A method, storage medium, apparatus, and program product for classifying wafer defects.

Abstract

This specification provides a method, storage medium, device, and program product for classifying wafer defects. The method utilizes a wafer classification model, which includes a feature extraction module and a classifier. The method includes: acquiring a defect image to be classified; determining several sub-defect images with the same image size but different defect proportions based on the defect image; extracting features from each sub-defect image using the feature extraction module to obtain defect features; and using the classifier to determine the defect type corresponding to the defect image to be classified based on the defect features. In the process of detecting wafer defects, the defect image to be classified is first processed to obtain several sub-defect images representing the defects in a magnified manner. This ensures that even when the defect size in the defect image is extremely small, the classification accuracy is still high when determining the defect type corresponding to the defect image to be classified using the sub-defect images.

Description

Technical Field

[0001] The embodiments in this specification belong to the field of data processing technology, and more specifically, relate to a method, storage medium, apparatus and program product for classifying wafer defects. Background Technology

[0002] In the field of wafer defect detection, AI algorithm detection has many advantages over traditional manual inspection—higher detection accuracy, higher detection efficiency, and lower labor costs.

[0003] Currently, inspection instruments can automatically identify defective wafers and acquire defect images targeting the defect area. However, determining the type of defect in the wafer still requires further analysis of the acquired defect images.

[0004] However, traditional machine learning models struggle to directly adapt to the high-precision analysis tasks of defect images. For example, with Convolutional Neural Networks (CNN) models, after a defect image is input into the CNN model, it undergoes multiple rounds of downsampling due to the limited parameter size of the CNN. However, wafer manufacturing demands extremely high precision, with surface roughness typically required at the sub-nanometer level. Consequently, some wafer defects are extremely small, and multiple rounds of downsampling can lead to a loss of defect information, significantly impacting the accuracy of wafer defect classification.

[0005] Therefore, the present invention provides a method, storage medium, apparatus and program product for classifying wafer defects. Summary of the Invention

[0006] The embodiments in this specification are intended to provide a method, storage medium, apparatus, and program product for classifying wafer defects.

[0007] To achieve the above objectives, this specification provides a method for classifying wafer defects. The method is executed using a wafer classification model, which includes a feature extraction module and a classifier. The method includes:

[0008] Obtain the image of the defect to be classified;

[0009] Based on the defect image to be classified, determine several sub-defect images with the same image size but different proportions of defect parts;

[0010] Using the feature extraction module, features are extracted from the several sub-defect images to obtain the features of each defect;

[0011] Using the classifier, the defect type corresponding to the defect image to be classified is determined based on the defect features.

[0012] In some implementations, based on the defect image to be classified, several sub-defect images with the same image size but different proportions of defect portions are determined, specifically including:

[0013] Several local images are obtained by using several screenshot frames of different sizes to capture the target regions containing defects in the defect image to be classified.

[0014] Each local image is magnified to the same size as the defect image to be classified, and the magnified local images and the defect image to be classified are used as several sub-defect images.

[0015] In some implementations, the defect type corresponding to the defect image to be classified is determined based on each defect feature, specifically including:

[0016] The various defect features are fused together to obtain the fused feature;

[0017] Based on the fusion features, the defect type corresponding to the defect image to be classified is determined.

[0018] In some implementations, the various defect features are fused to obtain fused features, specifically including:

[0019] The weighted sum of each defect feature yields the fused feature.

[0020] In some implementations, the various defect features are fused to obtain fused features, specifically including:

[0021] For any position in the fusion feature, the element at that position in the fusion feature is determined based on the maximum value of each element corresponding to that position in each defect feature, wherein the fusion feature and the defect feature have the same dimension.

[0022] In some implementations, the various defect features are fused to obtain fused features, specifically including:

[0023] By superimposing n defect features with N channels according to their channel dimensions, a joint defect feature with n×N channels is obtained.

[0024] The joint defect features are convolved using an n-channel 1×1-dimensional convolution kernel, and the convolution result is activated to obtain an N-channel fused feature.

[0025] Some implementations also include:

[0026] Obtain training samples, which include sample defect images and defect labels corresponding to the sample defect images;

[0027] Based on the sample defect images, determine several sub-sample images with the same image size but different proportions of defective parts;

[0028] Using the feature extraction module, feature extraction is performed on the several sub-sample images to obtain the defect features of each sample;

[0029] Using the classifier, the predicted score for each defect type corresponding to the defect image of each sample is determined based on the defect features of each sample;

[0030] The wafer classification model is trained based on the predicted score and the defect label.

[0031] A second aspect of this specification provides a computer-readable storage medium having a computer program stored thereon that, when executed in a computer, causes the computer to perform the method described in the first aspect.

[0032] A third aspect of this specification provides a computing device including a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method described in the first aspect.

[0033] This specification provides a fourth aspect of a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the method as described in the first aspect.

[0034] In the process of detecting wafer defects, at least one implementation method provided in this specification first processes the image of the defect to be classified to obtain several sub-defect images that represent the defects in magnified form. This ensures that when determining the defect type corresponding to the image of the defect to be classified using each sub-defect image, even if the defect size in the image of the defect to be classified is extremely small, the classification accuracy is still high. Attached Figure Description

[0035] To more clearly illustrate the technical solutions of the embodiments in this specification, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0036] Figure 1 This is a schematic diagram of the structure of a wafer sorting system provided in an embodiment of the present invention;

[0037] Figure 2 A flowchart illustrating a method for classifying wafer defects according to an embodiment of the present invention;

[0038] Figure 3This is a schematic diagram of a method for superimposing defect features according to an embodiment of the present invention;

[0039] Figure 4 This is a flowchart illustrating a training method for a wafer classification model provided in an embodiment of the present invention. Detailed Implementation

[0040] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0041] Figure 1 This specification illustrates a schematic diagram of a wafer classification system, which includes an image processing module and a wafer classification model. When the processing platform begins to identify the defect type corresponding to an image of defects to be classified, the image can be input into the wafer classification system pre-deployed in the processing platform. The wafer classification system first uses the image processing module to determine several sub-defect images based on the image of defects to be classified, and then inputs each sub-defect image into the wafer classification model to determine the defect type corresponding to the image of defects to be classified.

[0042] Figure 2 This specification illustrates a flowchart of a method for classifying wafer defects, which utilizes a pre-trained wafer classification model. The method includes:

[0043] Step S201: Obtain the image of the defect to be classified.

[0044] First, the processing platform can acquire images of defects to be classified that are pre-stored in the internal database of the processing platform or other communicable addresses, and input the images of defects to be classified into the image processing module.

[0045] The defect image to be classified can be obtained by acquiring the wafer to be inspected using an acquisition device. A single defect image may contain a complete wafer or a portion of a wafer to be classified; this specification does not impose any limitations. Furthermore, during defect inspection of each wafer, the inspection instrument can identify defective wafers to be classified using various common inspection methods such as X-ray inspection and laser scanning, and determine the location of the defects within the wafer. The acquisition device can then acquire an image targeting that location, ensuring that the acquired defect image clearly shows the defects on the wafer to be inspected.

[0046] Furthermore, since the inspection instrument can determine the defective part in the wafer to be inspected, when the acquisition device acquires the defective part in the wafer to be inspected as the center for image acquisition, the central area of ​​the acquired defective image to be classified represents the defective part; or it can acquire the image with a uniform relative pose and mark the area corresponding to the defective part in the acquired defective image to be classified. This specification does not limit this.

[0047] Step S203: Based on the defect image to be classified, determine several sub-defect images with the same image size but different proportions of defect parts.

[0048] After acquiring the defect image to be classified, the processing platform can input the defect image to the image processing module so that the image processing module can determine several sub-defect images.

[0049] Specifically, the image processing module can first enlarge the defect image to be classified into several enlarged defect images of different sizes according to a preset scaling ratio. For example, when the scaling ratio is 1:2, the length and width of the defect image to be classified can be enlarged by 2 times to obtain the enlarged defect image corresponding to the scaling ratio of 1:2. Then, according to the size of the defect image to be classified, the defect part in each of the enlarged defect images can be cropped to obtain a sub-defect image with the area of ​​the defect part being 2×2 times that of the original.

[0050] Among them, the process of enlarging the defect image to be classified into an enlarged defect image can be achieved using common interpolation methods such as nearest neighbor interpolation and bilinear interpolation, which will not be elaborated here.

[0051] Therefore, as Figure 1 As shown, the same defect can occupy more pixels in a sub-defect image. On the one hand, this avoids the loss of defect information during feature extraction in subsequent steps; on the other hand, defect information can account for a higher proportion of the defect features extracted from the sub-defect image.

[0052] In some implementations, the image of the defect to be classified can be directly used as a sub-defect image. Thus, when n-1 different scaling ratios are preset, a total of n sub-defect images can be obtained.

[0053] Step S205: Using the feature extraction module, perform feature extraction on the several sub-defect images respectively to obtain the features of each defect.

[0054] After the image processing module determines each sub-defect image, the respective defect images can be input into the wafer classification model. The feature extraction module of the wafer classification model can then be used to extract features from each sub-defect image to obtain the defect features.

[0055] In some implementations, the feature extraction module may include several feature extraction networks with identical structures but different parameters. Each feature extraction network may consist of several convolutional layers, and each network corresponds to a different scaling ratio. When processing sub-defect images using the feature extraction module, a sub-defect image is input into the corresponding feature extraction network according to its scaling ratio to obtain the defect features corresponding to that sub-defect image. Thus, by inputting n sub-defect images into the feature extraction module, it can utilize the n feature extraction networks to obtain n defect features.

[0056] In some implementations, such as Figure 1 As shown, the feature extraction module may include a single feature extraction network. When processing each sub-defect image using this feature extraction module, each sub-defect image is sequentially input into the feature extraction network to obtain the defect features corresponding to each sub-defect image. Thus, by inputting n sub-defect images into the feature extraction module, the module can obtain n defect features using one feature extraction network.

[0057] It should be noted that regardless of the implementation method, the dimensions of each defect feature are the same. Therefore, the defect features corresponding to the sub-defect image with a larger defect proportion can contain more defect information, while the defect features corresponding to the sub-defect image with a smaller defect proportion can contain more global information, as well as information on the relative relationship between the defect and the wafer.

[0058] Step S207: Using the classifier, determine the defect type corresponding to the defect image to be classified based on the defect features.

[0059] After determining each defect feature, the defect features can be input into the classifier to determine the prediction score of each defect type for the defect image to be classified. The prediction score can be further converted into the probability of identifying the defect image to be classified as each defect type. Thus, the defect type with the highest probability can be determined as the defect type of the defect image to be classified.

[0060] In some implementations, the classifier may include n sets of identical neural network layers, each set corresponding to a preset scaling ratio. Specifically, each set of neural network layers may include a Global Average Pooling Layer (GAP Layer), a flattening layer, and a Fully Connected Layer (FC Layer). The GAP Layer transforms the defect features of dimension W×H×C into a 1×1×C three-dimensional tensor. The flattening layer flattens this 1×1×C three-dimensional tensor into a one-dimensional vector. The FC Layer determines the probability score as the output based on the one-dimensional vector input to the flattening layer.

[0061] When using a classifier to determine the defect type corresponding to n input defect features, each defect feature can be input into a corresponding set of neural network layers to obtain n probability scores output by the n sets of neural network layers. The n probability scores are then weighted and fused, and softmax and other functions are applied to determine the probability distribution of each defect type corresponding to the defect image to be classified.

[0062] Since each defect feature is extracted from a different sub-defect image, and the defect portion in the sub-defect image is larger than that in the original defect image to be classified, the defect type can be determined by using each defect feature, which can utilize the defect information of smaller defects while ensuring the use of the relative relationship information between defects and wafers.

[0063] like Figure 2 The method for classifying wafer defects shown first processes the image of the defect to be classified during the detection of wafer defects, and obtains several sub-defect images that represent the defects in magnified form. This ensures that when the defect type corresponding to the image to be classified is determined using each sub-defect image, even if the defect size in the image to be classified is extremely small, the classification accuracy is still high.

[0064] In some implementations, such as Figure 2 In step S203 shown, several screenshot frames of different sizes are used to capture the target area containing the defect in the defect image to be classified, resulting in several local images. Each local image is enlarged to the same size as the defect image to be classified, and the enlarged local images and the defect image to be classified are used as several sub-defect images.

[0065] The processing platform can preset several screenshot frames of different sizes, each corresponding to a different scaling ratio. For example, when the scaling ratio is 1:2, the length and width of the screenshot frame can be 1 / 2 of the defect image to be classified. Thus, when the captured local image is enlarged to the same size as the defect image to be classified, the area of ​​the defect part in the enlarged sub-defect image is 2×2 times the original.

[0066] In some implementations, such as Figure 1 The wafer classification system shown in the diagram includes a classifier in the wafer classification model that specifically comprises a fusion module and a set of neural network layers, such as... Figure 2 In step S207, the defect features are fused to obtain fused features, and the defect type corresponding to the defect image to be classified is determined based on the fused features.

[0067] Specifically, after obtaining the defect features, the fusion module can be used to fuse the defect features to obtain fused features, which are then input into subsequent neural network layers to determine the defect type corresponding to the defect image to be classified.

[0068] The fusion of various defect features can be achieved by splicing together n N-channel defect features along the channel dimension to obtain an n×N-channel fused feature, or by aligning and fusing elements in the defect features to obtain an N-channel fused feature. This will not be elaborated upon in this specification.

[0069] Therefore, after obtaining the defect features, the defect features can be directly fused together. Only one set of neural network layers is needed to obtain the prediction score corresponding to the fused feature.

[0070] It should be noted that the specific structure of a group of neural network layers in the classifier can be referred to the structure described in step S207, and will not be repeated here; the parameter dimensions of each neural network layer in the classifier can be set according to the dimension of the defect features.

[0071] In some implementations, such as Figure 2 In step S207, the defect features are weighted and summed to obtain the fused features.

[0072] After inputting each defect feature into the classifier, the fusion module in the classifier can perform a weighted summation of the elements in the i-th row, j-th column, and k-th channel of each defect feature to obtain the element in the i-th row, j-th column, and k-th channel of the fused feature.

[0073] Therefore, in the process of obtaining the fusion features, the defect features corresponding to each scaling ratio are fully utilized, ensuring that no information contained in each defect feature is omitted.

[0074] The weights corresponding to each defect feature can be set by the user in advance; or set as hyperparameters and adjusted by the user based on the training effect after pre-training the wafer classification model.

[0075] In some implementations, such as Figure 2 In step S207, for any position in the fusion feature, the element at that position in the fusion feature is determined based on the maximum value of each element corresponding to that position in each defect feature. The fusion feature and the defect feature have the same dimension.

[0076] After each defect feature is input into the classifier, the fusion module in the classifier can determine the maximum value of the element in the i-th row, j-th column, k-th channel of each defect feature, and use it as the element in the i-th row, j-th column, k-th channel of the fused feature.

[0077] Therefore, in the process of obtaining the fusion features, only the maximum value of the corresponding element in each defect feature is used. When the value of the element in the defect feature is related to the defect information, the fusion feature can more centrally represent the defect information.

[0078] In some implementations, such as Figure 2 In step S207, for any position in the fusion feature, the element at that position in the fusion feature is determined based on the minimum value of each element corresponding to that position in each defect feature. The fusion feature and the defect feature have the same dimension.

[0079] The specific effect of obtaining the fusion feature based on the minimum value is similar to that of using the maximum value as described above, and will not be repeated here.

[0080] In some implementations, such as Figure 2 In step S207, n defect features with N channels are superimposed in terms of channel dimension to obtain a joint defect feature with n×N channels. The joint defect feature is then convolved using an n-channel 1×1-dimensional convolution kernel, and the convolution result is activated to obtain a fused feature with N channels.

[0081] The fusion module in this classifier can be a convolutional layer. After each defect feature is input into the classifier, the fusion module can first superimpose the defect features to obtain joint defect features, then perform convolution processing on the joint defect features, and further activate the result of the convolution processing to finally obtain the fused features.

[0082] Specifically, for methods that superimpose defect features to obtain joint defect features, such as... Figure 3As shown in the figure, defect feature A is represented by channel a1-channel a2-channel a3, defect feature B is represented by channel b1-channel b2-channel b3, and the combined defect feature is channel a1-channel b1-channel a2-channel b2-channel a3-channel b3. Of course, other feasible superposition methods can also be used to obtain the combined defect feature, and this specification does not impose any restrictions here.

[0083] During the convolution process, by using an n-channel 1×1-dimensional convolution kernel and setting the stride of the channel dimension to n, the fusion features of the N channels can be obtained from the joint defect features of the n×N channels.

[0084] This specification also provides a training method for a wafer classification model, which can also be executed by a processing platform. The process of this method is as follows: Figure 4 As shown, it includes:

[0085] S401: Obtain training samples, the training samples including sample defect images and defect labels corresponding to the sample defect images.

[0086] The sample defect image and the defect image to be classified obtained in step S201 can be in the same format, thereby reducing the interference of format differences on the classification results of the trained wafer classification model; the defect labels corresponding to the sample defect images can be manually labeled and determined.

[0087] S403: Based on the sample defect image, determine several sub-sample images with the same image size but different proportions of defective parts.

[0088] The specific process of executing step S403 can be found in step S203, and will not be repeated here.

[0089] It should be noted that when using, such as Figure 4 The wafer classification model trained using the method shown executes as follows: Figure 2 When using the method shown, the scaling ratio used in step S203 must be the same as the scaling ratio used in step S403 to ensure the effectiveness of subsequent steps.

[0090] S405: Using the feature extraction module, feature extraction is performed on the several sub-sample images to obtain the defect features of each sample.

[0091] The specific process of executing step S405 can be found in step S205, and will not be repeated here.

[0092] S407: Using the classifier, determine the prediction score of each defect type corresponding to the sample defect image based on the defect features of each sample.

[0093] The specific process of executing step S407 can be found in step S207, and will not be repeated here.

[0094] S409: Train the wafer classification model based on the predicted score and the defect label.

[0095] As mentioned above, the probability distribution of each defect type corresponding to the sample defect image can be further determined based on the prediction scores. The defect label is the labeled defect type corresponding to the sample defect image. Based on the probability corresponding to each defect type and the labeled defect type, the gradient of each model layer in the wafer classification model can be determined using a preset loss function. Then, the wafer classification model can be trained according to the preset learning rate and the determined gradient.

[0096] The preset loss function can be any common loss function such as cross-entropy loss, KL divergence, or cosine similarity loss. This manual does not impose any restrictions on this.

[0097] In some implementations, when executing such Figure 4 The method shown can be used to train the wafer classification model for multiple rounds, and then it can be tested or verified.

[0098] Specifically, the complete dataset can be divided into a training set, a test set, and a validation set. The representation of each sample in the dataset is the same as that of the training samples; each sample contains a defect image and its corresponding defect label. Samples in the training set can be used as training samples to perform operations such as... Figure 4 The training process is shown below; the samples in the test set can be used as test samples to adjust the hyperparameters (model parameter size, learning rate, etc.) of the wafer classification model. The testing process can be referenced as follows: Figure 4 The method shown; the samples in the validation set can be used as validation samples to verify the classification ability of the wafer classification model. The validation process can be referred to as follows: Figure 2 The method shown allows for the preset of several hyperparameters, which are then used to determine the values ​​of each hyperparameter during model training, thereby improving training efficiency and reducing model computational overhead.

[0099] It should be understood that the descriptions such as "first" and "second" in this article are merely for the sake of simplicity in description and to distinguish similar concepts, and do not have any other limiting function.

[0100] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0101] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0102] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of each example have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application. The software modules can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disks, removable disks, CD-ROMs, or any other form of storage medium known in the art.

[0103] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for classifying wafer defects, characterized in that, The method utilizes a wafer classification model, which includes a feature extraction module and a classifier. The feature extraction module comprises several feature extraction networks with identical structures but different parameters. The classifier comprises n sets of neural network layers with identical structures, each set of neural network layers corresponding to a preset percentage of defective parts. The method includes: Obtain the image of the defect to be classified; Based on the defect image to be classified, determine several sub-defect images with the same image size but different proportions of defect parts; Using the feature extraction module, features are extracted from the several sub-defect images to obtain the features of each defect; Using the classifier, the defect type corresponding to the defect image to be classified is determined based on the defect features. Specifically, based on the defect image to be classified, several sub-defect images with the same image size but different proportions of defect portions are determined, including: Several local images are obtained by using several screenshot frames of different sizes to capture the target regions containing defects in the defect image to be classified. Each local image is magnified to the same size as the defect image to be classified, and the magnified local images and the defect image to be classified are used as several sub-defect images. Based on the aforementioned defect features, the defect type corresponding to the defect image to be classified is determined, specifically including: Each defect feature is input into a set of neural network layers in the classifier according to the proportion of the corresponding defect part, and n probability scores are obtained from the output of n sets of neural network layers. Based on the weighted fusion result of the n probability scores, the defect type corresponding to the defect image to be classified is determined.

2. The method as described in claim 1, characterized in that, After obtaining the characteristics of each defect, the following is also included: The classifier is used to fuse the various defect features to obtain fused features; Based on the fusion features, the defect type corresponding to the defect image to be classified is determined.

3. The method as described in claim 2, characterized in that, The various defect features are fused to obtain the fused features, which specifically include: The weighted sum of each defect feature yields the fused feature.

4. The method as described in claim 2, characterized in that, The various defect features are fused to obtain the fused features, which specifically include: For any position in the fusion feature, the element at that position in the fusion feature is determined based on the maximum value of each element corresponding to that position in each defect feature, wherein the fusion feature and the defect feature have the same dimension.

5. The method as described in claim 2, characterized in that, The various defect features are fused to obtain the fused features, which specifically include: By superimposing n defect features with N channels according to their channel dimensions, a joint defect feature with n×N channels is obtained. The joint defect features are convolved using an n-channel 1×1-dimensional convolution kernel, and the convolution result is activated to obtain an N-channel fused feature.

6. The method as described in claim 1, characterized in that, Also includes: Obtain training samples, which include sample defect images and defect labels corresponding to the sample defect images; Based on the sample defect images, determine several sub-sample images with the same image size but different proportions of defective parts; Using the feature extraction module, feature extraction is performed on the several sub-sample images to obtain the defect features of each sample; Using the classifier, the predicted score for each defect type corresponding to the defect image of each sample is determined based on the defect features of each sample; The wafer classification model is trained based on the predicted score and the defect label.

7. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-6.

8. A computing device comprising a memory and a processor, wherein the memory stores executable code, and the processor, when executing the executable code, implements the method of any one of claims 1-6.

9. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the method as claimed in any one of claims 1-6.

Images