High-definition map watermark embedding and extraction method, device and medium

Abstract

The application provides a high-precision map watermark embedding and extracting method, device and medium, and relates to the technical field of digital geographic information. The watermark embedding method comprises the following steps: obtaining an original map containing helical line segments, circular arc segments, lane width, marking line width and lane offset; constructing a multi-dimensional carrier pool by extracting geometric segments and attributes; embedding the watermark by using absolute quantization index modulation; writing back the attribute values after quantizing the attribute carriers, calculating the geometric scaling invariant of the geometric carriers and obtaining target characteristic values after quantization; performing geometric constraint optimization on the helical line segments by taking length and starting curvature increment as variables to obtain the optimized length and end point coordinates, and directly calculating the new length of the circular arc segments; taking the subsequent segment of the geometric segment with length or end point coordinate change as a to-be-corrected object, accumulating the previous segment to update the length, calculating the new starting mileage and synchronously updating the mileage values of the auxiliary elements, and obtaining a map containing watermarks and topologically consistent. The obtained map can resist various attacks and maintain map precision.

Description

Technical Field

[0001] This invention relates to the field of digital geographic information technology, and in particular to a method, device and medium for embedding and extracting watermarks on high-precision maps. Background Technology

[0002] With the rapid development of intelligent connected vehicle technology, high-precision maps, as the core digital infrastructure of autonomous driving systems, have deeply penetrated the environmental perception, path planning, and decision-making control stages of vehicles due to their centimeter-level high-precision spatial information and rich road semantic logic. The production of high-precision maps involves high costs for surveying, data collection, and manual annotation, and has extremely high commercial value, making it a core object of copyright protection in the field of digital geographic information.

[0003] However, the openness and high value of high-precision map data make it face serious security risks. On the one hand, the illegal copying, distribution and infringement of data will bring huge economic losses to map providers; on the other hand, as the current international mainstream high-precision map exchange format, OpenDRIVE uses XML plaintext storage, which is easily read, parsed and saved as a new file by general text editors, making it difficult to define the ownership of data. Traditional encryption technology can only ensure the confidentiality of data during transmission and storage, but cannot solve the copyright ownership problem when the data is exposed in plaintext form in memory or application after decryption [4]. In contrast, digital watermarking technology can embed hidden information without changing the data format and usage attributes, and has become an effective means of realizing copyright confirmation.

[0004] Vector geographic data digital watermarking technology has achieved remarkable results in both theory and application after more than 20 years of development. Among the mainstream transform domain watermarking methods, existing techniques mostly utilize mathematical transformations to achieve frequency domain embedding in order to improve concealment. For example, the literature "Solachidis V, Pitas I. Watermarking polygonal lines using Fourier descriptors [J]. IEEE Computer Graphics and Applications, 2004, 24(3): 44-51.", "Xu Dehe, Zhu Changqing, Wang Zhiming. DFT-based blind watermarking algorithm for vector geographic data [J]. Acta Geodaetica et Cartographica Sinica, 2010, 39(4): 425-429.", and "Zhu Changqing, Xu Dehe. Fourier transform of vector geographic data and its application in digital watermarking [J]. Science of Surveying and Mapping, 2010, 35(4): 5-7." utilizes Discrete Fourier Transform (DFT) to convert spatial coordinates into complex form for watermarking; the literature "Yang Chengsong, Zhu Changqing. Watermarking algorithm for vector geographic data based on Discrete Cosine Transform [J]. Science of Surveying and Mapping, 2008, 33(5): 108-110.", and "Yang Chengsong, Zhu Changqing, Zhang Ruolin. Vector map watermarking algorithm based on second-generation wavelet transform [J]. Journal of Image and Graphics, 2010, 15(3): 511-516.》Hidden information in low-frequency coefficients based on discrete cosine transform (DCT) and discrete wavelet transform; the paper "Ren Na, Zhu Changqing, Lan Xuhui. Zero watermarking algorithm for vector geographic data based on singular value decomposition [J]. Acta Geodaetica et Cartographica Sinica, 2014, 43(4): 423-429." proposes a watermarking scheme based on singular value decomposition (SVD). However, such transform domain methods are essentially mathematical perturbations of spatial domain coordinates, making it difficult to achieve absolute control over coordinate deviations. High-precision maps require centimeter-level absolute spatial accuracy. Any unconstrained coordinate offset will destroy the smoothness of lane lines and road network topology, bringing fatal safety hazards to autonomous driving.

[0005] To avoid the accuracy loss caused by coordinate perturbation, existing technologies have proposed watermarking methods based on spatial domain geometric features. For example, the paper "Wang X, Niu X. A distance-based watermarking algorithm for vector maps using Hilbert curve [J]. Multimedia Tools and Applications, 2017, 76(2): 1851-1869." uses Hilbert curves to establish feature pairs and embeds watermarks by modifying the relative distance between feature pairs. In addition, methods using topological relationships such as polygon area ratio or the angle between adjacent line segments have also been proposed. However, most of these methods are aimed at simple line and surface features composed of discrete point coordinates in traditional GIS. Unlike traditional vector maps, OpenDRIVE high-precision maps adopt highly structured parametric representations, and their road reference lines rely heavily on continuous mathematical equations such as spirals, arcs, and cubic polynomials. Traditional geometric feature algorithms cannot be compatible with such parametric mapping logic, resulting in insufficient carriers when dealing with long straight roads or simple road network scenarios common in high-precision maps.

[0006] In recent years, research on digital watermarking for high-precision maps in OpenDRIVE format has gradually emerged, and existing technologies have utilized the XML text structure characteristics of OpenDRIVE for information hiding. For example, the literature "19" discloses a text information hiding method based on XML structure, attempting to embed data using the arrangement order of tag attributes. The literature "Li Zhaocan, Wang Liming, Ge Sijiang, et al. A big data plain text watermarking method based on orthogonal coding [J]. Computer Science, 2019, 46(12): 148-154." proposes a big data plain text watermarking method based on orthogonal coding in the field of text watermarking, embedding identification information through specific encoding rules. In order to enhance concealment, the literature "Zhang Zhenyu, Li Qianmu, Qi Yong. Text watermarking design based on invisible characters [J]. Journal of Nanjing University of Science and Technology, 2017, 41(4): 405-411." proposes a text watermarking design idea based on invisible characters, utilizing the characteristics of Unicode encoding for information steganography. Furthermore, the paper "Lü Xuchao, Ren Na, Zhou Qifei, Zhu Changqing. Digital watermarking algorithm for OpenDrive format high-precision maps with invisible characters [J]. Journal of Geoinformation Science, 2024, 26(9): 2026-2037." proposes a digital watermarking method based on invisible characters for OpenDRIVE high-precision maps. This method uses zero-width characters as the watermark carrier, embedding them into XML tag attributes while ensuring the usability of the high-precision map data. However, steganography based on text encoding or structure inevitably suffers from robustness defects. During the distribution and use of high-precision map data, operations such as format cleaning and cross-format conversion are frequently encountered, which can completely erase the watermark information, resulting in extremely poor robustness.

[0007] In summary, while watermarking methods based on the transform domain offer good concealment, they are prone to compromising the absolute spatial accuracy of high-precision maps. Watermarking methods based on spatial geometric features can avoid accuracy loss, but they face the problem of limited carriers in the parametric representation of high-precision maps. Watermarking methods based on text structure or encoding achieve zero coordinate loss, but watermark loss can occur during operations such as format cleaning and cross-format conversion. Summary of the Invention

[0008] This invention provides a high-precision map watermark embedding and extraction method, device, and medium, aiming to solve the problems of existing watermarking technologies, such as difficulty in balancing geometric accuracy and anti-attack capabilities, lack of geometric carriers under parametric representation, and the susceptibility of text-based watermarks to erasure during format cleaning and cross-format conversion. By constructing a multi-dimensional carrier pool containing geometric shapes and road attributes, and using absolute quantization index modulation to achieve orthogonal decoupled embedding of multiple carriers, and combining Levenberg-Marquardt optimization and topology chain correction strategies, this invention achieves stable resistance to strong attacks such as geometric transformations, extreme clipping, and format cleaning while ensuring map geometric continuity and topological consistency, thus meeting the safety application requirements of autonomous driving systems.

[0009] In a first aspect, the present invention provides a method for embedding watermarks in high-precision maps, the method comprising:

[0010] Acquire raw high-precision map data, which includes road reference line geometric segments, lane width attributes, road marking width attributes, and lane center offset attributes;

[0011] Based on the hierarchical structure of the original high-precision map data, spiral segments and arc segments are extracted from the geometric segments of the road reference lines as geometric shape carriers, and lane width attributes, road marking width attributes, and lane center offset attributes are extracted as road attribute carriers to construct a multi-dimensional carrier pool including geometric shape carriers and road attribute carriers.

[0012] The original copyright information is converted into a watermark bitstream through error correction encoding. An absolute quantization index modulation strategy is used to embed the watermark into each carrier in the multidimensional carrier pool. Specifically, for road attribute carriers, the absolute physical values ​​of the road attribute carriers are quantized and modulated to obtain modified attribute values, and the modified attribute values ​​are written back to the corresponding attribute nodes to complete the watermark embedding of the road attribute carriers. For geometric shape carriers, the geometric scaling invariant of the geometric shape carriers is calculated, and the geometric scaling invariant is quantized and modulated to obtain the target feature value.

[0013] For the quantized and modulated spiral segment carrier, the target termination curvature is pre-calculated based on the target feature value, and geometric constraint optimization is performed using the length and initial curvature increment as optimization variables to obtain the optimized spiral segment length and the optimized spiral segment endpoint coordinates; for the arc segment carrier, the new length of the arc segment is calculated based on the target feature value.

[0014] The geometric segments whose lengths have changed relative to the original lengths of the corresponding geometric segments, or whose endpoint coordinates have changed relative to the original endpoint coordinates of the corresponding geometric segments, are designated as modified geometric segments. The subsequent geometric segments of these modified segments are designated as objects to be corrected. The updated lengths of the preceding geometric segments of the objects to be corrected are accumulated to calculate the new starting mileage of the objects to be corrected. Simultaneously, the mileage values ​​of the associated auxiliary data elements of the objects to be corrected are updated. This process transmits and compensates for the geometric deformation caused by watermark embedding along the topological chain, resulting in high-precision map data with embedded watermarks.

[0015] In one possible design, the geometric scaling invariant... Calculate using the following formula:

[0016] ;

[0017] in, Let be the curvature of the arc segment. Let be the terminal curvature of the helical segment. The length of the current geometric segment is denoted as Spiral, where Spiral is a spiral segment and Arc is a circular arc segment; the geometric scaling invariant remains constant when the map is subjected to a global scaling attack.

[0018] In one possible design, the watermark embedding method for each carrier in the multidimensional carrier pool using an absolute quantization index modulation strategy includes:

[0019] The target feature value is calculated using the following formula:

[0020] ;

[0021] in, To quantize the step size, V represents the binary watermark bits to be embedded, and V is the original feature value of the current carrier. The target feature value is defined by Round, which is the rounding function.

[0022] For the road attribute carrier, the modified attribute value is Write back directly to the data;

[0023] For a geometrically shaped carrier, the target feature value Used to calculate the modified geometric parameters.

[0024] In one possible design, for the quantized and modulated helical segment carrier, the target termination curvature is pre-calculated based on the target feature value. Geometric constraint optimization is performed using the length and initial curvature increment as optimization variables to obtain the optimized helical segment length and optimized helical segment endpoint coordinates, including:

[0025] Based on the target feature value, the target termination curvature is calculated using the following formula:

[0026] ;

[0027] in, For the target feature value, The length of the current geometric segment;

[0028] With the length of the current geometric segment and initial curvature increment To optimize the variables, a least-squares objective function with geometric constraints is constructed:

[0029] ;

[0030] Where min is the minimization function, J is the least squares objective function for optimization, and u is the x-coordinate of the endpoint of the spiral segment in the local coordinate system obtained after optimization. t v is the original x-coordinate of the endpoint of the spiral segment in the local coordinate system in the original high-precision map, and v is the y-coordinate of the endpoint of the spiral segment in the local coordinate system obtained after optimization calculation. t θ represents the original ordinate of the endpoint of the spiral segment in the local coordinate system from the original high-precision map, and θ is the heading angle of the endpoint of the spiral segment obtained through optimization calculation. t w1 is the original heading angle of the endpoint of the spiral line segment in the original high-precision map, w2 is the weighting coefficient of the endpoint position error, w3 is the weighting coefficient of the regularization constraint of the initial curvature change.

[0031] In one possible design, the new starting mileage of the object to be corrected is calculated by accumulating the updated lengths of its preceding geometric segments, and the mileage values ​​of the associated data elements of the object to be corrected are updated synchronously, including:

[0032] Get by A road composed of an ordered sequence of geometric segments, if the first... Geometric segments If a geometric segment has been modified, then apply this to all subsequent geometric segments. Perform the following mileage recursion sequentially:

[0033] ;

[0034] in, and For the updated starting mileage of the i-th and (i-1)-th geometric segments, This represents the length of the (i-1)th geometric segment after the update.

[0035] Will Assign the mileage value to the associated data element of the i-th geometric segment to complete the synchronous update of the mileage coordinates of subsequent geometric segments and their associated data elements.

[0036] Secondly, this invention provides a high-precision map watermark extraction method, applied to high-precision map data with embedded watermarks obtained by the high-precision map watermark embedding method described in the first aspect and various possible designs of the first aspect, wherein the high-precision map watermark extraction method includes:

[0037] Based on the high-precision map data with embedded watermark, geometric shape carriers and road attribute carriers are extracted, a multi-dimensional carrier pool is constructed, and the feature values ​​of each carrier are calculated.

[0038] A linear confidence weight is constructed based on the surrounding distance from the feature value of the carrier to the nearest embedding point; based on the preset carrier type weight, the decision results scattered throughout the graph are aggregated into the corresponding bits using a global hash mapping to determine the weighted voting statistics.

[0039] The weighted voting statistics are used to make a sign decision to obtain a watermark bit sequence. Random errors are corrected by a decoder to recover the copyright information.

[0040] In one possible design, the linear confidence weights are constructed based on the wraparound distance from the carrier's feature values ​​to the nearest embedding point using the following formula:

[0041] ;

[0042] in, For linear confidence weights, Let quantization step size be 'max', and max be the maximum value function. The characteristic value of the carrier The surrounding distance to the nearest embedding point; the linear confidence weight takes a maximum value of 1 at the embedding point and a minimum value of 0 at the quantization center and decision boundary;

[0043] The weighted voting statistics Calculate using the following formula:

[0044] ;

[0045] in, To be hashed to the first A collection of carriers of bits. For the first The hard decision symbol of a carrier, W type,k For the first The carrier type weight of each carrier, W conf,k For the first The confidence weight of each carrier.

[0046] In one possible design, the method for obtaining the watermark bit sequence by performing sign decision on the weighted voting statistics includes:

[0047] The weighted voting statistics for each bit are used to make a decision using the following formula:

[0048] ;

[0049] in, Let S be the decision result for the j-th watermark bit, and let 1 be the indicator function. j Output 1 if the value is greater than 0, otherwise output 0, thus obtaining the watermark bit sequence.

[0050] Thirdly, the present invention provides an electronic device, comprising: at least one processor and a memory; the memory storing computer execution instructions; the at least one processor executing the computer execution instructions stored in the memory, causing the at least one processor to perform the high-precision map watermark embedding method as described in the first aspect and various possible designs of the first aspect, or the high-precision map watermark extraction method as described in the second aspect and various possible designs of the second aspect.

[0051] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the high-precision map watermark embedding method as described in the first aspect and various possible designs of the first aspect, or the high-precision map watermark extraction method as described in the second aspect and various possible designs of the second aspect.

[0052] The high-precision map watermark embedding and extraction method, device, and medium provided by this invention have at least the following beneficial effects:

[0053] 1. This invention, based on the parametric representation characteristics of high-precision maps, extracts spiral segments and arc segments from road reference lines as geometric shape carriers. Simultaneously, it extracts lane width, pavement marking width, and lane center offset from the attribute layer as road attribute carriers, constructing a multi-dimensional carrier pool containing both geometric shape and road attribute carriers. This multi-dimensional carrier pool can uncover attribute parameter spaces that are easily overlooked by traditional methods, helping to alleviate the problem of relatively scarce geometric carriers in existing technologies for long straight roads or simple road networks in high-precision maps, and achieving a high watermark embedding capacity per unit of data.

[0054] 2. This invention directly employs absolute physical value quantization index modulation for road attribute carriers, while calculating geometric scaling invariants before performing quantization index modulation for geometric shape carriers. Since road attributes and geometric definitions are physically isolated and stored in high-precision map data, absolute quantization index modulation does not rely on local reference values, effectively mitigating self-referencing crosstalk errors caused by embedding operations altering reference values ​​in traditional quantization index modulation. Simultaneously, an orthogonal decoupling relationship is formed between different carrier types, allowing multiple carrier types to be safely embedded in parallel, reducing cross-interference and improving the reliability of watermark extraction.

[0055] 3. This invention targets helical segment carriers, using length and initial curvature increments as optimization variables to construct a least-squares objective function that includes endpoint position error constraints and heading angle error constraints. The Levenberg-Marquardt algorithm is used to solve this function. Under the condition of meeting watermark embedding requirements, the endpoint position tolerance can be controlled to the sub-micrometer level, and the heading angle tolerance to within 0.0001 rad, which helps maintain the continuity of the road reference line. For circular arc segment carriers, the new length is directly calculated based on the target feature values, which is less likely to introduce additional errors. The above geometric constraint optimization mechanism can reduce the impact of the watermark embedding process on the absolute spatial accuracy of the map to a low level, thereby meeting the centimeter-level accuracy requirements of autonomous driving.

[0056] 4. This invention reduces the risk of road topology breakage through a topology chain correction strategy. When a geometric segment experiences changes in length or endpoint coordinates due to watermark embedding, the topology chain correction strategy sequentially updates the mileage of all subsequent geometric segments and synchronously assigns the updated mileage values ​​to the associated auxiliary data elements. Through the synergistic effect of cascaded mileage updates and optimizer endpoint constraints, the geometric deformation of a single geometric segment can be smoothly propagated and compensated along the topology chain, maintaining overall map topology consistency without changing the original endpoint coordinates of subsequent geometric segments. This effectively reduces the possibility of road breakage or misalignment caused by watermark embedding.

[0057] 5. Based on the inherent characteristics of geometric scaling invariants, this invention possesses natural immunity to global translation and rotation attacks, and can maintain relative stability of feature values ​​against global proportional scaling attacks. Since the watermark information is distributed across multiple carriers throughout the road reference line and attribute layer, even if the map suffers a high proportion of cropping attacks, sufficient watermark information may still be retained in the remaining carriers. After weighted soft decision and hash aggregation, the copyright identifier can usually be correctly recovered. For format cleaning and cross-format conversion attacks, this invention embeds the watermark into the geometric shape and physical attributes themselves, rather than relying on text structure features such as XML tag order or invisible characters. Therefore, format cleaning operations are less likely to erase the watermark information, helping to overcome the weakness of existing text structure watermarks in resisting destruction.

[0058] 6. After embedding the watermark, the curvature distribution of the road reference lines, lane width, road marking width, and lane offset remain continuous and smooth, making it less prone to introducing additional high-frequency jitter or abrupt geometric changes. Dynamic simulations have verified that the watermarked map can support autonomous driving systems in completing normal path planning and vehicle control without significantly disrupting road smoothness, thus meeting the basic requirements for safe autonomous driving applications. Attached Figure Description

[0059] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0060] Figure 1 A flowchart of a high-precision map watermark embedding method provided in an embodiment of the present invention;

[0061] Figure 2 A flowchart of a high-precision map watermark extraction method provided in an embodiment of the present invention;

[0062] Figure 3 The image shows a visualization of the original experimental data provided in the embodiments of the present invention; wherein, (a) data a; (b) data b; (c) data c;

[0063] Figure 4 This is a schematic diagram of a simulated driving experiment provided in an embodiment of the present invention;

[0064] Figure 5 The following are comparison diagrams showing the visual effects provided for embodiments of the present invention; wherein, (a) before watermark embedding; (b) after watermark embedding;

[0065] Figure 6 A comparison of text display effects provided for embodiments of the present invention; wherein, (a) contains the watermark of the present invention; (b) contains the watermark of document I;

[0066] Figure 7 The figures show a comparison of experimental results for translation, scaling, and rotation attacks provided in the embodiments of the present invention; wherein, (a) translation attack experimental results; (b) scaling attack experimental results; and (c) rotation attack experimental results.

[0067] The accompanying drawings have illustrated specific embodiments of the invention, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art through reference to specific embodiments. Detailed Implementation

[0068] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention as detailed in the appended claims.

[0069] In the technical solution of this invention, the collection, storage, use, processing, transmission, provision and disclosure of relevant data and information comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0070] It should be noted that in the embodiments of the present invention, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of the present invention. However, they do not mean that the inventor has used or necessarily used the solution.

[0071] The technical solution of the present invention and how the technical solution of the present invention solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of the present invention will now be described with reference to the accompanying drawings.

[0072] To address the limitations of existing watermarking technologies in protecting the copyright of OpenDRIVE high-precision maps, which struggle to balance robustness and geometric accuracy, this invention provides a high-precision map watermark embedding method. This method constructs a multi-dimensional carrier pool containing geometric shapes and road attributes based on parametric representation characteristics. It employs absolute quantization index modulation to achieve orthogonal decoupled embedding of multiple carriers and utilizes Levenberg-Marquardt optimization and topological chain correction strategies to fundamentally guarantee the geometric continuity and topological consistency of the map. Figure 1 As shown, the high-precision map watermark embedding method includes the following steps S101-S105.

[0073] S101: Obtain the original high-precision map data, which includes road reference line geometric segments, lane width attributes, road marking width attributes, and lane center offset attributes.

[0074] It should be noted that the original high-precision map data can specifically adopt the OpenDRIVE format. OpenDRIVE is a high-precision map exchange standard widely used in autonomous driving simulation. It uses XML for structured storage, and its road model is centered on reference lines, describing the road shape through parametric geometric segments such as straight lines, spirals, and arcs. Semantic information such as lane width, pavement marking width, and lane center offset is recorded through attribute nodes. The acquisition of original high-precision map data can include vehicle-mounted LiDAR mapping, satellite image extraction, or importing from third-party data sources. This embodiment uses the OpenDRIVE format as an example for illustration; however, those skilled in the art should understand that OpenDRIVE is merely an exemplary format and does not constitute a limitation on the scope of protection of this invention.

[0075] S102: Based on the hierarchical structure of the original high-precision map data, spiral segments and arc segments are extracted from the geometric segments of the road reference lines as geometric shape carriers, and lane width attributes, road marking width attributes, and lane center offset attributes are extracted as road attribute carriers to construct a multi-dimensional carrier pool that includes geometric shape carriers and road attribute carriers.

[0076] Taking OpenDRIVE as an example, this embodiment constructs a carrier pool containing three orthogonal dimensions based on the hierarchical structure of OpenDRIVE, achieving a significant expansion of the embedding capacity. For example, by traversing the OpenDRIVE DOM tree, spiral lines, arc geometric segments, lane width, lane marking width, and lane offset attributes can be extracted to construct a multi-dimensional carrier pool.

[0077] For geometrically shaped carriers, OpenDRIVE reference lines are primarily composed of straight lines, helices, and arcs. Among these, the curvature distribution of helices and arcs naturally possesses resistance to translation and rotation. To resist global scaling attacks, this embodiment extracts its geometric scaling invariants. This geometric scaling invariant It is positively correlated with the total turning angle of the curve:

[0078] ;

[0079] in, Let be the curvature of the arc segment. Let be the terminal curvature of the helical segment. The length of the current geometric segment is represented by Spiral, which represents a spiral segment, and Arc, which represents a circular arc segment.

[0080] When map data is affected by a scaling factor During a global scaling attack, its curvature transformation is Length transformation The product of the two It remains strictly constant. This intrinsic property makes it an ideal carrier of scaling-resistant geometry.

[0081] For road attribute carriers, the physical geometry definition of a road is as follows in the OpenDRIVE standard specification. <geometry>The semantic attributes of roads are physically isolated and decoupled from road semantic infrastructure. These semantic attributes are typically expressed using cubic polynomials. Perform a parameterized description, where the constant term This represents the absolute physical measure of the attribute at the starting point. This represents the coefficient of the linear term, characterizing the linear rate of change of the road attribute with mileage. The coefficient represents the quadratic term, indicating the rate of change of the road's attributes. The coefficients of the cubic term are used as higher-order adjustment factors for fine fitting of nonlinear transition edges. This represents the local axial mileage relative to the starting point of the current road segment. Represents the current local mileage The corresponding road attribute calculation value.

[0082] Based on the above characteristics, this embodiment selects the following three types of core semantic attributes as extension carriers:

[0083] ① Lane width is determined by <lane>under the node <width>The attribute definition includes a constant term representing the initial absolute width of the lane.

[0084] ②The width of the road markings is from <roadmark>The attribute definition includes a constant term that represents the actual physical width of the traffic markings.

[0085] ③ Lane center offset from <lanes>under the node <laneoffset>The attribute definition includes a constant term that represents the absolute lateral offset of the reference line relative to the actual geometric center of the road.

[0086] In conventional map coordinate translation and rotation, the absolute scalar values ​​represented by the above three attributes are not affected by global coordinate system transformation. To avoid numerical crosstalk between geometric and attribute carriers during joint embedding, this embodiment subsequently uses an absolute quantization index modulation strategy, fundamentally achieving orthogonality and decoupling between different types of carriers.

[0087] S103: The original copyright information is converted into a watermark bitstream through error correction encoding. An absolute quantization index modulation strategy is used to embed the watermark into each carrier in the multidimensional carrier pool. Specifically, for road attribute carriers, the absolute physical values ​​of the road attribute carriers are quantized and modulated to obtain modified attribute values, and the modified attribute values ​​are written back to the corresponding attribute nodes to complete the watermark embedding of the road attribute carriers. For geometric shape carriers, the geometric scaling invariant of the geometric shape carriers is calculated, and the geometric scaling invariant is quantized and modulated to obtain the target feature value.

[0088] Specifically, in this embodiment, the original copyright information is encoded using BCH(15,7,2) error correction to generate a watermark bitstream. The frame format includes a 4-bit flag field and an 8-bit length field. The length field is used for frame self-checking at the extraction end. (Bitstream establishment) The index relationship between the map carriers is established. A unique identifier for each carrier is calculated using the FNV-1a hash algorithm, and the bitstream is then configured based on the calculation results. Specific bits in They are pseudo-randomly assigned to the corresponding carriers.

[0089] In the absolute quantization index modulation strategy, different embedding methods are used for two different types of carriers. Specifically, the quantization index modulation strategy (QIM) used in this embodiment differs from the traditional QIM watermarking method. Traditional QIM watermarking methods often use local mean or reference values ​​as normalization benchmarks, but the embedding operation itself changes the reference value, leading to self-referencing errors that cause extraction errors. To completely eliminate this self-referencing crosstalk problem, this embodiment adopts an absolute QIM strategy for the attribute carrier, directly quantizing and modulating the absolute physical value of the attribute without relying on any external reference value.

[0090] Lane width, lane marking width, and lane offset are all independent scalar attributes, with values ​​defined independently by OpenDRIVE's attribute layer and physically isolated from the geometric definition of the reference lines. Therefore, embedding operations on attribute carriers do not affect the invariant calculations of geometric carriers, and vice versa. This orthogonal decoupling characteristic allows multiple types of carriers to be safely embedded in parallel without any cross-interference.

[0091] For coordinate system-level attacks, attribute carriers are naturally immune to translation and rotation transformations because these scalar attributes do not depend on the global coordinate system. Global scaling... When the value is multiplied by a factor of 1, the absolute value of the attribute carrier will be scaled accordingly. Its scaling robustness is determined by the scaling invariant of the geometric carrier. Provides collaborative protection. In practical applications, map format conversion and coordinate system transformation mainly involve translation and rotation, and the probability of global proportional scaling attacks is extremely low.

[0092] When using quantization indexed modulation (QIM) for bit embedding, let the quantization step size be... The binary watermark bit to be embedded is For the current carrier, let its original feature value be V, then the target feature value is... The calculation formula is:

[0093] ;

[0094] Where Round is the rounding function, V is the original feature value of the current carrier; for attribute carriers, its original feature value is the absolute value of the attribute, and for geometric carriers, its original feature value is the scaling invariant. .

[0095] For the three attributes of lane width, lane marking width, and lane offset, since their modification does not involve reference line integration, they naturally satisfy geometric continuity. The modified attribute values ​​can be directly written back to the data without geometric optimization. However, for geometrically shaped carriers, further geometric constraint optimization steps are required to complete the watermark embedding.

[0096] S104: For the helical segment carrier after quantization modulation, the target termination curvature is pre-calculated based on the target feature value. Geometric constraint optimization is performed with the length and initial curvature increment as optimization variables to obtain the optimized helical segment length and the optimized helical segment endpoint coordinates. For the circular arc segment carrier, the new length of the circular arc segment is calculated based on the target feature value.

[0097] In this embodiment, QIM modulation is applied to the mapped carrier features. For a helical carrier, the LM optimizer is activated to solve for the optimal parameters that satisfy the geometric constraints; for a circular arc carrier, the new length is calculated directly. .

[0098] In this embodiment, multi-order geometric continuity is introduced as a boundary constraint criterion when adjusting the carrier parameters. Geometric continuity is defined as spatial continuity, which requires that the three-dimensional spatial coordinates of adjacent geometric segments completely coincide at the joints between segments in order to eliminate data gaps between segments. Geometric continuity is defined as continuity in the direction of the first derivative, requiring that the tangent vectors of the curves at the point of connection between segments remain consistent to ensure a smooth transition in heading angle when autonomous vehicles cross segments. Specifically, for the geometric vehicle, the scaling invariant is modified. The curvature or length of the curve needs to be changed. For helical carriers, this embodiment first pre-calculates the target termination curvature according to the QIM formula. As a hard constraint, the Levenberg-Marquardt (LM) nonlinear optimization algorithm is then introduced to optimize the length. and initial curvature increment To optimize the variables, a least-squares objective function with geometric constraints is constructed:

[0099] ;

[0100] Where min is the minimization function, J is the least squares objective function for optimization, and u is the x-coordinate of the endpoint of the spiral segment in the local coordinate system obtained after optimization. t v is the original x-coordinate of the endpoint of the spiral segment in the local coordinate system in the original high-precision map, and v is the y-coordinate of the endpoint of the spiral segment in the local coordinate system obtained after optimization calculation. t θ represents the original ordinate of the endpoint of the spiral segment in the local coordinate system from the original high-precision map, and θ is the heading angle of the endpoint of the spiral segment obtained through optimization calculation. t The original heading angle of the endpoint of the spiral segment in the original high-precision map is w1, w2 is the weighting coefficient of the endpoint position error, w3 is the weighting coefficient of the regularization constraint of the initial curvature change. (First term) Constrain endpoint position errors to ensure Continuity; second term Constrain endpoint heading angle error to ensure Continuity; Third term To regularize the initial curvature variation and prevent the optimizer from over-adjusting the initial curvature, the weights are set to... , , The optimized model described above satisfies the watermark embedding requirements while ensuring positional tolerance. Heading tolerance .

[0101] For a circular arc carrier, due to its curvature The length is constant and the new length can be calculated directly. No need to enable LM optimization.

[0102] S105: Geometric segments whose lengths have changed relative to the original lengths of the corresponding geometric segments after optimization (either the length of the spiral segment or the length of the arc segment after optimization) or whose endpoint coordinates have changed relative to the original endpoint coordinates of the corresponding geometric segment are considered as modified geometric segments. The subsequent geometric segments of the modified geometric segments are considered as objects to be corrected. The updated lengths of the preceding geometric segments of the objects to be corrected are accumulated to calculate the new starting mileage of the objects to be corrected. The mileage values ​​of the associated data elements of the objects to be corrected are updated synchronously. This process transmits and compensates for the geometric deformation caused by watermark embedding on the topological chain, resulting in high-precision map data with embedded watermark.

[0103] In this embodiment, a topology-preserving chain-like correction strategy is used to embed the watermark into minor adjustments to the length or endpoint position of individual geometric segments, propagating it forward along the ordered sequence of geometric segments on the road reference line. Since the starting mileage of subsequent geometric segments depends entirely on the cumulative length of preceding geometric segments, synchronously accumulating and updating the mileage value of each subsequent geometric segment ensures the continuity of the entire road's linear reference system without altering its original endpoint coordinates. Simultaneously, the sub-millimeter-level constraints on endpoint positions and heading angles imposed by the Levenberg-Marquardt optimizer in step S104 strictly maintain the adjacency relationship between modified geometric segments and subsequent geometric segments. This avoids topological breaks or mileage misalignments caused by local geometric deformations, ensuring the integrity of the road structure of the watermarked map and the availability of subsequent autonomous driving applications.

[0104] Specifically, in the OpenDRIVE high-precision map standard, road reference lines consist of a chain-like topology formed by a series of interconnected geometric segments. When LM optimization fine-tunes the length of a certain geometric segment... In OpenDRIVE, because the starting position of subsequent geometric segments is recorded using an absolute coordinate system, changes in the length of a single node can disrupt its geometric adjacency with subsequent nodes. To address this issue, this paper proposes a topology-preserving chain-based correction strategy, which propagates and compensates for geometric deformations caused by watermark embedding along the topology chain.

[0105] Suppose a certain road is... A sequence of ordered geometric segments Composition. If the first Geometric segments If the optimization results in a change in length or a change in the endpoint position (i.e., the geometric segment has been modified), then all subsequent geometric segments... Perform the following corrections:

[0106] (1) Mileage advance update

[0107] Starting mileage of all subsequent road sections The summation needs to be based on the actual length of the preceding road segments to maintain the continuity of the linear reference system.

[0108] ;

[0109] in, and For the updated starting mileage of the i-th and (i-1)-th geometric segments, This is the length of the (i-1)th geometric segment after the update.

[0110] The above updates also cover subsequent geometric segments and their associated ancillary data elements. coordinate.

[0111] (2) Geometric continuity guarantee

[0112] For the endpoints Spatial continuity is ensured in this embodiment by using endpoint constraints of the LM optimizer, which controls the deviation between the end point of the modified geometric segment and the start point of the subsequent segment to the sub-millimeter level, thereby maintaining the geometric continuity of the entire graph while keeping the original coordinates unchanged.

[0113] Example 2:

[0114] This invention provides a high-precision map watermark extraction method, applied to high-precision map data with embedded watermarks obtained by the high-precision map watermark embedding method in Embodiment 1, such as... Figure 2 As shown, the high-precision map watermark extraction method includes the following steps S201-S203.

[0115] S201: Based on the high-precision map data with embedded watermarks, extract the geometric shape carrier and road attribute carrier, construct a multi-dimensional carrier pool, and calculate the feature values ​​of each carrier.

[0116] In this embodiment, high-precision map data with embedded watermark is used as the map to be tested. A carrier pool is constructed and feature values ​​are calculated according to the same rules as the embedding end. .

[0117] S202: Construct linear confidence weights based on the wraparound distance from the feature value of the carrier to the nearest embedding point; based on the preset carrier type weights, use global hash mapping to aggregate the decision results scattered throughout the graph to the corresponding bit positions to determine the weighted voting statistics.

[0118] To improve extraction accuracy under low signal-to-noise ratio conditions, this embodiment employs a soft-decision mechanism. Definition The surrounding distance to the nearest embedding point is Constructing linear confidence weights This makes it take the maximum value of 1 at the embedding point and the minimum value of 0 at the quantization center and decision boundary:

[0119] ;

[0120] in, For linear confidence weights, Let quantization step size be 'max', and max be the maximum value function. The characteristic value of the carrier The surrounding distance to the nearest embedding point.

[0121] At the same time, carrier type weights are introduced. Different weights are assigned based on the noise resistance of different carrier types, such as geometric carrier 1.2, lane width 1.0, road markings 0.8, and lane offset 0.6. Using a global hash mapping relationship, the decision results scattered throughout the entire graph are aggregated to their corresponding bits. Above. Definition For the first The hard decision symbol of each carrier, then the weighted vote count for:

[0122] ;

[0123] in, To be hashed to the first The set of carriers of bits, W type,k For the first The carrier type weight of each carrier, W conf,k For the first The confidence weight of each carrier.

[0124] S203: Perform symbol decision on the weighted voting statistics to obtain the watermark bit sequence, correct random errors through the decoder, and recover the copyright information.

[0125] Weighted voting statistics The following final sign decision is made:

[0126] ;

[0127] in, Let S be the decision result for the j-th watermark bit, and let 1 be the indicator function. j Output 1 if the value is greater than 0, otherwise output 0, thus obtaining the watermark bit sequence.

[0128] After the final symbol decision is executed, the BCH decoder corrects any random errors that may occur during transmission and restores the copyright information.

[0129] Example 3:

[0130] To comprehensively verify the effectiveness of the watermarking method (including watermark embedding and extraction methods) proposed in this invention for copyright protection and integrity authentication of OpenDRIVE format high-precision maps, this embodiment designed multiple sets of comparative experiments. The experiments focused on examining the applicability, imperceptibility, and robustness of the watermarking method, and compared it with existing high-precision map watermarking methods. This embodiment selected three data sets for the experiment: data a, data b, and data c, with sizes of 2066 kilobytes, 2252 kilobytes, and 26722 kilobytes, respectively. Figure 3 As shown.

[0131] The experimental environment was a 64-bit Windows 11 operating system; the programming language was C++20; the development environment was Visual Studio 2022; the OpenDRIVE viewer was the OdrViewer online viewer; the simulation used the Esmini 2.55.0 open-source OpenDRIVE simulator; and the text comparison software was Beyond Compare 4.4.3.

[0132] To verify the usability of the watermarking method proposed in this invention in practical autonomous driving applications, this embodiment designed a simulation test based on Esmini. The aim is to ensure that although the method makes minor adjustments to the map's geometric parameters, it does not disrupt the smoothness of the road or affect the vehicle's driving safety.

[0133] This experiment uses the open-source emulator Esmini as the testing platform. An automated batch processing test method was built, and the specific process is as follows: three steps.

[0134] Step 1: Dynamic scene automatically generated.

[0135] For each road in the OpenDRIVE map under test, a corresponding Open SCENARIO (.xosc) standard scene file is automatically generated. The vehicle model adopts a standard four-wheel dynamics model, and the vehicle is set to travel along the lane centerline of the current road reference line. The initial speed is set to... To simulate typical driving conditions on urban expressways and suburban roads.

[0136] Step 2, simulation execution.

[0137] The generated scene was batch-processed and simulated using Esmini's background no-rendering mode. The simulation step size was set to... This ensures that vehicle movement can be captured at the millisecond level.

[0138] Step 3: Data recording and parsing.

[0139] During the simulation, the vehicle's trajectory coordinates are recorded in real time. Heading angle lateral acceleration and road curvature The data is collected and output as a CSV log for subsequent analysis.

[0140] To quantify the impact of watermarking on map usability, this experiment selected maximum lateral jerk and maximum curvature jump as core evaluation metrics. These two metrics correspond to ride safety and geometric smoothness in autonomous driving, respectively.

[0141] Jerk is the derivative of acceleration with respect to time, reflecting how quickly a vehicle's acceleration changes. In autonomous driving, lateral jerk... It is a key physical quantity for measuring the degree of steering wheel vibration caused by abrupt changes in road geometry. Its calculation formula is:

[0142] ;

[0143] in Let t be the lateral acceleration of the vehicle, and t be the time. The time interval. If there are road geometric connections... Discontinuous Pulsating mutations will occur. Based on the security standards for high-precision maps, the security threshold set for this experiment is... for:

[0144] ;

[0145] like If so, it is determined that the watermark embedding in this section of the road disrupts the dynamic stability.

[0146] Curvature jumps reflect road geometry Continuity. In an ideal OpenDRIVE map, the curvature should remain continuous at the junctions of spirals and arcs or straight lines. Curvature jumps. The calculation formula is:

[0147] ;

[0148] in, To test the road reference line path, The curvature at the starting point of the subsequent geometric segment, Let be the curvature of the endpoint of the preceding geometric segment.

[0149] Considering the accuracy requirements of high-precision maps, this experiment sets a strict geometric continuity threshold. for:

[0150] ;

[0151] like If so, it is determined that the watermarking method has disrupted the geometric smoothness of the map.

[0152] In addition to the quantitative indicators mentioned above, the experiment also needs to calculate the proportion of simulated vehicles that successfully complete the predetermined trajectory, i.e., the trajectory pass rate. This trajectory pass rate is also used as an evaluation indicator. If a vehicle deviates from its lane, collides, or the simulator crashes due to geometric errors during the simulation, the test is considered a failure.

[0153] Following the above applicability analysis method, and after testing, the data with embedded watermarks can meet the requirements for high-precision maps, and the rendering effect of the simulated driving experiment is as follows: Figure 4 As shown.

[0154] To verify the impact of embedded watermarks on the visual presentation of maps, the experiment used the OpenDRIVE online viewer odrviewer to examine the differences in visualization effects before and after watermark embedding.

[0155] like Figure 5 As shown in the visualization comparison, the high-precision map data with embedded watermarks maintains a high degree of visual consistency in overall spatial form, and no significant geometric distortion is presented in the regular view. The watermarking method still demonstrates excellent concealment and imperceptibility in the overall data visualization effect.

[0156] To evaluate the robustness and concealment of the watermarking method in non-Unicode encoded environments, this experiment visually compared data embedded with the watermark of this invention and the data described in Reference I in VSCode, as shown below. Figure 6 As shown. Observation reveals that the zero-width character scheme used in Reference I leads to parsing errors and produces obvious garbled characters in restricted environments due to encoding compatibility issues; conversely, the watermarking method proposed in this invention does not rely on special character encoding and can perfectly maintain the text presentation of the original data. Experimental results confirm that, under restricted display conditions lacking Unicode support, the watermark obtained by the method of this invention has good visual imperceptibility.

[0157] It should be noted that Reference I refers to "Lv Xuchao, Ren Na, Zhou Qifei, Zhu Changqing. Digital watermarking algorithm for high-precision maps in OpenDrive format with invisible characters [J]. Journal of Geoinformation Science, 2024, 26(9):2026-2037."

[0158] Table 1 shows the impact of different watermarking methods on file size changes. The data indicates that the method in Reference I resulted in a data increment of more than 18%, significantly increasing the file size. In contrast, the method proposed in this invention successfully embeds the watermark while maintaining a size essentially the same as the original data, effectively avoiding the risk of exposing the watermark due to sudden changes in file size. Thus, it demonstrates significant advantages in both concealment and storage efficiency.

[0159] Table 1: File Size Changes

[0160]

[0161] Robustness is a core indicator for evaluating whether a digital watermarking method can still maintain information integrity and be accurately extracted after being attacked. To objectively assess the anti-attack capability of the method in this invention, this experiment selected the normalized correlation coefficient (NC) as the evaluation criterion.

[0162] Let the original watermark information sequence be... The watermark information sequence extracted after the attack is The sequence length is This experiment uses the bit-matching rate based on the XNOR operation as the formula for calculating the NC value:

[0163] ;

[0164] in This represents the XOR operation, where the i-th original watermark information sequence... The sequence of watermark information extracted after the i-th attack The result is 1 if the two are the same, and 0 otherwise. The above formula reflects the similarity between the extracted watermark and the original embedded information. When the extracted information is completely consistent with the original embedded information, it is generally considered that... It has effective watermark extraction capabilities. In this experiment, 0.7 was set as the judgment threshold.

[0165] To verify the advancement of the method of this invention, this experiment selected the robust watermarking method proposed in Reference I as a benchmark for comparison. The experiment conducted comprehensive comparative tests against geometric transformation attacks, random cropping attacks, and format cleaning attacks.

[0166] 1) Robustness to geometric transformation attacks

[0167] Geometric transformation attacks are the most common operations in high-precision map processing, including coordinate system transformation, projection transformation, and editing. The experiment performed translation, rotation, and scaling attacks on watermarked data, and the test results are as follows. Figure 7 As shown, the results indicate that the method of the present invention maintains a perfect extraction rate of NC=1.0 under all conventional geometric transformation attacks.

[0168] 2) Robustness comparison against clipping attacks

[0169] Data truncation is one of the most destructive attack methods in high-precision map distribution. It directly deletes large amounts of data such as lanes and intersections, destroying the integrity of watermark information. To verify the resilience of the method in this invention under extreme conditions, this experiment conducted a comparative test against data truncation attacks.

[0170] The experiment randomly cut the map into regions according to different cropping ratios, and tested the method of the present invention using data c, comparing it with the methods in references II, III and I.

[0171] It should be noted that Reference II refers to "Li Zhaocan, Wang Liming, Ge Sijiang, et al. A method for watermarking plain text in big data based on orthogonal coding [J]. Computer Science, 2019, 46(12): 148-154.", and Reference III refers to "Zhang Zhenyu, Li Qianmu, Qi Yong. Design of text watermark based on invisible characters [J]. Journal of Nanjing University of Science and Technology, 2017, 41(4): 405-411.".

[0172] Table 2 Results of the Pruning Attack Experiment

[0173]

[0174] As shown in Table 2, the NC values ​​of methods II and III decreased significantly when the cropping ratio reached 90%, and those of method I also decreased after exceeding 90%. In contrast, the method of the present invention still has an NC value of 1 even when cropping 98%, and can extract perfectly, indicating that the method of the present invention has extremely strong anti-cropping ability.

[0175] 3) Comparison of robustness to anti-format cleaning

[0176] In the actual production and distribution of OpenDRIVE high-definition maps, data frequently needs to flow between different editing software, simulation platforms, and databases. This process inevitably involves operations such as format standardization and cleaning, as well as data format conversion. Although these operations do not change the geometric topology and logical semantics of the map, they reset the underlying text encoding and storage structure of the file, posing a significant challenge to the robustness of digital watermarks.

[0177] To verify the superiority of the method of this invention in dealing with such high-order semantic attacks, this experiment designed a comparative experiment between the method of this invention and the text watermarking method based on invisible characters in Reference I.

[0178] Table 3 Results of the format cleaning attack experiment

[0179]

[0180] As shown in Table 3, after standardized format cleaning and data format conversion, the invisible character watermark embedded in Document I was filtered or truncated because it could not pass the strict verification of the underlying data type, resulting in complete information loss. In contrast, the method of this invention is not affected by the above processing flow and can successfully extract the complete digital watermark.

[0181] The experimental results above demonstrate that the method of this invention, while ensuring the visual and storage insensitivity of the map, exhibits extremely high robustness against strong attacks such as geometric transformations, 98% extreme cropping, and format cleaning. Esmini dynamics simulations verify that the method of this invention does not disrupt road smoothness and fully meets the safety application requirements of autonomous driving systems.

[0182] In summary, the method of this invention constructs a multidimensional orthogonal carrier pool by exploiting the parametric characteristics of OpenDRIVE and employs an absolute QIM strategy to eliminate self-referencing crosstalk. By combining LM nonlinear optimization and topology-preserving chain correction, it effectively avoids the risk of topological breakage caused by coordinate perturbations while achieving high-capacity information embedding. Multidimensional simulations and comparative experiments confirm that the method of this invention can perfectly resist destructive attacks such as translation scaling, large-scale region cropping, and cross-format cleaning while ensuring centimeter-level absolute spatial accuracy and dynamic usability, thus solving the technical problem of poor resistance to damage in existing text structure watermarks.

[0183] Example 4:

[0184] This invention provides an electronic device. The electronic device may include a processor and a memory, wherein the processor and the memory can communicate; exemplarily, the processor and the memory communicate via a communication bus.

[0185] The processor executes computer execution instructions stored in memory, causing the processor to perform the scheme in the above embodiments. The processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0186] The communication bus can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The system bus can be divided into address bus, data bus, control bus, etc. Transceivers are used to enable communication between database access devices and other computers (e.g., clients, read-write libraries, and read-only libraries). Memory may include random access memory (RAM) and may also include non-volatile memory.

[0187] The electronic device provided in this embodiment of the invention can be the terminal device described in the above embodiments.

[0188] This invention also provides a computer-readable storage medium storing computer instructions. When the computer instructions are executed on a computer, the computer performs the technical solutions of the high-precision map watermark embedding method or the high-precision map watermark extraction method described in the above embodiments.

[0189] This invention also provides a computer program product, which includes a computer program stored in a computer-readable storage medium. At least one processor can read the computer program from the computer-readable storage medium. When the at least one processor executes the computer program, it can implement the technical solutions of the high-precision map watermark embedding method or the high-precision map watermark extraction method described in the above embodiments.

[0190] In the several embodiments provided by this invention, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms.

[0191] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to implement the solution of this embodiment according to actual needs.

[0192] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each module can exist physically separately, or two or more modules can be integrated into one unit. The unit composed of the above modules can be implemented in hardware or in the form of hardware plus software functional units.

[0193] The integrated modules described above, implemented as software functional modules, can be stored in a computer-readable storage medium. These software functional modules, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute certain steps of the methods of the various embodiments of the present invention.

[0194] It should be understood that the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly manifested as execution by a hardware processor, or execution by a combination of hardware and software modules within the processor.

[0195] The memory may include high-speed RAM, and may also include non-volatile storage (NVM), such as at least one disk storage device, and may also be a USB flash drive, external hard drive, read-only memory, disk or optical disc, etc.

[0196] Buses can be Industry Standard Architecture (ISA) buses, Peripheral Component Interconnect (PCI) buses, or Extended Industry Standard Architecture (EISA) buses, etc. Buses can be categorized into address buses, data buses, control buses, etc.

[0197] The aforementioned storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0198] An exemplary storage medium is coupled to a processor, enabling the processor to read information from and write information to the storage medium. Alternatively, the storage medium can be an integral part of the processor. The processor and storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and storage medium can exist as discrete components in an electronic control unit or main control device.

[0199] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0200] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.< / laneoffset> < / lanes> < / roadmark> < / width> < / lane> < / geometry>

Claims

1. A method for embedding watermarks in high-precision maps, characterized in that, The method includes: Acquire raw high-precision map data, which includes road reference line geometric segments, lane width attributes, road marking width attributes, and lane center offset attributes; Based on the hierarchical structure of the original high-precision map data, spiral segments and arc segments are extracted from the geometric segments of the road reference lines as geometric shape carriers, and lane width attributes, road marking width attributes, and lane center offset attributes are extracted as road attribute carriers to construct a multi-dimensional carrier pool including geometric shape carriers and road attribute carriers. The original copyright information is converted into a watermark bitstream through error correction encoding. An absolute quantization index modulation strategy is used to embed the watermark into each carrier in the multidimensional carrier pool. Specifically, for road attribute carriers, the absolute physical values ​​of the road attribute carriers are quantized and modulated to obtain modified attribute values, and the modified attribute values ​​are written back to the corresponding attribute nodes to complete the watermark embedding of the road attribute carriers. For geometric shape carriers, the geometric scaling invariant of the geometric shape carriers is calculated, and the geometric scaling invariant is quantized and modulated to obtain the target feature value. For the quantized and modulated spiral segment carrier, the target termination curvature is pre-calculated based on the target feature value, and geometric constraint optimization is performed using the length and initial curvature increment as optimization variables to obtain the optimized spiral segment length and the optimized spiral segment endpoint coordinates; for the arc segment carrier, the new length of the arc segment is calculated based on the target feature value. The geometric segments whose lengths have changed relative to the original lengths of the corresponding geometric segments, or whose endpoint coordinates have changed relative to the original endpoint coordinates of the corresponding geometric segments, are designated as modified geometric segments. The subsequent geometric segments of these modified segments are designated as objects to be corrected. The updated lengths of the preceding geometric segments of the objects to be corrected are accumulated to calculate the new starting mileage of the objects to be corrected. Simultaneously, the mileage values ​​of the associated auxiliary data elements of the objects to be corrected are updated. This process transmits and compensates for the geometric deformation caused by watermark embedding along the topological chain, resulting in high-precision map data with embedded watermarks.

2. The high-precision map watermark embedding method according to claim 1, characterized in that, The geometric scaling invariant Calculate using the following formula: ； in, Let be the curvature of the arc segment. Let be the terminal curvature of the helical segment. The length of the current geometric segment is denoted as Spiral, where Spiral is a spiral segment and Arc is a circular arc segment; the geometric scaling invariant remains constant when the map is subjected to a global scaling attack.

3. The high-precision map watermark embedding method according to claim 1, characterized in that, The methods for embedding watermarks into each carrier in the multidimensional carrier pool using an absolute quantization index modulation strategy include: The target feature value is calculated using the following formula: ； in, To quantize the step size, V represents the binary watermark bits to be embedded, and V is the original feature value of the current carrier. The target feature value is defined by Round, which is the rounding function. For the road attribute carrier, the modified attribute value is Write back directly to the data; For a geometrically shaped carrier, the target feature value Used to calculate the modified geometric parameters.

4. The high-precision map watermark embedding method according to claim 1, characterized in that, For the quantized and modulated helical segment carrier, the target termination curvature is pre-calculated based on the target feature value. Geometric constraint optimization is performed using the length and initial curvature increment as optimization variables to obtain the optimized helical segment length and optimized helical segment endpoint coordinates, including: Based on the target feature value, the target termination curvature is calculated using the following formula: ； in, For the target feature value, The length of the current geometric segment; With the length of the current geometric segment and initial curvature increment To optimize the variables, a least-squares objective function with geometric constraints is constructed: ； Where min is the minimization function, J is the least squares objective function, and u is the x-coordinate of the endpoint of the spiral segment in the local coordinate system obtained after optimization. t v is the original x-coordinate of the endpoint of the spiral segment in the local coordinate system in the original high-precision map, and v is the y-coordinate of the endpoint of the spiral segment in the local coordinate system obtained after optimization calculation. t θ represents the original ordinate of the endpoint of the spiral segment in the local coordinate system from the original high-precision map, and θ is the heading angle of the endpoint of the spiral segment obtained through optimization calculation. t w1 is the original heading angle of the endpoint of the spiral line segment in the original high-precision map, w2 is the weighting coefficient of the endpoint position error, w3 is the weighting coefficient of the regularization constraint of the initial curvature change.

5. The high-precision map watermark embedding method according to claim 1, characterized in that, The new starting mileage of the object to be corrected is calculated by accumulating the updated lengths of its preceding geometric segments, and the mileage values ​​of its associated data elements are updated synchronously, including: Get by A road composed of an ordered sequence of geometric segments, if the first... Geometric segments If a geometric segment has been modified, then apply this to all subsequent geometric segments. Perform the following mileage recursion sequentially: ； in, and For the updated starting mileage of the i-th and (i-1)-th geometric segments, This represents the length of the (i-1)th geometric segment after the update. Will Assign the mileage value to the associated data element of the i-th geometric segment to complete the synchronous update of the mileage coordinates of subsequent geometric segments and their associated data elements.

6. A method for extracting watermarks from high-precision maps, applied to high-precision map data with embedded watermarks obtained by the high-precision map watermark embedding method as described in any one of claims 1 to 5, characterized in that, The high-precision map watermark extraction method includes: Based on the high-precision map data with embedded watermark, geometric shape carriers and road attribute carriers are extracted, a multi-dimensional carrier pool is constructed, and the feature values ​​of each carrier are calculated. A linear confidence weight is constructed based on the surrounding distance from the feature value of the carrier to the nearest embedding point; based on the preset carrier type weight, the decision results scattered throughout the graph are aggregated into the corresponding bits using a global hash mapping to determine the weighted voting statistics. The weighted voting statistics are used to make a sign decision to obtain a watermark bit sequence. Random errors are corrected by a decoder to recover the copyright information.

7. The high-precision map watermark extraction method according to claim 6, characterized in that, Based on the wraparound distance from the carrier's feature values ​​to the nearest embedding point, a linear confidence weight is constructed using the following formula: ； in, For linear confidence weights, Let quantization step size be 'max', and max be the maximum value function. The characteristic value of the carrier The surrounding distance to the nearest embedding point; the linear confidence weight takes a maximum value of 1 at the embedding point and a minimum value of 0 at the quantization center and decision boundary; The weighted voting statistics Calculate using the following formula: ； in, To be hashed to the first A collection of carriers of bits. For the first The hard decision symbol of a carrier, W type,k For the first The carrier type weight of each carrier, W conf,k For the first The confidence weight of each carrier.

8. The high-precision map watermark extraction method according to claim 7, characterized in that, The methods for obtaining the watermark bit sequence by performing sign decision on the weighted voting statistics include: The weighted voting statistics for each bit are used to make a decision using the following formula: ； in, Let S be the decision result for the j-th watermark bit, and let 1 be the indicator function. j Output 1 if the value is greater than 0, otherwise output 0, thus obtaining the watermark bit sequence.

9. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory to implement the high-precision map watermark embedding method as described in any one of claims 1-5 or the high-precision map watermark extraction method as described in any one of claims 6-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the high-precision map watermark embedding method as described in any one of claims 1-5 or the high-precision map watermark extraction method as described in any one of claims 6-8.

Images