CN114298881B - Vector map watermark processing method and terminal based on gradient lifting decision tree - Google Patents
Vector map watermark processing method and terminal based on gradient lifting decision tree Download PDFInfo
- Publication number
- CN114298881B CN114298881B CN202111269781.0A CN202111269781A CN114298881B CN 114298881 B CN114298881 B CN 114298881B CN 202111269781 A CN202111269781 A CN 202111269781A CN 114298881 B CN114298881 B CN 114298881B
- Authority
- CN
- China
- Prior art keywords
- vector map
- decision tree
- data
- watermark
- vertex
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Image Processing (AREA)
- Editing Of Facsimile Originals (AREA)
Abstract
The invention provides a vector map watermark processing method and a terminal based on a gradient lifting decision tree, wherein the method comprises the following steps: s101: preprocessing the vector map, and acquiring a maximum disturbance area through a Delaunay triangulation network; s102: embedding watermarks according to the non-feature points in the data, identifying and correcting points to be corrected in the vertexes according to the position relation between the vertexes and the corresponding maximum disturbance areas and constraint conditions, and detecting the watermarks; s103: performing feature engineering processing on the generated training set and test set to train a gradient lifting decision tree model, judging whether the test result of the model meets a preset condition, if so, executing S104, otherwise, executing S103; s104: and predicting the radius of the maximum disturbance area according to the model, and embedding and correcting the watermark in the vector map through the predicted maximum disturbance and constraint conditions. The invention can more efficiently check and correct the problems caused in the watermark embedding process, and ensure the data quality of the vector map while realizing that the watermark is not damaged.
Description
Technical Field
The invention relates to the field of vector map watermarking, in particular to a vector map watermarking method and a vector map watermarking terminal based on a gradient lifting decision tree.
Background
The vector map is an important strategic resource for national economy, national defense construction and social development, not only can accurately express the spatial position of the elements, but also can describe other important spatial characteristics such as topological relation, geometric shape and the like of the elements in detail. From the perspective of national geographic information security, the security problem of the vector map relates to various aspects such as national security, scientific and technological cooperation, intellectual property protection, data sharing and the like. In recent years, research on vector map watermarking technology is becoming a hot spot for research and application in the vector map security field. When specific watermark information is embedded into a vector map, although basic characteristics of digital watermarks such as robustness, invisibility, safety and the like can be satisfied, vertex coordinates of elements in data can be disturbed, and important spatial features such as topological relations and geometric features of the elements are destroyed, so that the usability of the watermark-containing vector map is reduced.
Therefore, in the process of vector map copyright protection research, how to design a high-quality and high-efficiency watermark processing scheme which can meet the requirements of invisibility and robustness of digital watermarks and can also keep the topological relation and the geometric characteristics of a vector map becomes a problem to be solved urgently.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a vector map watermark processing method and a terminal based on a gradient lifting decision tree, which can check and correct data change caused by watermark embedding by combining a maximum disturbance area and constraint conditions, improve the accuracy of training data, optimize the radius of the maximum disturbance area by establishing a gradient lifting decision tree model, improve the efficiency of a vector map coordinate domain watermark processing algorithm, more efficiently check and correct the problems of data topology relation errors, geometric feature change and the like caused in the watermark embedding process, and ensure the data quality of a vector map while preventing the watermark from being damaged.
In order to solve the above problems, the present invention adopts a technical solution as follows: a vector map watermarking processing method based on a gradient boost decision tree comprises the following steps: the vector map watermarking processing method based on the gradient lifting decision tree comprises the following steps: s101: preprocessing a vector map, constructing a Delaunay triangulation network according to the preprocessed data of the vector map, and acquiring the maximum disturbance area of each vertex in the grid through the Delaunay triangulation network; s102: watermark embedding is carried out according to the characteristic points and the non-characteristic points in the data, points to be corrected in the vertexes are identified according to the position relation between the vertexes and the corresponding maximum disturbance areas and constraint conditions, the coordinates of the points to be corrected are corrected, and watermark detection is carried out on the data of the vector map; s103: generating a training set and a test set according to data of a vector map with watermarks detected, performing feature engineering processing on the training set and the test set, training a gradient lifting decision tree model based on the processed and generated features, and judging whether a test result of the trained gradient lifting decision tree model meets a preset condition, if so, executing S104, and if not, executing S103; s104: and predicting the radius of the maximum disturbance area according to the gradient lifting decision tree model, and embedding and correcting the watermark in the vector map according to the predicted maximum disturbance and the constraint condition.
Further, the step of preprocessing the vector map specifically includes: and fusing adjacent polygons in the vector map, and converting the fused data of the vector map into line data.
Further, the step of obtaining the maximum perturbation area of each vertex in the mesh through the Delaunay triangulation network specifically includes: and calculating the radius of an inscribed circle of each triangle in the Delaunay triangulation network, and acquiring the maximum perturbation region of each vertex according to the minimum radius of the inscribed circle in the adjacent triangle of the vertex.
Further, the step of embedding the watermark according to the feature points and the non-feature points in the data specifically includes: and acquiring characteristic points and non-characteristic points in the data through a Douglas-Pock algorithm, and embedding the watermark into the non-characteristic points.
Further, the step of identifying the point to be corrected in the vertex according to the position relationship between the vertex and the corresponding maximum disturbance area and the constraint condition specifically includes: judging whether the vertex is positioned in the maximum disturbance area or not according to the position relation; if so, determining that the vertex is not the point to be corrected; if not, carrying out topology constraint and direction constraint check on the vertex, and identifying the point to be corrected according to the check result.
Further, the step of performing watermark detection on the data of the vector map specifically includes: extracting a global distance sequence of the vector map, reordering the global distance sequence to obtain a first set, obtaining two subsets of the first set, calculating a variance ratio of the two subsets, and obtaining a watermark detection result according to the variance ratio.
Further, the step of performing feature engineering processing on the training set and the test set includes: and respectively carrying out characteristic engineering processing on the training set and the test set to generate characteristics, and forming a training data set and a test data set according to the generated characteristics, wherein the characteristics comprise a variance ratio, the number of non-characteristic points, the abscissa value of the non-characteristic points, the ordinate value of the non-characteristic points, the number of characteristic points, the abscissa value of the characteristic points, the ordinate value of the characteristic points, the maximum change rate of an included angle, the change rate of a perimeter, the change rate of an area and the radius of a maximum disturbance area.
Further, the step of determining whether the test result of the trained gradient boost decision tree model meets a preset condition specifically includes: judging whether the maximum allowable error requirement is met or not according to the test result; if yes, determining that a preset condition is met; if not, determining that the preset condition is not met.
Further, the step of predicting the radius of the maximum perturbation region according to the gradient boosting decision tree model specifically includes: and inputting the data of the preprocessed vector map and the maximum allowable error into the gradient lifting decision tree model, and acquiring the radius of the maximum disturbance area according to an output result.
Based on the same inventive concept, the invention further provides an intelligent terminal, which comprises a processor and a memory, wherein the memory stores a computer program, the processor is in communication connection with the memory, and the vector map watermarking method based on the gradient boosting decision tree is executed according to the computer program.
Compared with the prior art, the invention has the beneficial effects that: the method can check and correct the data change caused by watermark embedding by combining the maximum disturbance area and the constraint condition, improves the accuracy of training data, optimizes the radius of the maximum disturbance area by establishing a gradient lifting decision tree model, improves the efficiency of a vector map coordinate domain watermark processing algorithm, more efficiently checks and corrects the problems of data topology relation errors, geometric feature changes and the like caused in the watermark embedding process, and ensures the data quality of the vector map while preventing the watermark from being damaged.
Drawings
FIG. 1 is a flowchart of an embodiment of a vector map watermarking method based on a gradient boosting decision tree according to the present invention;
FIG. 2 is a flowchart of another embodiment of a gradient boosting decision tree-based vector map watermarking method according to the present invention;
FIG. 3 is a diagram illustrating an embodiment of data preprocessing in a gradient boosting decision tree-based vector map watermarking method according to the present invention;
fig. 4 is a schematic diagram of an embodiment of generating a maximum disturbance area in the vector map watermarking processing method based on the gradient boosting decision tree.
Detailed Description
The present invention will be further described with reference to the accompanying drawings and the detailed description, and it should be noted that any combination of the embodiments or technical features described below can be used to form a new embodiment without conflict.
Referring to fig. 1-4, fig. 1 is a flowchart illustrating a vector map watermarking method based on a gradient boosting decision tree according to an embodiment of the present invention; FIG. 2 is a flowchart of another embodiment of a gradient boosting decision tree-based vector map watermarking method according to the present invention; FIG. 3 is a schematic diagram illustrating an embodiment of data preprocessing performed by the gradient boosting decision tree-based vector map watermarking method according to the present invention; fig. 4 is a schematic diagram of an embodiment of generating a maximum disturbance area in a vector map watermarking method based on a gradient lifting decision tree, where a in fig. 3 is a diagram showing original data of a vector map, b is a diagram showing data after fusion of the vector map, c is a diagram showing line data of the vector map, a diagram a in fig. 4 is a diagram after a constraint triangular net is constructed by the vector map, b is a schematic diagram of a global maximum disturbance area of the vector map, and c is a schematic diagram of a partial maximum disturbance area of the diagram b. The vector map watermarking method based on the gradient boost decision tree of the invention is explained in detail with reference to fig. 1-4.
In this embodiment, the intelligent terminal executing the vector map watermarking method based on the gradient boost decision tree may be a mobile phone, a tablet computer, a notebook computer, a desktop computer, a server, or other intelligent terminals capable of acquiring vector map data and performing watermarking according to the vector map data.
In this embodiment, the vector map watermarking method based on the gradient boosting decision tree includes:
s101: the vector map is preprocessed, a Delaunay triangular network is constructed according to the preprocessed data of the vector map, and the maximum disturbance area of each vertex in the grid is obtained through the Delaunay triangular network.
In this embodiment, the embedding of the watermark will cause the geometric coordinates of the elements to shift, which may cause two adjacent surface elements to overlap or create a "gap". In order to avoid the above situation, the step of preprocessing the vector map specifically includes: and fusing adjacent polygons in the vector map, and converting the data of the fused vector map into line data.
In a specific embodiment, in the data preprocessing process of the vector map, all adjacent polygons of the vector map are fused, wherein fusion is performed by using a fusion tool in the professional software ArcGIS software or C # calls an open source code-fusion function in ArcEngine, and the intersection points between the fused polygon vertex, common edge and polygon boundary line (the common edge is an edge shared by two or more polygons, and the polygon boundary line refers to an edge of a polygon) are extracted, and parameters for fusion are default values. And converting the fused data of the vector map into line data by using a surface line-turning tool in ArcGIS.
In this embodiment, the line data obtained by preprocessing is calculated by a constrained triangulation algorithm to obtain a Maximum Perturbation Region (MPR) of the vertex.
In a specific embodiment, the Delaunay triangulation network is generated using the 3D analysis tool in the ArcGIS software or the nearest neighbor method (thieson polygon method) component in the ArcGIS Engine component library. The process is as follows: the Thiessen polygon uses an extreme boundary interpolation method, and only the nearest single point is used for regional interpolation. The Thiessen polygon divides the region into sub-regions by data point location, each sub-region containing a data point, the distance of each sub-region to the data point within it being less than any distance to other data points, and the value is assigned using the data point within it. The connecting lines connecting all the data points form a Delaunay triangulation network.
The Delaunay triangulation network is a triangulation network, and the thieson polygon method can generate a larger MPR region, so that after watermark embedding, fewer and less drastic corrections are generated; furthermore, it can tolerate more conventional input primitives, such as arcs.
In other embodiments, the Delaunay triangulation network may also be obtained by a distance ranking method.
In this embodiment, the step of obtaining the maximum perturbation area of each vertex in the mesh through the Delaunay triangulation network specifically includes: and calculating the radius of an inscribed circle of each triangle in the Delaunay triangulation network, and acquiring the maximum perturbation region of each vertex according to the minimum radius of the inscribed circle in the adjacent triangle of the vertex. Wherein, the adjacent triangle of the vertex is the triangle sharing the vertex. The minimum inscribed circle radius is determined as the minimum inscribed circle radius.
S102: and embedding watermarks according to the characteristic points and the non-characteristic points in the data, identifying points to be corrected in the vertexes according to the position relation between the vertexes and the corresponding maximum disturbance areas and the constraint conditions, correcting the coordinates of the points to be corrected, and detecting the watermarks of the data of the vector map.
In this embodiment, the step of embedding the watermark according to the feature points and the non-feature points in the data specifically includes: and acquiring feature points and non-feature points in the data after vector map preprocessing by a Douglas-Pock algorithm, and embedding the watermark into the non-feature points. The characteristic points are head and tail points, concave points and convex points of a polygon, the other vertexes are non-characteristic points, and the watermark embedding mode is the prior art.
In a specific embodiment, the watermark embedding is performed by using a coordinate domain vector map watermark algorithm which keeps similar geometric shape, wherein the watermark is embedded into non-feature points of the preprocessed vector map data in a mode of modifying coordinate values of geographic elements, and the watermark embedding step comprises the following steps: (1) extracting a distance sequence from an original vector map as a data carrier to be embedded with a watermark; (2) watermark information is embedded in the extracted distance sequence.
In a specific embodiment, the step of extracting the global distance sequence and reordering comprises: for line data of any one vector map, firstly, feature points of the line data are extracted through a Douglas-Pock algorithm, the sequence of the extracted feature points is defined as { p1, p2, p3, p4, p5, \8230; (pn) }, and other points are non-feature points. Then, all the feature points are connected in sequence to obtain a simplified line sequence { l1, l2, l3, l4, \ 8230;, ln }. And (3) respectively calculating the vertical distance Dn from each non-feature point to the corresponding simplified line ln (the simplified line corresponding to the non-feature point is determined according to the position and line of the non-feature point and the feature point in the online data), and finally respectively constructing a distance sequence Dn = { d1, d2, d3, \ 8230;, dn } from the vertical distance Dn in each simplified line. Obviously, for any vector line layer, the global distance sequence I can be expressed as a formula
Since the order of the global distance sequence I obtained by the initial construction is affected by the reading order of the vector map, in order to make the watermark algorithm have certain robustness to the data order, the global distance sequence I needs to be reordered. Firstly, calculating the geometric center O of all feature points { p1, p2, \ 8230;, pn } and the midpoint coordinate on of each simplified line ln, then sequencing the distances between on and O in sequence, and finally, reordering the global distance sequence I by using the obtained arrangement sequence to finally obtain the reordered distance sequence Io = { Do1, do2, \ 8230;, don }.
Embedding watermark information in the extracted distance sequence includes: a binarized pseudo random noise sequence N of only 0 and 1 is obtained by the key K. All subsets Don of the global distance sequence I are divided into two sets IA and IB by the sequence N. In general, for data from a vector map containing a sufficient number of vertices, the sum of the lengths of IA and IB are very close, and as such, they have similar sample means and sample variances, i.e.
When all the distance values in the set IA are kept unchanged, each distance in the set IB is multiplied by a custom value mu (0 < mu < 1), and the mean value and the variance of the set IB before and after change satisfy the formula
In order to maintain the coordinate accuracy of the vector map, μmust satisfy max (IB) x (1- μ). Ltoreq.τ, max (IB) being the maximum distance in the set IB. Where τ is the maximum allowable error of the original vector map. Meanwhile, when the Douglas-Pock algorithm is applied, the simplified threshold value ThD can be determined by a formula ThD = tau/(1-mu), finally, the sets IA and I' B are combined into the distance sequence IW of the embedded watermark, and the new coordinates of the vertex are calculated by the IW, so that the vector map data GW of the embedded watermark is obtained. Parameters for watermark embedding include: (1) the maximum allowable error tau of the coordinates is 0.2m; (2) the watermark embedding strength factor mu 2 is 0.5; (3) the threshold TD of the douglas-pock algorithm is about 0.683; (4) a binarized pseudo random noise sequence 1000001 (1000001 is the result after Key binarization) of only 0 and 1 is obtained by Key (Key is used to divide the data set into two uniform parts).
In this embodiment, the step of identifying the point to be corrected in the vertex according to the position relationship between the vertex and the corresponding maximum perturbation region and the constraint condition specifically includes: judging whether the vertex is positioned in the maximum disturbance area or not according to the position relation; if yes, determining that the vertex is not the point to be corrected; if not, carrying out topology constraint and direction constraint check on the vertex, and identifying the point to be corrected according to the check result.
Specifically, for the data after watermark embedding, the position relationship between each vertex and its corresponding MPR is determined: (1) If the vertex is located within the MPR, the vertex does not need to be corrected; (2) If not, determining that the vertex is a dirty point, and checking the dirty point by using topological constraint and direction constraint. The topological constraint mainly judges whether the adjacent edge of the vertex is intersected with other elements or the vertex itself in the watermark embedding process (the intersection of the adjacent edge is that another polygon or a suspension point is formed after the edge of one polygon is intersected with the edge), and the direction constraint mainly judges whether the direction of the arc section connected with the vertex is changed excessively (theta is a difference threshold value between the directions of the arc section before and after the watermark is embedded. If the vertex violates one or more of the constraints, the vertex is determined to be a point to be corrected.
In the embodiment, for the point to be corrected which violates the topological constraint and the directional constraint, the vertex is corrected by adopting a coordinate adjustment method based on the topological association of the homonymy point.
In this embodiment, the step of performing watermark detection on the data of the vector map specifically includes: extracting a global distance sequence of the vector map, reordering the global distance sequence to obtain a first set, obtaining two subsets of the first set, calculating a variance ratio of the two subsets, and obtaining a watermark detection result according to the variance ratio, wherein the steps of extracting the global distance and reordering are the same as the above embodiment.
In a specific embodiment, the douglas-pock algorithm is used to extract the global distance sequence of the vector map data, and the douglas-pock algorithm uses the same threshold TD as the watermark embedding above, and obtains the first set Iw according to the result of the global distance sorting. The first set Iw is divided into two subsets IA and IB by the binary pseudo-random noise sequence N obtained using the key K above (the resulting binary numbers are divided into two subsets by the first 0 or 1 by adding the bit operations), and the variances VA and VB of the two subsets are calculated, respectively, and the ratio is denoted as R = VB/VA. If R is less than or equal to Th e (mu) 2 And 1), determining existence of watermark information of a top point corresponding to the global distance sequence, wherein Th is a simplification threshold and is used when a Douglas algorithm is applied. Th = τ/(1- μ), τ being the maximum allowable error of the vector map. Mu is a custom value mu (0 < mu < 1). In order to maintain the coordinate accuracy of the vector map, μmust satisfy max (IB). Times. (1- μ). Ltoreq.τ. If R is not less than or equal to Th e (mu) 2 1), then no watermark information is detected, sayThe vertex is not embedded with the watermark originally and is not processed.
S103: generating a training set and a test set according to data of the vector map with the detected watermark, performing feature engineering processing on the training set and the test set, training a gradient boost decision tree model based on the processed generated features, and judging whether a test result of the trained gradient boost decision tree model meets a preset condition, if so, executing S104, and if not, executing S103.
In a specific embodiment, in the vector map data which is corrected by the watermark and can detect the watermark, 80% of the data is randomly taken as a training set, and the rest 20% of the data is taken as a testing set.
In this embodiment, the step of performing feature engineering processing on the training set and the test set includes: respectively carrying out feature engineering processing on the training set and the test set to generate features, and forming the training data set and the test data set according to the generated features, wherein the features comprise a variance ratio, the number of non-feature points, the abscissa value of the non-feature points, the ordinate value of the non-feature points, the number of feature points, the abscissa value of the feature points, the ordinate value of the feature points, the maximum change rate of an included angle, the change rate of a perimeter, the change rate of an area and the radius of a maximum disturbance area.
In a specific embodiment, performing feature engineering processing includes: taking R values (multiple R values; the distance sequence is a set, the global distance sequence is a set composed of multiple distance sequences, one vector layer has only one global distance sequence, the global distance sequence is ordered, and the distance sequences in the global distance sequence are reordered) obtained in watermark detection as a characteristic x1 1 ,x1 2 ，…，x1 n . The number of non-feature points is taken as the feature x2 1 ,x2 2 ，…，x2 n . The abscissa value of the non-feature point is taken as the feature x3 1 ,x3 2 ，…，x3 n . The ordinate value of the non-feature point is taken as the feature x4 1 ,x4 2 ，…，x4 n . The number of feature points is taken as feature x5 1 ,x5 2 ，…，x5 n . Taking the abscissa value of the feature point as the feature x6 1 ,x6 2 ，…，x6 n . The ordinate value of the feature point is set as feature x7 1 ,x7 2 ，…，x7 n . The uncorrected data of the vector map containing the watermark is compared with the data of the corresponding original vector map, and the maximum change rate of the included angle (the included angle is the minimum positive angle formed by the intersection of two straight lines, one polygon has a plurality of vertexes, and the maximum change rate of the included angle is the maximum one of the change rates of the included angle of each vertex) is taken as the characteristic x8 1 ,x8 2 ，…，x8 n . Comparing the uncorrected watermarked data with the original vector data taking the rate of change of perimeter (the rate of change of perimeter is equal to the absolute value of the difference between the perimeter of the original polygon and the perimeter of the corresponding watermarked polygon divided by the perimeter of the original polygon) as the feature x9 1 ,x9 2 ，…， x9 n . Comparing the uncorrected watermarked vector map data with the corresponding original vector map data takes the area change rate (the area change rate is equal to the absolute value of the difference between the original polygon area and the corresponding watermarked polygon area divided by the original polygon area) as the feature x10 1 ,x10 2 ，…，x10 n . Taking the acquired radius of the maximum disturbance region MPR of each vertex as the characteristic y 1 ,y 2 ，y 3 ，…，y n . Cleaning the above-mentioned characteristics generated by training set to obtain training data set T 1 ＝{(x 1 ,y 1 ),(x 2 ,y 2 ),…,(x n ,y n ) }. Cleaning the above-mentioned characteristics generated by test set to obtain test data set T 2 ＝{(x 1 ,y 1 ),(x 2 ,y 2 ),…,(x n ,y n )}。
In a specific embodiment, the step of training the gradient boosting decision tree model specifically includes: inputting the acquired training data set into a GBTRegegression model of a machine learning module Spark ML of open source software Spark software; setting initial parameters of the model, executing a spark ML program, performing model training, and recording AUC (Area Under Current, AUC is a performance index for measuring the quality of the machine learning model, and the value range of AUC is between 0.5 and 1. The AUC is closer to 1.0, the model effect is better, and when the AUC is equal to 0.5, the model effect is lowest and has no application value). Continuously adjusting parameters of the model for training; and selecting the model with the highest AUC value for model test.
In this embodiment, the step of determining whether the test result of the trained gradient boost decision tree model satisfies the preset condition specifically includes: judging whether the maximum allowable error requirement is met or not according to the test result; if yes, determining that a preset condition is met; if not, determining that the preset condition is not met. And acquiring a test result by inputting data in the test data set into the gradient lifting decision tree model.
S104: and predicting the radius of the maximum disturbance area according to the gradient lifting decision tree model, and embedding and correcting the watermark in the vector map according to the predicted maximum disturbance and the constraint condition.
In this embodiment, the step of predicting the radius of the maximum perturbation region according to the gradient boosting decision tree model specifically includes: and inputting the data of the preprocessed vector map and the maximum allowable error into a gradient lifting decision tree model, and acquiring the radius of the maximum disturbance area according to an output result.
After the radius of the maximum disturbance area is obtained through prediction, watermark embedding and correction are carried out on the vector map according to the maximum disturbance area and the constraint condition, wherein the watermark embedding and correction mode is the same as the watermark embedding and correction mode adopted in the embodiment.
The method and the device embed the watermark information by utilizing the MPR, the direction constraint and the vertex coordinate adjusted vector map coordinate domain watermark algorithm, and optimize the MPR based on the gradient lifting decision tree algorithm, thereby more efficiently and more accurately checking and correcting the problems of data topology relation errors, geometric feature changes and the like caused in the watermark embedding process, and realizing the purpose of ensuring the data quality of the vector map while not damaging the watermark.
Based on the same inventive concept, the invention also provides an intelligent terminal.
In this embodiment, the intelligent terminal includes a processor and a memory, the memory stores a computer program, the processor is connected to the memory in a communication manner, and the vector map watermarking method based on the gradient boost decision tree according to the computer program is executed.
In some embodiments, memory may include, but is not limited to, high speed random access memory, non-volatile memory. Such as one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid-state storage devices. The Processor may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the Integrated Circuit may also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, or a discrete hardware component.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. A vector map watermarking processing method based on a gradient boost decision tree is characterized by comprising the following steps:
s101: preprocessing a vector map, constructing a Delaunay triangular network according to the preprocessed data of the vector map, and acquiring the maximum disturbance area of each vertex in the grid through the Delaunay triangular network;
s102: watermark embedding is carried out according to characteristic points and non-characteristic points in the data preprocessed by the vector map, points to be corrected in the vertexes are identified according to the position relation between the vertexes and the corresponding maximum disturbance areas and constraint conditions, coordinates of the points to be corrected are corrected, and watermark detection is carried out on the data of the vector map;
s103: generating a training set and a test set according to data of a vector map with watermarks detected, performing feature engineering processing on the training set and the test set, training a gradient lifting decision tree model based on the processed and generated features, and judging whether a test result of the trained gradient lifting decision tree model meets a preset condition, if so, executing S104, and if not, executing S103;
s104: and predicting the radius of the maximum disturbance area according to the gradient lifting decision tree model, and embedding and correcting the watermark in the vector map according to the predicted maximum disturbance and the constraint condition.
2. The gradient boosting decision tree-based vector map watermarking method according to claim 1, wherein the step of preprocessing the vector map specifically includes:
and fusing adjacent polygons in the vector map, and converting the fused data of the vector map into line data.
3. The gradient boosting decision tree-based vector map watermarking method according to claim 1, wherein the step of obtaining the maximum perturbation region of each vertex in the mesh through the Delaunay triangulation network specifically comprises:
and calculating the radius of an inscribed circle of each triangle in the Delaunay triangulation network, and acquiring the maximum perturbation region of each vertex according to the minimum radius of the inscribed circle in the adjacent triangle of the vertex.
4. The vector map watermarking method based on the gradient boosting decision tree as claimed in claim 1, wherein the step of performing watermark embedding according to the feature points and non-feature points in the data preprocessed by the vector map specifically includes:
and acquiring characteristic points and non-characteristic points in the data through a Douglas-Pock algorithm, and embedding the watermark into the non-characteristic points.
5. The vector map watermarking method based on the gradient boosting decision tree as claimed in claim 1, wherein the step of identifying the point to be corrected in the vertex according to the position relationship between the vertex and the corresponding maximum perturbation region and the constraint condition specifically includes:
judging whether the vertex is positioned in the maximum disturbance area or not according to the position relation;
if so, determining that the vertex is not the point to be corrected;
if not, carrying out topology constraint and direction constraint check on the vertex, and identifying the point to be corrected according to the check result.
6. The vector map watermarking method based on the gradient boosting decision tree as claimed in claim 1, wherein the step of performing watermark detection on the data of the vector map specifically includes:
extracting a global distance sequence of the vector map, reordering the global distance sequence to obtain a first set, obtaining two subsets of the first set, calculating a variance ratio of the two subsets, and obtaining a watermark detection result according to the variance ratio.
7. The vector map watermarking method based on the gradient boosting decision tree as claimed in claim 6, wherein the step of performing feature engineering processing on the training set and the test set comprises:
and respectively carrying out characteristic engineering processing on the training set and the test set to generate characteristics, and forming a training data set and a test data set according to the generated characteristics, wherein the characteristics comprise a variance ratio, the number of non-characteristic points, the abscissa value of the non-characteristic points, the ordinate value of the non-characteristic points, the number of characteristic points, the abscissa value of the characteristic points, the ordinate value of the characteristic points, the maximum change rate of an included angle, the change rate of a perimeter, the change rate of an area and the radius of a maximum disturbance area.
8. The gradient boosting decision tree-based vector map watermarking method according to claim 1, wherein the step of judging whether the test result of the trained gradient boosting decision tree model satisfies a preset condition specifically comprises:
judging whether the maximum allowable error requirement is met or not according to the test result;
if so, determining that a preset condition is met;
if not, determining that the preset condition is not met.
9. The gradient boosting decision tree-based vector map watermarking method according to claim 1, wherein the step of predicting the radius of the maximum perturbation area according to the gradient boosting decision tree model specifically comprises:
and inputting the data of the preprocessed vector map and the maximum allowable error into the gradient lifting decision tree model, and acquiring the radius of the maximum disturbance area according to an output result.
10. An intelligent terminal, characterized in that the intelligent terminal comprises a processor, a memory, the memory stores a computer program, the processor is connected with the memory in communication, according to the computer program, the vector map watermarking method based on gradient boosting decision tree according to any one of claims 1-9 is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111269781.0A CN114298881B (en) | 2021-10-29 | 2021-10-29 | Vector map watermark processing method and terminal based on gradient lifting decision tree |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111269781.0A CN114298881B (en) | 2021-10-29 | 2021-10-29 | Vector map watermark processing method and terminal based on gradient lifting decision tree |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114298881A CN114298881A (en) | 2022-04-08 |
| CN114298881B true CN114298881B (en) | 2023-01-06 |
Family
ID=80964515
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111269781.0A Active CN114298881B (en) | 2021-10-29 | 2021-10-29 | Vector map watermark processing method and terminal based on gradient lifting decision tree |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114298881B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101149835A (en) * | 2007-10-29 | 2008-03-26 | 中国人民解放军信息工程大学 | Map data rasterizing based robust blind water mark embedding and extraction method |
| CN108961141A (en) * | 2018-07-04 | 2018-12-07 | 张新长 | Dual Zero watermarking method, system, storage medium and the server of map vector |
| CN109409499A (en) * | 2018-09-20 | 2019-03-01 | 北京航空航天大学 | One kind being based on deep learning and the modified track restoration methods of Kalman filtering |
| CN111707238A (en) * | 2020-05-29 | 2020-09-25 | 广东省国土资源测绘院 | Method and system for generating aviation digital orthophoto map |
| CN113284036A (en) * | 2021-06-15 | 2021-08-20 | 苏州科技大学 | Vector map reversible watermarking algorithm based on virtual grid interval mapping |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111242825B (en) * | 2019-12-17 | 2022-09-23 | 中国人民解放军海军大连舰艇学院 | Water depth feature-based ENC electronic nautical chart zero-watermarking method |
| CN111784084B (en) * | 2020-08-17 | 2021-12-28 | 北京市城市规划设计研究院 | Travel generation prediction method, system and device based on gradient lifting decision tree |
-
2021
- 2021-10-29 CN CN202111269781.0A patent/CN114298881B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101149835A (en) * | 2007-10-29 | 2008-03-26 | 中国人民解放军信息工程大学 | Map data rasterizing based robust blind water mark embedding and extraction method |
| CN108961141A (en) * | 2018-07-04 | 2018-12-07 | 张新长 | Dual Zero watermarking method, system, storage medium and the server of map vector |
| CN109409499A (en) * | 2018-09-20 | 2019-03-01 | 北京航空航天大学 | One kind being based on deep learning and the modified track restoration methods of Kalman filtering |
| CN111707238A (en) * | 2020-05-29 | 2020-09-25 | 广东省国土资源测绘院 | Method and system for generating aviation digital orthophoto map |
| CN113284036A (en) * | 2021-06-15 | 2021-08-20 | 苏州科技大学 | Vector map reversible watermarking algorithm based on virtual grid interval mapping |
Non-Patent Citations (9)
| Title |
|---|
| 3D隐写模型与方法研究;周航;《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》;20210115;全文 * |
| BlackMarks: Blackbox Multibit Watermarking for Deep Neural Networks;Huili Chen 等;《arXiv:1904.00344》;20190331;全文 * |
| Topology-Preserving and Geometric Feature-Correction Watermarking of Vector Maps;Xu Xi 等;《IEEE Access》;20200212;全文 * |
| TOPOLOGY-PRESERVING WATERMARKING OF VECTOR GRAPHICS;STEFAN HUBER 等;《International Journal of Computational Geometry & Applications》;20141231;全文 * |
| Watermarking 3D CAPD models for topology verification;ZhiyongSu 等;《Computer-Aided Design》;20130731;第45卷(第7期);全文 * |
| 二维矢量地图脆弱水印技术研究;王娜娜;《中国博士学位论文全文数据库 信息科技辑》;20180615;全文 * |
| 保持几何形状的矢量地图水印算法应用分析;杨娜娜 等;《北京测绘》;20210430;第35卷(第4期);全文 * |
| 保持拓扑关系与几何特征的矢量地图水印处理方法;杨娜娜 等;《测绘科学技术学报》;20181231;第35卷(第4期);全文 * |
| 海量数据组织与管理研究的现状与展望;洪镇填;《地理空间信息》;20141231;第12卷(第5期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114298881A (en) | 2022-04-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Salinas et al. | Structure‐aware mesh decimation | |
| Chao et al. | A high capacity 3D steganography algorithm | |
| JP2984633B2 (en) | Reference image creation method and pattern inspection device | |
| CN109472870B (en) | Model matching method based on grid reconstruction and multi-influence-domain correction | |
| Zhang et al. | Dual contouring for domains with topology ambiguity | |
| CN106934842B (en) | Method and device for identifying closed area | |
| JP2009211626A (en) | Image processing device and method | |
| US20220392139A1 (en) | Ordering of Patch Selection in Tessellation Operations | |
| CN109886279A (en) | Image processing method, device, computer equipment and storage medium | |
| KR102158390B1 (en) | Method and apparatus for image processing | |
| CN110569845A (en) | test paper image correction method and related device | |
| US9131193B2 (en) | Image-processing device removing encircling lines for identifying sub-regions of image | |
| CN114298881B (en) | Vector map watermark processing method and terminal based on gradient lifting decision tree | |
| TWI353562B (en) | A video processing apparatus and related method to | |
| CN113284036B (en) | Vector map reversible watermarking algorithm based on virtual grid interval mapping | |
| JP4756436B2 (en) | Pattern recognition apparatus, pattern recognition method, and pattern recognition program | |
| JP2006337167A (en) | Inspection method of low-contrast flaw under cyclic noise, and inspection method of low-contrast flaw under repeated pattern | |
| US10679049B2 (en) | Identifying hand drawn tables | |
| CN114616571A (en) | Semiconductor layout environment around points of interest | |
| CN108876691B (en) | Self-adaptive reversible watermarking method and system | |
| CN111179375B (en) | Automatic generation method and device of contour graph of building bearing platform and computer storage medium | |
| CN113763558A (en) | Information processing method and device, equipment and storage medium | |
| CN112199460A (en) | Vector map long and narrow arc segment identification method | |
| KR100965843B1 (en) | A method on extracting the boundary vector of object obtained from laser scanning data | |
| TWI407365B (en) | Method for data clustering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |