Dynamic Visual Identification Method Of Waterline For Ship

As important basis for cargo port accounting,draft survey is the main method to measure the weight of import and export goods currently.However,the traditional manual observation method is inefficient,difficult to ensure the accuracy and fairness,and has a high risk coefficient,which cannot meet the development needs of port digitization and informatization.A camera installed in the fixed place is employed to capture the waterline in the existing image recognition technology.In order to improve the shortcomings of the existing fixed camera capturing waterline method,such as complicated arrangement and difficulty in observing outboard side,this paper studies the fast and accurate positioning of the ship's waterline area with dynamically changing position in the entire picture and proposes a dynamic identification method of waterline area.Final determination of draught value based on the identified area provides the foundation for accurate and real-time draft survey.1.Proposing dynamic identification method of waterline region based on statistical selection for contourIn order to simplify the complicated arrangement of fixed camera devices,drones are used to capture the waterline of ships.But the waterline area in the images may vary with the unstable flight path.Thus,a dynamic recognition method for waterline area is proposed in this paper.For the dynamically changing waterline area in the image position,a statistic-based selection for contour based on the knowledge of waterline characteristics is employed to locate the interesting horizontal areas containing the waterline.Aiming at the change of illumination conditions and the interference of ship's water trace,this paper carries out color space conversion based on the recognized waterline horizontal area.In the L*a*b color space,combined with K-means++ clustering and watershed algorithm guided by knowledge,the waterline is accurately segmented by using the color and spatial information of pixels.Therefore,the position of waterline is located in the vertical direction,and the task of dynamic identification of interesting waterline area is completed.The experimental results for images under complex environments show that the proposed method can detect the waterline area that changes dynamically in the entire ship image,which not only ensures the integrity of effective information of hull characters,but also reduces the complexity of algorithm.2.Proposing region correction method to deal with different tilt types and achieving the segmentation of large and small hull charactersThe character segmentation effect and the accuracy of draught value recognition are affected by the tilted character caused by the drone shooting angle.An area correction method for different types of area tilt and performs tilt correction on the identified waterline area is summarized in this paper.According to the waterline and ship character trends,the slopes of the vertical and horizontal directions are calculated.Then the source points of the original image are determined to solve the transformation matrix for projective transformation to complete the tilt correction of the waterline area.The experimental results show that the method can deal with different types of regional tilt effectively to complete the tilt correction of the area,such as horizontal tilt mode,vertical miscut mode,horizontal miscut mode,general tilt mode.According to the special arrangement of the hull characters,the hull characters are extracted by projecting in the vertical and horizontal direction from the large and small.The relative positions of the characters are recorded at the same time to provide a basis for the final determination of the draft.3.Achieving the character recognition and the ship draft determinationThe simplified VGG network MiniVGG is selected to recognize the extracted hull waterline characters in this paper.The network has higher recognition accuracy and faster algorithm running speed,and experiments show that it performs well in various classification performance indexes and can effectively recognize the extracted ship characters.The fixed height value of the character is used to convert true distance represented by each pixel in image to achieve distance measurement conversion.The pixel distance between the position of the waterline and the first hull character on the waterline is converted into an actual distance,and the draft of a single picture ship is obtained according to the calculation formula of the draft.Because ships are in the waves during draft survey,it is inevitable that there is a wave surge and resulting the readings of the waterline in fluctuations.The water surface in the actual port environment belongs to the sine wave in the regular wave.The experiment shows that the wave surface fitted out by the single picture draft value shows obvious sine wave characteristics in the time period.The wave is analyzed and the draft value is finally given,which provides a scientific and reasonable reference for the measurement.In this paper,through the research on shortcomings of the existing image recognition technology in detecting ship waterline,a dynamic identification method of waterline region is proposed.The detection of ship waterline and final determination of draught value is completed based on the identified area.Scientific and effective identification methods is provided for managers,which has certain guiding significance for port automation and follow-up measures.The thesis includes 44 figures,8 tables and 87 references.