bureau. Since the MSA provides that AIS equipment isrequired by over300gross tonnage, vessels can be left out by using AIS monitoring;while radar will be limited by meandering waterway and it is also costly; what’s more,CCTV can only observe the vessel navagation, but not taking operation. The use ofimage processing to monitor the vessel navigation has various advantages, such aslow cost, none leaving vessels, detecting accurately and low impact of terrain. So it issignificance for not only monitoring the status of vessel navigation, maintenancesafety, but also measuring the vessel speed, type, length in the future.In this paper, with the MSA regulatory and specifications of navigation, thevessel navigation information collection point is determined by the basis of a fullinvestigation. Secondly, target image segmentation can be done after the collection ofvessel navigation information. Finally, the vessel characters can be calculated bytarget image segmentation, all of the characters have the property of RSTinvariability.The main contents of this paper include the following:(1) The vessel navigation information is collected and the video imagepreprocessing can be done in this part. Firstly, image exchange is processed. Secondly,image filtering and image enhancement are processed, so that the difference betweenforeground and background is more obvious.(2) Vessel target video image segmentation is done in this part. Firstly, severalcommon methods of target segmentation are introduced, and compare which methodis determined. Secondly, Ostu method is used to select the threshold, combines withbackground rebuilt. Therefore, a image deference method is proposed, it can detectthe vessels in different angle(face or side face), different weather(sunny foggy orcloudy). Then the method is detected by experiment. (3) Character distilling are detected based on the vessel binary image. First of all,some invariants which have RST invariability are introduced. Secondly, after theintroduction of the invariants, many new character distilling methods are proposed,such as the ratio of target area to its protruded shape area, cirque division, curvaturemesure, the distance of the center to contour pixel points. Those algorithms’implementation methods are given.This study provides a kind of vessel object segmentation algorithm, thenproposed many new character distilling methods, which have RST invariability. All ofthem play an important role in dynamic monitoring of navigation and protection ofthe safety of navigation.