Electronic chart system water depth point dense query method

Abstract

The invention relates to a water depth point dense query method for an electronic chart system, which can overcome the defect that the availability of a water depth point query function in the existing electronic chart system is limited. Comprising the following steps: (1) water depth data extraction: reading water depth map layer data from a chart file, and obtaining and storing a water depth point record set; (2) gridding processing: establishing a gridding model by using the scattered water depth data, setting step lengths in horizontal and vertical directions, and selecting a proper starting point to draw a grid; and (3) water depth data reconstruction: firstly, adopting a missing value complementing technology to complement missing values in the water depth grid to form a complete water depth grid model, and then adopting a super-resolution technology to improve the water depth point density to form a high-resolution water depth grid model. And (4) water depth data query: when a user clicks a screen, obtaining a water depth grid model obtained by reconstruction, comparing geographic coordinates at the clicked position with geographic coordinates in the grid model, and returning a query result.

Description

Technical Field

[0001] The invention relates to the field of marine information processing, and in particular to a method for densely querying water depth points in an electronic nautical chart system. Background Art

[0002] In electronic charts, a sounding point is a data point that represents the depth of a specific location at the bottom of the ocean or water body. This data is crucial to fields such as navigation safety, hydrographic surveys, marine engineering, and marine scientific research. In electronic chart systems, the depth value of the point is usually displayed at the screen coordinates corresponding to the sounding point position in the chart file data. The sounding point query function in the electronic chart system means clicking the screen coordinates with the mouse. If there is a sounding point at that location, the query returns the sounding point value. If there is no sounding point at that location, a null value is returned. However, due to the limited number of sounding points in electronic chart data files, the sounding values ​​can only be obtained at limited locations. The query results for most locations are null values, and the availability of the query function is limited. Summary of the invention

[0003] The purpose of the present invention is to propose a method for dense query of water depth points in an electronic chart system to address the defect of limited availability of water depth point query function in an electronic chart system. The method is a complete set of dense query solutions for water depth points, including water depth data extraction, water depth data gridding processing, water depth data reconstruction and water depth data query. The technical solution is as follows: A method for densely querying water depth points in an electronic chart system: (1) Extraction of water depth data. Read water depth data from the chart file. Electronic chart data is divided into layers. Water depth data is stored in the "SOUNDG" layer of the chart, which is a point layer. Read the layer data, obtain the water depth point record set, extract the data in the water depth point set and store it. (2) Grid processing. Use scattered water depth data to build a grid model. Set the specified step size in the horizontal and vertical directions, select a suitable starting point to draw a straight line to form a grid. Each water depth point is scattered in a different grid to generate a grid model. (3) Depth data reconstruction. Depth data reconstruction is divided into two stages: missing value completion and super-resolution. First, the missing value completion technology is used to complete the missing values ​​in the depth grid to form a complete depth grid model. Then, the super-resolution technology is used to increase the density of depth points to form a high-resolution depth grid model. The missing value completion is performed using a conditional generative adversarial network, and the super-resolution is performed using a Transformer network. (4) Water depth data query. When the user clicks on the screen, the screen coordinates of the query point are converted into projection coordinates and then into geographic coordinates according to the geographic coordinate system used by the electronic nautical chart system. The reconstructed water depth grid model is obtained, and the geographic coordinates of the clicked point are compared with the geographic coordinates in the grid model. If the comparison is successful, the geographic coordinates and water depth point value of the point are used as the query results and highlighted on the nautical chart. The beneficial effects of the present invention are: (1) The present invention ensures the integrity of water depth grid data through spatial interpolation technology, improves the accuracy of water depth data through super-resolution technology, and finally realizes intensive query of water depth points. (2) The present invention can more effectively utilize existing water depth data resources through a dense query method, reduce the need for additional data collection, reduce data acquisition costs, and improve data processing efficiency. (3) The present invention supports users to quickly obtain water depth information through simple click operations, thereby improving the convenience and intuitiveness of user interaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0004] Figure 1 It is the overall structure diagram of the present invention; DETAILED DESCRIPTION

[0005] To make the technical solution of the present invention clearer, the present invention is further described below in conjunction with the accompanying drawings. The present invention is specifically implemented according to the following steps: (1) Water depth data extraction Electronic chart data is divided into information layers. The water depth data is stored in the "SOUNDG" layer of the chart, which is a point layer. The data needs to be extracted for reconstruction of the elevation model. The water depth point extraction process is: read the "SOUNDG" layer of the electronic chart data, which is the water depth point layer, obtain the layer record set, traverse the set, and obtain the geometric information of each record object, that is, the water depth point set. Extract the water depth point set data and store the set data. (2) Grid processing The grid model is established by using scattered water depth data. The grid structure is simple and has uniform resolution, which is convenient for retrieval and storage. It is also easy to combine with navigation data and has good practicality. The grid processing method used is: first set the specified step size in the horizontal and vertical directions, select a suitable starting point to draw a straight line to form a grid, and each water depth point is scattered in different grids to generate a grid model. The generated grid model is represented by the following matrix: Among them, Z 0,0 Indicates the elevation value of the upper left corner of the chart, Z i,j Indicates the elevation value of the lower right corner of the chart. (3) Depth data reconstruction The reconstruction of water depth data is divided into two stages: missing value completion and super-resolution. Missing value completion refers to finding the elevation value of an unknown point based on several adjacent known elevation data points. In the real marine elevation measurement scenario, the measured elevation data is often scattered and cannot directly form a regular DEM, so it needs to be reconstructed through missing value completion. Super-resolution refers to the use of low-resolution water depth data to obtain high-resolution water depth data. This method can obtain high-resolution water depth data without increasing additional measurement costs and ensuring the quality of reconstruction. The water depth data is reconstructed using a deep learning method, each coordinate point is regarded as the pixel position of a two-dimensional image, and the height value is regarded as the image grayscale value, so that the water depth data reconstruction can be converted into a two-dimensional image reconstruction. The missing value completion is performed using a conditional generative adversarial network. The network is trained using a public ocean elevation dataset, the original water depth image is cropped, and the sampled data is obtained by random sampling in proportion as scattered water depth data. The sampled data is input into the network generator as conditional data to guide the generator to generate outputs that meet the conditional description. The sampled data, original data, and generated data are spliced ​​and sent to the discriminator. The discriminator needs to determine whether it is original data or generated data, and at the same time identify whether they conform to the distribution of the sampled data. In terms of the network model, classic models such as cycleGAN and styleGAN can be used. In order to better restore the terrain features, the slope and aspect indicators in the DEM data are introduced into the loss function to make the generated image more consistent with the original terrain, and z is used. i,j represents the elevation value of the (i, j) position in the DEM data, cellsize represents the resolution of the image, dx and dy represent the slope in the horizontal and vertical directions respectively, and the slope value of the (i, j) position is slope i,j The formula is: The slope aspect indicates the direction of the slope, which is usually expressed as the angle from the x-axis clockwise to the slope vector, ranging from 0 to 360°. Each DEM data can generate a slope map and an aspect map. The more similar the slope map and aspect map of the generated data are to the real data, the more the generated terrain conforms to the original terrain. Super-resolution is performed using the Transformer network. The network is trained using a public ocean elevation dataset, and the original water depth image is downsampled to obtain low-resolution water depth data. The low-resolution data is fed into the Tramsformer network to guide the network to generate a high-resolution image that matches the original image. The network uses a reconstruction model based on the Swin Transformer, which is divided into a shallow feature extraction module, a deep feature extraction module, and a reconstruction module. The shallow feature extraction module consists of a convolutional layer to extract low-frequency information in the image, and the deep feature extraction module consists of a Swin Transformer block to extract deep information in the image. The reconstruction module fuses shallow and deep features to generate the final high-resolution image. Transformer can overcome the problem that convolution cannot pay attention to the global terrain and improve the expression ability of the model. In addition, in order to solve the irreversible texture degradation problem caused by downsampling, a reference image mechanism is introduced, and images with the same terrain are selected as reference images. Feature matching is performed on low-resolution images and reference images to provide additional terrain information for high-resolution image reconstruction, helping to improve the details and clarity of the target image. In the improved attention module, the low-resolution image is upsampled to obtain an image of the same size as the high-resolution image, corresponding to Q in the attention mechanism. A reference image of the same size and the same terrain area is selected, corresponding to V in the attention mechanism. The reference image is downsampled and upsampled in turn to obtain a reference image of unchanged size but degraded features, corresponding to K in the attention mechanism. (4) Water depth data query The water depth data query function supports users to query the water depth point values ​​on the nautical chart by clicking. The query process is as follows: after turning on the query function, the user clicks on the nautical chart interface to obtain the screen coordinates of the clicked point; according to the geographic coordinate system used by the electronic nautical chart system, the screen coordinates of the query point are converted into projection coordinates and then into geographic coordinates; the reconstructed water depth grid model is obtained, and the geographic coordinates of the clicked point are compared with the geographic coordinates in the grid model. After the comparison is successful, the geographic coordinates and water depth point values ​​at that point are used as the query results and highlighted on the nautical chart.

Claims

1. A method for densely querying water depth points in an electronic nautical chart system, characterized in that: It includes three parts: depth data processing, depth data reconstruction and depth data query. It reads depth layer data from chart files, obtains a set of depth point records, extracts data from the depth point set and stores it, and uses scattered depth data to establish a grid model. It uses missing value completion technology to complete the missing values ​​in the depth grid to form a complete depth grid model, and uses super-resolution technology to increase the density of depth points to form a high-resolution depth grid model. It supports users to point-select and query depth points on the electronic chart interface.

2. The method for densely querying water depth points in an electronic nautical chart system according to claim 1, characterized in that: Depth data processing: read depth data from the chart file. The depth data is stored in the "SOUNDG" layer of the chart, which is a point layer. Read the layer, obtain the depth point record set, extract the data in the depth point set and store it, and use the scattered depth data to establish a grid model. First, set the specified step size in the horizontal and vertical directions, select a suitable starting point to draw a straight line to form a grid. Each depth point is scattered in different grids to generate a grid model.

3. The method for densely querying water depth points in an electronic nautical chart system according to claim 1, characterized in that: Depth data reconstruction: deep learning methods are used to reconstruct depth data. Each coordinate point is regarded as the pixel position of a two-dimensional image, and the height value is regarded as the grayscale value of the image. The depth data reconstruction is converted into two-dimensional image reconstruction. The missing value is completed using a conditional generative adversarial network. The original depth image is randomly sampled to obtain the sampled data. As scattered depth data, the sampled data is input into the generator of the network as conditional data to guide the generator to generate outputs that meet the conditional description. In terms of the network model, classic models such as cycleGAN and styleGAN can be used. In order to better restore the terrain features, the slope and aspect indicators in the DEM data are introduced into the loss function to make the generated image more consistent with the original terrain. Super-resolution uses a Swin-based The reconstruction model of Transformer is divided into a shallow feature extraction module, a deep feature extraction module and a reconstruction module. The shallow feature extraction module is composed of convolutional layers, which overcomes the problem that convolution cannot pay attention to the global terrain and improves the expression ability of the model. In order to solve the irreversible texture degradation problem caused by downsampling, a reference image mechanism is introduced. Images with the same terrain are selected as reference images, and feature matching is performed on low-resolution images and reference images to provide additional terrain information for high-resolution image reconstruction, helping to improve the details and clarity of the target image.

4. The method for densely querying water depth points in an electronic nautical chart system according to claim 1, characterized in that: Depth data query: The depth data query function supports users to query the depth point values ​​on the chart by clicking on them. The query process is as follows: after the query function is turned on, the user clicks on the chart interface to obtain the screen coordinates of the clicked point. According to the geographic coordinate system used by the electronic chart system, the screen coordinates of the query point are converted into projection coordinates and then into geographic coordinates; Get the reconstructed water depth grid model, compare the geographic coordinates of the clicked location with the geographic coordinates in the grid model, and if the comparison is successful, use the geographic coordinates and water depth point value of the location as the query result and highlight them on the nautical chart.

Images