CPU-gpu-based group target rcs efficient parallel computing method

Abstract

The application discloses a CPU-GPU-based group target RCS high-efficiency parallel calculation method, which converts the linear reconstruction of group target characteristic current from a cycle-level calculation mode to a matrix-level calculation mode of a GPU platform, adopts a multi-layer fast multipole method, and accelerates the calculation of electromagnetic coupling between targets under a CPU-GPU heterogeneous parallel framework, wherein the calculation task division of aggregation and configuration matrix filling is based on the target and facet dimension, the calculation task division of transfer matrix filling is based on the grouping box pair and beam dimension, and the calculation and superposition of group target far-field scattering are based on the parallel processing mode of the target, facet and angle dimension. Through the cooperative calculation of CPU and GPU, the application can significantly improve the RCS calculation efficiency and scalability of large-scale group targets.

Description

Technical Field

[0001] This invention belongs to the field of radar stealth technology, specifically relating to a CPU-GPU-based method for efficient parallel computation of the RCS of swarm targets. Background Technology

[0002] With the rapid development of electromagnetic engineering, groups of targets composed of a large number of identical array elements are widely used in practical applications. Compared with single targets, the overall electromagnetic scattering characteristics of groups of targets are affected by multiple scattering and strong electromagnetic coupling between targets, making it difficult to describe by simply superimposing the scattering characteristics of individual targets. This complex coupling effect greatly increases the difficulty of radar detection and target characteristic acquisition, placing higher demands on the accuracy and efficiency of electromagnetic scattering modeling and numerical calculation methods for groups of targets. Therefore, achieving efficient solutions to large-scale group target electromagnetic scattering problems while ensuring computational accuracy is not only of significant theoretical research importance, but also has broad application value in engineering applications such as complex electromagnetic environment assessment.

[0003] In the analysis of target electromagnetic scattering characteristics and the calculation of radar cross section (RCS), existing techniques typically employ numerical calculation methods based on full-wave electromagnetic theory. This involves discretizing the target and solving for the distribution of electromagnetic fields or induced currents to obtain scattering characteristics. To reduce computational complexity, various fast algorithms have been proposed to approximate or hierarchically process the electromagnetic interactions between targets, improving computational efficiency to some extent. However, when the target consists of a large number of repetitive units, or when RCS calculations are required for groups of targets under multiple attitudes and frequencies, large-scale unknowns still need to be processed, along with numerous matrix operations and iterative solutions. The computational process is highly dependent on traditional CPU platforms, with limited overall parallelism, insufficient computational efficiency and scalability, making it difficult to meet the needs of rapid analysis and engineering applications. With further increases in computational scale, even with fast algorithms and structure reuse strategies, the electromagnetic coupling between targets still needs repeated calculations, leading to a continuous increase in computation time and storage overhead, becoming a major bottleneck restricting the electromagnetic scattering analysis of large-scale group targets. Therefore, simply relying on algorithm-level optimization is no longer sufficient to further improve overall computational performance.

[0004] Against this backdrop, there is a need for an electromagnetic scattering calculation method that can be adapted to high-parallelism computing platforms, making full use of the advantages of parallel computing hardware such as GPUs in large-scale matrix operations and data parallel processing, so as to significantly improve the efficiency of electromagnetic scattering and RCS calculation of group targets while ensuring computational accuracy. Summary of the Invention

[0005] The purpose of this invention is to provide a CPU-GPU-based method for efficient parallel computation of the RCS of large-scale group targets, which can achieve efficient and accurate computation of the RCS of large-scale group targets.

[0006] The technical solution to achieve the purpose of this invention is as follows: a high-efficiency parallel computing method for group target RCS based on CPU-GPU, the specific steps of which are as follows:

[0007] Step 1: Solve for the eigenvalues ​​and characteristic currents of a single target based on the generalized eigenvalue equations. Linearly reconstruct the initial induced currents on the surface of each target using a finite number of characteristic current terms. Convert the linear reconstruction from the original loop-level computation mode to matrix-level parallel computation. Divide the matrix operation task into multiple matrix blocks and allocate each matrix block to a thread block of the GPU. Multiple threads within the thread block complete the computation in parallel, thereby achieving parallel acceleration of the initial induced current calculation process on the target surface.

[0008] Step 2: Employing the multi-layer fast multipole method, the electromagnetic coupling between targets in a group of targets is accelerated under a CPU-GPU heterogeneous parallel architecture to obtain the coupling induced current. An octree structure required for MLFMA is constructed on the CPU (host side), and wavenumber and unit vector parameters are calculated. These parameters are then transferred to the GPU's global memory (device side). In the calculation of aggregation and configuration factors, each triangle is assigned a thread, and the discrete surface elements within the target are processed in parallel as independent computational units. In the calculation of transfer factors, each box pair is assigned a thread block, and each beam is assigned a thread. Transfer operations between different grouped box pairs are allocated to parallel computing resources, and multiple beam components are calculated in parallel within each grouped box.

[0009] Step 3: Superimpose the initial induced current of the target in Step 1 with the inter-target coupling induced current in Step 2 to calculate the total scattering field of the group of targets; based on the principle of allocating one thread to each triangle, perform parallel calculation and superposition of the far-field scattering contribution of the corresponding inner edge of each triangle element under the conditions of incident angle and receiving angle to obtain the far-field scattering result of each target; perform corresponding phase calculation of the scattering field of each target in the CPU and superimpose it into the total scattering field to calculate the RCS of the group of targets.

[0010] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0011] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the above-described method.

[0012] A computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described method.

[0013] Compared with the prior art, the significant advantages of the present invention are as follows: The present invention provides a CPU-GPU-based efficient parallel computing method for group target RCS. It adopts a CPU and GPU collaborative computing architecture, reconstructs the traditional scalar or loop-based electromagnetic scattering calculation process into a matrix-level parallel computing mode, and combines a multi-level parallel partitioning strategy at the target level, surface element level and plane wave component level to fully exploit the computing power of the GPU in large-scale parallel computing, avoid the efficiency bottleneck caused by the traditional serial solution method, and effectively improve the overall efficiency of group target electromagnetic scattering calculation. Attached Figure Description

[0014] Figure 1 This is a schematic diagram of the overall process of a CPU-GPU-based efficient parallel computing method for group target RCS provided by the present invention.

[0015] Figure 2 This is a schematic diagram of the parallel strategy for aggregation and configuration matrix filling.

[0016] Figure 3 This is a schematic diagram of the parallel strategy for filling the transition matrix.

[0017] Figure 4 This is a schematic diagram of a far-field superposition parallel strategy.

[0018] Figure 5 This is a comparison chart of the RCS results for a flock of 48 birds.

[0019] Figure 6 This is a comparison chart of the RCS results for a swarm of 100 drones. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Combination Figure 1 This invention provides a CPU-GPU-based method for efficient parallel computation of group target RCS, the method comprising:

[0022] Step 1: Based on the generalized eigenvalue equation, the eigenvalues ​​and characteristic currents of a single target are obtained. The initial induced currents on the surface of each target are then linearly reconstructed using a finite number of characteristic current terms. This linear reconstruction is transformed from the original loop-level computation mode to matrix-level parallel computation. The matrix operation task is divided into multiple matrix blocks, and each matrix block is allocated to a thread block of the GPU. Multiple threads within the thread block complete the computation in parallel, thereby achieving parallel acceleration of the initial induced current calculation process on the target surface.

[0023] First, the impedance matrix of the target needs to be constructed based on the equivalence principle, and the corresponding eigenvalues ​​and characteristic currents need to be solved using the generalized eigenvalue equations. Based on the extracted eigenvalues ​​and characteristic modes, the induced current on the target surface is approximated as a superposition of finite number of characteristic currents:

[0024]

[0025] in, This represents the induced current on the target surface. Indicates the number of truncated terms. This represents the pattern weight coefficient corresponding to each feature pattern. This represents the nth characteristic current;

[0026] The characteristic currents of each target in the target group are reconstructed using induced current independently. The computational flow for linear reconstruction of the induced current is reorganized, transforming the original loop-level computation into matrix-level parallel computation.

[0027]

[0028] in, Indicates the first of each objective The first mode corresponds to the th The characteristic current corresponding to the inner edge of the strip. Indicates the first The first target corresponds to the The pattern weight coefficients corresponding to each pattern Indicates the first The first target corresponds to the The induced current corresponding to the inner edge of the strip.

[0029] First, the characteristic current, normalized characteristic current, and corresponding mode weight parameters required for reconstructing the induced current are transferred to the GPU's global memory (which is the device side) all at once, and then reused in the initial induced current reconstruction and subsequent coupled iterative calculations, thereby reducing the number of data transfers between the GPU and the CPU and improving overall computational efficiency.

[0030] The matrix operation task is divided into multiple matrix blocks, and each matrix block is assigned to a thread block on the GPU, where multiple threads within the thread block perform the calculations in parallel. By calling the matrix operation acceleration interface on the GPU, the parallel computing capabilities of the GPU in large-scale matrix operations are fully utilized, achieving fine-grained data parallelism and effectively improving the computational efficiency of the mode linear reconstruction induced current.

[0031] Step 2: Employing the Multilevel Fast Multipole Method (MLFMA) in a CPU-GPU heterogeneous parallel architecture, the electromagnetic coupling between targets in a swarm is accelerated to obtain the coupled induced current. The CPU, acting as the host, constructs the octree structure required for MLFMA and calculates parameters such as wavenumber and unit vector, transferring these parameters to the GPU's global memory. The calculation of aggregation and configuration factors adopts the idea of ​​"one thread per triangle," treating discrete surface elements within the target as independent computational units for parallel processing. The calculation of transfer factors employs the ideas of "one thread block per box pair" and "one thread per beam," allocating transfer operations between different grouped box pairs to parallel computing resources and performing parallel calculations on multiple beam components within each grouped box.

[0032] MLFMA is used to accelerate the calculation of electromagnetic coupling between targets in a target swarm:

[0033]

[0034] in Expressed as permeability in free space, Represents angular frequency. Indicates wave number; It is made of objects The aggregation of coupling effects generated by the induced currents at various source points on the current level box or the propagation of coupling effects from the center of the current level box to the center of the parent level box. It is a translation of the coupling effect between different boxes at the same level. From object The decoupling of the coupling effect from the center of the parent box to the center of the child box, or the distribution of the coupling effect from the center of the finest level box to the object. A single field point on the surface, It is an object For objects The coupled right-hand vector; The double integral on a sphere representing a unit radius of spectral space. Represents the scattered wave vector. Represents objects From the center of the box to the object The position vector of the box's center. Represents objects From the center of the box to the object The distance from the center of the box.

[0035] The specific expressions for each factor are as follows:

[0036] Configuration factor: ;

[0037] Aggregation factor: ;

[0038] Transfer factor: ;

[0039] in It is an object The unknown quantity inside the box. It is an object The unknown quantity inside the box; Indicates from object From the center of the box to the object The position vector of the box's center. Representation Object The position vector from the unknown quantity inside the box to the center of the corresponding box. Representation Object The position vector from the unknown quantity inside the box to the center of the corresponding box; , Representing objects respectively and The corresponding basis functions within; It is a configuration factor. It is a aggregation factor. It is a transfer factor. The wavenumber represents free space; Indicates unit dyadicity, This represents the integral region contained within a certain box in the object. Represents the entire integration region of the object. It is a transverse projection dyadic operator.

[0040] First, the octree structure required for MLFMA is built on the CPU, and the corresponding wavenumber, unit vector and other parameters are calculated and transferred to the global memory of the GPU.

[0041] The calculation process for aggregation and configuration factors adopts the idea of ​​"one thread per triangle." A parallel mapping method based on a two-dimensional mesh is used, specifically mapping the target index to the first dimension of the mesh, and mapping the block index of a single target triangle facet to the second dimension of the mesh. Discrete facets within the target are treated as independent computational units for parallel processing. By using parallel threads to process different targets and multiple facets within them simultaneously, the parallel calculation and filling of aggregation and configuration factors are completed. Specific parallel strategies are as follows: Figure 2 As shown.

[0042] The calculation of the transfer factor adopts the idea of ​​"allocating one thread block per box pair" and "dividing each beam into one thread." A parallel mapping method based on a one-dimensional grid is used, specifically mapping the grouped box pair indices to the first dimension of the grid to distribute the transfer calculations of different grouped box pairs across parallel computing resources. Within each thread block, multiple plane wave components belonging to the same grouped box pair complete the transfer factor calculation in a thread-level parallel manner, achieving efficient parallelization of the overall computation. Specific parallel strategies are as follows: Figure 3 As shown.

[0043] Step 3: Superimpose the initial induced current of the target from Step 1 with the inter-target coupling induced current from Step 2 to calculate the total scattered field of the group of targets. Using the idea of ​​"one thread per triangle," the far-field scattering contributions of the corresponding inner edges of each triangle element under the conditions of incident and receiving angles are calculated in parallel and superimposed to obtain the far-field scattering results of each target. In the CPU, the phase of each target's scattered field is calculated and superimposed to form the total scattered field, thereby calculating the RCS of the group of targets.

[0044] The total scattered field of a group of targets is a linear superposition of the scattered fields of each individual target, as shown below:

[0045]

[0046] in It is the scattered wave vector. It is the incident wave vector. It is the first The position vector of the target centroid, the angle of incidence is The receiving angle is This indicates that the computational complexity of calculating the scattered electric field of a cluster of targets is related to the number of targets. They have a linear relationship. Indicates pitch angle, Indicates azimuth; It is the overall scattered field of a group of targets. It is the scattering field of each individual target.

[0047] First, the target geometric information required for far-field superposition calculation (including the vertex coordinates, normal vectors, and area of ​​each discrete triangular element) and the target surface induced current data obtained by linear reconstruction in step 2 are transferred to the global memory of the GPU.

[0048] A parallel mapping approach based on a two-dimensional computational grid is adopted to divide the far-field computation task. Specifically, the combined index of the receiving angle is mapped to the first dimension of the grid, and the block index of the total discrete triangular facets of the target group is mapped to the second dimension of the grid. The dimensions of the receiving angle and the discrete facets of the target surface are considered simultaneously, so that the far-field scattering contributions of each angle and each triangle facet corresponding to the inner edge can be executed in parallel as independent computational units and superimposed.

[0049] During GPU kernel execution, each parallel thread determines its corresponding angle combination and triangle element number based on its location within the computational grid. Using the discrete integral formula for electromagnetic scattering and pre-stored geometric parameters and induced current data in the GPU's global memory, it calculates in parallel the far-field scattering contribution of each angle and the corresponding inner edge of each triangle element, and then superimposes the scattered fields under the corresponding target at the corresponding angle. The specific parallel strategy is as follows: Figure 4 As shown.

[0050] The far-field scattering results of each target are transferred from the GPU to the CPU. The phase of the scattered fields of each target is calculated in the CPU and superimposed to form the total scattered field. The RCS of the group of targets is then calculated.

[0051] The present invention will now be described in detail with reference to the embodiments and accompanying drawings.

[0052] Example

[0053] This embodiment constructs and analyzes a flock of 48 birds and a swarm of 100 fixed-wing UAVs. The distance between the centers of mass of two adjacent birds is 0.65m; the distance between the centers of mass of two adjacent UAVs is 3.3m. The wingspan * body length of the bird model is approximately... The wingspan * fuselage size of the drone model is approximately In real flight, the attitudes of targets often exhibit random variations due to environmental factors such as airflow and wind speed. To simulate this dynamic characteristic, this example randomly generates the attitude angles (including pitch, yaw, and roll angles) of birds and drones within a set range during electromagnetic modeling to more realistically reflect the actual scattering scenario. A plane wave is incident, the operating frequency of the 48 birds is 13 GHz, and the incident angle is... , The number of discrete unknowns for a single bird is 79,674, the number of patterns selected for each bird is 2,000, and the number of unknowns for the entire flock is 3,824,352. The swarm of 100 drones operates at a frequency of 3 GHz and an incident angle of [missing information]. , There are 36,717 discrete unknowns for a single drone, 1,400 modes selected for each drone, and 3,671,700 unknowns for the entire drone swarm. Figure 5 , Figure 6 The calculation results of the RCS for two group target structures are presented respectively. The red solid line represents the RCS calculated by FEKO-MLFMA, and the blue dashed line represents the RCS calculated by CPU-GPU collaborative acceleration. Table 1 shows the efficiency of group target RCS calculation in this embodiment.

[0054] Table 1. Statistics on the computational efficiency of RCS for group targets

[0055]

[0056] Among them, FEKO-MLFMA is the Fast Multipole Algorithm in the commercial software FEKO, CM-MLFMA is the Eigenmodulus-Fast Multipole Algorithm, and CM-MLFMA-GPU is a hardware platform acceleration of the Eigenmodulus-Fast Multipole Algorithm using a GPU.

[0057] from Figure 5 , Figure 6 It can be seen that this embodiment can accurately calculate the RCS of the target. The coupling process in Table 1 includes the transfer factor and the aggregation-transfer configuration process. According to the comparison in Table 1, it can be seen that after using the GPU to accelerate the key calculation modules such as induced current reconstruction, aggregation-transfer-configuration factor and far-field superposition in parallel, the total calculation time in this embodiment is greatly reduced, achieving an order-of-magnitude improvement in calculation efficiency, with a speedup ratio of about 21, which can quickly calculate the RCS of large-scale group targets.

[0058] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention. All components not explicitly stated in this embodiment can be implemented using existing technology.

Claims

1. A CPU-GPU-based method for efficient parallel computation of group target RCS, characterized in that, include: Step 1: Solve for the eigenvalues ​​and characteristic currents of a single target based on the generalized eigenvalue equation, and reconstruct the initial induced currents on the surface of each target using a finite number of characteristic current terms. The linear reconstruction is transformed from the original loop-level computation mode to matrix-level parallel computation. The matrix operation task is divided into multiple matrix blocks, and each matrix block is assigned to a thread block of the GPU. Multiple threads within the thread block complete the computation in parallel, thereby achieving parallel acceleration of the initial induced current calculation process of the target surface. First, the impedance matrix of the target needs to be constructed based on the equivalence principle, and the corresponding eigenvalues ​​and characteristic currents are solved using the generalized eigenvalue equations. Based on the extracted eigenvalues ​​and characteristic currents, the induced current on the target surface is approximated as a superposition of finite number of characteristic currents. ； in, This represents the induced current on the target surface. Indicates the number of truncated terms. This represents the mode weighting coefficient corresponding to each characteristic current. This represents the nth characteristic current; In a group of targets, the characteristic currents of each target are reconstructed using induced current, and the calculation processes are independent of each other; GPU parallel computing transforms the original loop-level calculations into matrix-level parallel calculations. ； in, Indicates the first of each objective The first mode corresponds to the th The characteristic current corresponding to the inner edge of the strip. Indicates the first The first target corresponds to the The pattern weight coefficients corresponding to each pattern Indicates the first The first target corresponds to the The induced current corresponding to the inner edge of the strip; The matrix operation task is divided into multiple matrix blocks, and each matrix block is assigned to a thread block of the GPU, where multiple threads within the thread block complete the calculation in parallel. Step 2: Employing the multi-layer fast multipole method, the electromagnetic coupling between targets in a group of targets is accelerated under a CPU-GPU heterogeneous parallel architecture to obtain the coupling induced current. An octree structure required for MLFMA is constructed on the CPU (host side), and wavenumber and unit vector parameters are calculated. These parameters are then transferred to the GPU's global memory (device side). In the calculation of aggregation and configuration factors, each triangle is assigned a thread, and the discrete surface elements within the target are processed in parallel as independent computational units. In the calculation of transfer factors, each box pair is assigned a thread block, and each beam is assigned a thread. Transfer operations between different grouped box pairs are allocated to parallel computing resources, and multiple beam components are calculated in parallel within each grouped box. Step 3: Superimpose the initial induced current of the target in Step 1 with the inter-target coupling induced current in Step 2 to calculate the total scattering field of the group of targets; based on the principle of allocating one thread to each triangle, perform parallel calculation and superposition of the far-field scattering contribution of the corresponding inner edge of each triangle element under the conditions of incident angle and receiving angle to obtain the far-field scattering result of each target; perform corresponding phase calculation of the scattering field of each target in the CPU and superimpose it into the total scattering field to calculate the RCS of the group of targets.

2. The efficient parallel computing method for group target RCS based on CPU-GPU according to claim 1, characterized in that, In step 2, MLFMA is used to accelerate the computation of electromagnetic coupling between targets in a group of targets under a CPU-GPU heterogeneous parallel architecture: ； in Expressed as permeability in free space, Represents angular frequency. Indicates wave number; It is made of objects The aggregation of coupling effects generated by the induced currents at various source points on the current level box or the propagation of coupling effects from the center of the current level box to the center of the parent level box. It is a translation of the coupling effect between different boxes at the same level. From object The decoupling of the coupling effect from the center of the parent box to the center of the child box, or the distribution of the coupling effect from the center of the finest level box to the object. A single field point on the surface, It is an object For objects The coupled right-hand vector; The double integral on a sphere representing a unit radius of spectral space. Represents the scattered wave vector. Represents objects From the center of the box to the object The position vector of the box's center. Represents objects From the center of the box to the object The distance between the centers of the boxes; The specific expressions for each factor are as follows: Configuration factor: ; Aggregation factor: ; Transfer factor: ; in It is an object The unknown quantity inside the box. It is an object The unknown quantity inside the box; Indicates from object From the center of the box to the object The position vector of the box's center. Representation Object The position vector from the unknown quantity inside the box to the center of the corresponding box. Representation Object The position vector from the unknown quantity inside the box to the center of the corresponding box; , Representing objects respectively and The corresponding basis functions within; It is a configuration factor. It is a aggregation factor. It is a transfer factor. The wavenumber represents free space; Indicates unit dyadicity, This represents the integral region contained within a certain box in the object. Represents the entire integration region of the object. It is a transverse projection dyadic operator; First, the octree structure required for MLFMA is built on the CPU, and the corresponding wave number and unit vector parameters are calculated and transferred to the global memory on the GPU. In the process of calculating aggregation and configuration factors, based on the idea of ​​allocating one thread to each triangle, a parallel mapping method based on two-dimensional grid is adopted. The target index is mapped to the first dimension of the grid, and the block index of the single target triangle surface element is mapped to the second dimension of the grid. The discrete surface elements inside the target are treated as independent computing units for parallel processing. In the calculation of the transfer factor, based on the idea of ​​allocating a thread block for each box pair and dividing each beam into a thread, a parallel mapping method based on a one-dimensional grid is adopted to map the grouped box pair index to the first dimension of the grid, so as to realize the distribution of transfer calculation of different grouped box pairs on parallel computing resources; within each thread block, multiple plane wave components belonging to the same grouped box pair complete the calculation of the transfer factor in a thread-level parallel manner, realizing the efficient parallelization of the overall calculation.

3. The efficient parallel computing method for group target RCS based on CPU-GPU according to claim 2, characterized in that, In step 3, the total scattered field of the group of targets is formed by the linear superposition of the scattered fields of each individual target, as shown below: ； in It is the scattered wave vector. It is the incident wave vector. It is the first One goal, It is the first The position vector of the target centroid, the angle of incidence is The receiving angle is , Indicates pitch angle, Indicates azimuth; It is the overall scattered field of a group of targets. It is the scattered field of each individual target; The far-field computation task is divided by a parallel mapping method based on a two-dimensional computational grid. Specifically, the combined index of the receiving angle is mapped to the first dimension of the grid, and the block index of the total discrete triangular surface elements of the target group is mapped to the second dimension of the grid. At the same time, the dimensions of the receiving angle and the discrete surface elements of the target surface are considered, so that the far-field scattering contribution of each angle and each triangular surface element can be executed in parallel as independent computational units and superimposed. Finally, the phase of each target's scattering field is calculated in the CPU and superimposed to form the total scattering field, and then the RCS of the group of targets is calculated.

4. An electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-3.

6. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method described in any one of claims 1-3.

Images