An Enhancement Method and System for Irregular Tactile Graphics Data

Abstract

This invention relates to data processing technology, specifically to a method and system for enhancing irregular tactile graphic data. The method includes: eliminating outlier data in the tactile graphic data and standardizing the tactile graphic data; filling the tactile graphic data using rectangular or cross-shaped filling; determining the receptive field of a rectangle centered on the filled pixel for rectangular filling; determining the receptive field of a cross-shaped filling using the filled pixel as the center, where the receptive field encompasses all pixels in the corresponding row and column of the entire tactile graphic data; performing a weighted summation of all valid pixel values ​​within the receptive field followed by nonlinear projection, and using the projection result as the pixel value of the center pixel; and performing random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data. This invention solves the problems of missing edge information and confusion caused by low-resolution sensing that easily occur during the acquisition of tactile graphic data.

Description

Technical Field

[0001] This invention relates to data processing technology, specifically to an enhancement method and system for irregular tactile graphic data. Background Technology

[0002] In the current field of tactile perception, capturing irregularly patterned sensor data using high-dimensional sensors and processing it with visual methods has become a trend, especially in processing hand and foot pressure data. However, some studies show that cross-application of visual models to physiological signals may not necessarily improve performance. This is mainly due to two factors: First, the difference in resolution and data volume between the two types of data makes it difficult for visual models to find appropriate complexity to adapt to physiological data, ultimately affecting task performance. Second, the limitations of the data acquisition process. For example, most existing high-resolution sensors do not have high sensitivity, which may result in some loss of edge information. Furthermore, readout circuits are prone to row and column crosstalk at high resolutions. These factors reduce the amount of effective information in the data and create a vicious cycle with the first problem.

[0003] To address these issues, most researchers currently favor improving visual models to enhance accuracy, but this approach is costly. This invention shifts to data augmentation, improving the quality of tactile data by increasing its information content and randomness, thereby altering its original distribution.

[0004] However, current mainstream data augmentation methods are all vision-based. Conventional augmentation operations include, but are not limited to, flipping, cropping, adding noise, and masking. Furthermore, these conventional augmentation methods require the data itself to possess a certain degree of robustness. Tactile signals, as physiological signals, lack strong robustness due to their limited dimensionality. Therefore, when these augmentation methods are applied to tactile data, they often fail to achieve satisfactory augmentation results and may even disrupt the normal distribution of the data. There is a research gap regarding augmentation methods for physiological signals, especially tactile signals; therefore, this invention proposes a targeted augmentation scheme for tactile data. Summary of the Invention

[0005] To address the shortcomings and deficiencies of existing technologies, this invention proposes an enhancement method and system for irregular tactile graphic data. This method solves the problems of missing edge information and confusion caused by low-resolution sensing that are prone to occur during the acquisition of tactile graphic data. It increases the information content and diversity of tactile graphic data and improves the perceptual performance of the model.

[0006] The enhancement method adopted in this embodiment of the invention is as follows: an enhancement method for irregular tactile graphic data, comprising the following steps:

[0007] S1. Eliminate outlier data in the tactile graphic data and standardize the tactile graphic data.

[0008] S2. Use rectangular fill or cross fill to fill the tactile graphic data;

[0009] Rectangular filling uses the pixel to be filled as the center to determine the receptive field of the rectangle. All valid pixel values ​​within the receptive field are weighted and summed, and then non-linearly projected. The projection result is used as the pixel value of the center pixel. After one filling is completed, the center pixel is regarded as a valid pixel, and the pixel kernel is moved to the next pixel to continue the next filling.

[0010] Cross-filling determines the receptive field centered on the pixel to be filled. The receptive field is all pixels in the corresponding row and column of the entire tactile graphics data. The weighted sum of all valid pixel values ​​within the receptive field is then non-linearly projected, and the projection result is used as the pixel value of the center pixel.

[0011] S3. Perform random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data.

[0012] Preferably, in step S2, the pixel value of the center pixel in the rectangular filling is calculated using the following formula:

[0013]

[0014] Where k is the preset receptive field size; valid represents the number of valid pixels within the receptive field; (i,j) represents the relative coordinates of a pixel within the receptive field; X(i,j) represents the pixel value of the center pixel obtained during filling; F(i,j) represents the pixel value corresponding to pixel (i,j) in the receptive field; M(i,j) indicates whether pixel (i,j) is a valid pixel, if it is valid, the pixel is "1", otherwise it is "0"; F represents the matrix of pixels within the receptive field; M represents the discrimination matrix for each pixel within the receptive field to determine whether it is a valid pixel; the sigmoid function represents projecting the summation value between 0 and 1.

[0015] Preferably, the pixel value of the center pixel in step S2 is calculated using the following formula:

[0016]

[0017] Where X(i,j) represents the pixel value of the center pixel obtained during filling; X(i,a) represents the pixel in the i-th row and a-th column within the receptive field, and X(b,j) represents the pixel in the b-th row and j-th column within the receptive field; the sigmoid function represents projecting the summation value between 0 and 1.

[0018] Preferably, in step S3, the random migration processing uses the idea of ​​convex combination to splice the cut-off parts of the two tactile graphic data sample inputs. The first tactile graphic data sample input is used as the enhanced sample object, and a rectangle is cut out from the first tactile graphic data sample input. The original data in the rectangle and the data part of the same shape randomly selected from the second tactile graphic data sample input after being filled with the same type of data are convexly combined. The same size shape is cut out from the randomly selected second tactile graphic data sample input and incorporated into the first tactile graphic data sample input.

[0019] Furthermore, the random migration process in step S3 is represented as follows:

[0020] x′ a =U⊙x a +V⊙x b

[0021]

[0022] where x a and x b Input tactile graphic data samples after they have undergone the same type of data filling processing; define U and V to represent the corresponding binary masks, that is, to extract x. a x b The positions where U∈{1,p} are deleted, retained, or weighted and merged from two haptic graphic data sample inputs. 32×32 V∈{0,1-p} 32×32 p as x a x b The coefficients for fusing two tactile graphic data samples at the shearing site; ⊙ represents element-wise multiplication, ||·|| represents selecting the shape with 1 in the matrix. It indicates that the shapes are exactly the same.

[0023] The enhancement system technical solution adopted in this embodiment of the invention is: an enhancement system for irregular tactile graphic data, comprising the following modules:

[0024] The preprocessing module is used to eliminate outlier data in the tactile graphic data and to standardize the tactile graphic data.

[0025] The data filling module is used to fill haptic graphic data with rectangular or cross-shaped fill.

[0026] Rectangular filling uses the pixel to be filled as the center to determine the receptive field of the rectangle. All valid pixel values ​​within the receptive field are weighted and summed, and then non-linearly projected. The projection result is used as the pixel value of the center pixel. After one filling is completed, the center pixel is regarded as a valid pixel, and the pixel kernel is moved to the next pixel to continue the next filling.

[0027] Cross-filling determines the receptive field centered on the pixel to be filled. The receptive field is all pixels in the corresponding row and column of the entire tactile graphics data. The weighted sum of all valid pixel values ​​within the receptive field is then non-linearly projected, and the projection result is used as the pixel value of the center pixel.

[0028] The random migration module performs random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data.

[0029] Compared with the prior art, the technical effects achieved by the present invention include:

[0030] 1. This invention can be easily embedded and integrated offline into network training; after data augmentation, the information content of tactile graphic data will be increased, and the deep learning model can learn more useful knowledge from it to help with classification, and the confusion between different categories will also be alleviated due to the random transfer of data parts.

[0031] 2. This invention can improve the quality of tactile perception images, helping the model to better complete perception tasks. Compared with the existing technology of improving task performance by improving the model, this invention is lower in cost, simpler and more intuitive, and the effect is more obvious.

[0032] 3. This invention is not only applicable to irregular tactile graphic data, but can also be easily applied to other physiological signals, such as EEG signals and plantar pressure signals, thus satisfying the method's versatility over a wide range and possessing certain application value. Attached Figure Description

[0033] Figure 1 This is a flowchart of an enhancement method for irregular tactile graphic data in an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of the haptic graphic data format;

[0035] Figure 3 These are schematic diagrams of data filling, where (a) is a schematic diagram of rectangular filling and (b) is a schematic diagram of cross filling;

[0036] Figure 4 This is a schematic diagram of random migration processing. Detailed Implementation

[0037] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0038] Example

[0039] This embodiment provides an enhancement method for irregular tactile graphic data, such as... Figure 1 As shown, this method can standardize the acquired tactile graphic data, adding more information to the original tactile graphic data while increasing the differences between different categories. It addresses the limitations of the front-end acquisition process from the back-end data processing stage, improving perception performance. Specifically, the entire method can include the following steps:

[0040] S1. Preprocess the tactile graphic data.

[0041] Tactile graphic data can be acquired through a sensing unit (e.g., a sensor). Let the tactile graphic data acquired by the sensor be... And assume it is as follows Figure 2 The format in which it is represented. To initially improve the quality and accuracy of haptic graphics data, preprocessing is first required. Preprocessing mainly consists of two steps:

[0042] S11. Eliminate outlier data in tactile graphic data.

[0043] First, we define a theoretical upper limit threshold x for haptic graphics data globally. u Lower limit threshold x l Then iterate through all tactile graphic data. If any tactile graphic data x contains a sensor unit whose return value exceeds the upper limit threshold x, then... u In cases where the tactile graphic data is located, the image frame containing that data is deemed invalid. This phenomenon typically corresponds to the sensor unit encountering an abnormally sharp object; if retained, its excessive noise would affect the normal distribution of the entire data in subsequent processing. Simultaneously, if all values ​​in any tactile graphic data x are less than the lower threshold x... l If this happens, the image frame containing the tactile graphic data will be considered a non-contact frame, containing no information, and thus deemed an invalid frame.

[0044] S12. Standardize the tactile graphic data x to obtain standardized data x. norm The formula for standardization is:

[0045]

[0046] To preserve the physical meaning of the information in the data, the standardization of tactile graphic data will be carried out within a preset acquisition time, meaning that the tactile graphic data of different acquisition objects will be standardized separately.

[0047] This embodiment defines the acquisition process for different acquisition objects as different acquisition rounds. In each acquisition round, x ave x max and x min These represent the average, maximum, and minimum values ​​of the subset of tactile graphic data in this collection round, respectively. The Clip(·) function is used to restrict the values ​​in parentheses to the range [0,1]. This is done because the return values ​​of the sensor units generally have a relatively large threshold range. Standardization will give different tactile graphic data the same dimensions and units, making the return values ​​of all sensor units on the same order of magnitude, and can also accelerate the convergence speed during model training.

[0048] S2. Fill in the tactile graphic data.

[0049] When high-resolution sensor data used for human tactile perception is processed by convolutional neural networks or other image processing methods, the input is often required to be a regular rectangular structure. To ensure the integrity of this rectangular structure, existing techniques fill in "0" elements in locations where there are no sensors. For most algorithms, these zero elements used for padding have no effect on the uniformity of all data. If these invalid zero elements could be supplemented with valid information, it would undoubtedly expand the feature dimension of the original input, helping the model better distinguish different categories of data from another perspective. Previous research has proposed solutions to the problem of missing original image regions, specifically by filling the missing input regions with salient regions of other image categories according to certain coefficients and then fusing them according to certain weighted coefficients. However, the pixel level of tactile images is too low to directly transfer pixels from other images for padding. This embodiment proposes a new data padding strategy.

[0050] Based on the different neighborhood ranges of the pixels to be filled, this embodiment proposes two filling methods. The first is rectangular filling, which mainly relies on the effective information around the blank sensing pixels to fill; the second is cross filling, which mainly relies on the information of the corresponding row and column of the sensing unit to fill. The two filling methods will be described in detail below.

[0051] 1) Rectangular Fill

[0052] When filling pressure values ​​according to the hand region, there will be many blank areas in the upper part of the thumb and the gaps between the other four fingers. These areas can first be filled using methods similar to convolution kernels. Regarding the specific principles of rectangular filling, see... Figure 3As shown in sub-figure (a), all dashed grid parts represent a filling receptive field. The rectangular filling determines the receptive field of the rectangle with the pixel to be filled as the center. All valid pixel values ​​in the receptive field are weighted and summed and then nonlinearly projected. The projection result is used as the pixel value of the center pixel. When a filling is completed, the center pixel will be regarded as a valid pixel. The pixel kernel is transferred to the next pixel to continue the next filling. This process is repeated in the pattern of formula (2) until all blank pixels are filled.

[0053]

[0054] Where k is the preset receptive field size, with a default value of 7; valid represents the number of valid pixels within the receptive field; (i,j) represents the relative coordinates of a pixel within the receptive field; X(i,j) represents the pixel value of the center pixel obtained during filling; F(i,j) represents the pixel value corresponding to pixel (i,j) in the receptive field; M(i,j) indicates whether pixel (i,j) is a valid pixel, if it is valid, the pixel is "1", otherwise it is "0"; F represents the matrix of pixels within the receptive field; M represents the discrimination matrix for each pixel within the receptive field, i.e., the valid mask.

[0055] By combining this filling method, this embodiment can capture overall information from different regions within the receptive field, thereby improving the input dimension and robustness of the network.

[0056] 2) Cross fill

[0057] The principle of cross-fill is as follows: Figure 3 As shown in sub-figure (b), the receptive field is determined with the filled pixel as the center. The weighted sum of all valid pixel values ​​within the receptive field is then nonlinearly projected, and the projection result is used as the pixel value of the center pixel. Compared with rectangular filling, the receptive field of the center pixel in cross filling is no longer limited to a rectangle, but extends to all pixels in the corresponding rows and columns of the entire tactile graphics data, as shown in formula (3):

[0058]

[0059] Where X(i,j) represents the pixel value of the center pixel obtained during filling. The pixel neighborhood of this filling method is all pixels in the i-th row and j-th column, so a and b in the formula will traverse from 0 to 32; X(i,a) represents the pixel in the i-th row and a-th column within the receptive field, and X(b,j) represents the pixel in the b-th row and j-th column within the receptive field. After traversal, the information of the corresponding column and row of the filled unit will be mapped to the blank pixel.

[0060] However, simply summing the pixels within the receptive field does not change the minimum linearly independent set of the original input vector. Therefore, this embodiment performs a non-linear projection on the summation result after summation, projecting the summation value between 0 and 1 based on a sigmoid function. Through this data filling method, a blank element in the finger gap region can capture information from the corresponding regions of all fingers except the thumb, thus fully utilizing broader and more global data association information and significantly enhancing the effectiveness of the filling. Compared to rectangular filling, cross-shaped filling has a wider receptive field and can extract more global details.

[0061] In this embodiment, the filling starts from the blank elements in the center of the hand (such as the webbed sides of the fingers and the web between the thumb and forefinger) and moves outwards to fill the blank elements. After each pixel is filled, it is also used as a valid value to fill the next pixel. After all the tactile graphic data is filled, the input feature space dimension of the network will be improved to a certain extent, thereby increasing the amount of data information and achieving the effect of data augmentation.

[0062] S3. Perform random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data.

[0063] When a limited number of sensory pixels are used to perceive similar objects, the sensor images projected from the specific contact gesture surface may not show significant differences due to the similarity in object shape, weight, and other surface physical properties. If the corresponding tactile frames are directly used for classification, the model will struggle to identify the correct results, leading to confusion between similar objects caused by physical limitations. However, tactile perception often exhibits a clear temporal variation; different surface materials (such as roughness and softness) affect tactile interactions, resulting in individual differences in the corresponding perception process. If the contextual information within a perception process can be complemented, the differences in tactile perception frames for similar objects can be amplified, the inter-class distance can be improved, and the model can better distinguish objects. Based on this, this embodiment proposes random transfer processing to help improve the variability and randomness of tactile graphic data; the random transfer processing process is manifested as follows: Figure 4 As shown, the tactile graphic data is enhanced according to formulas (4) and (5).

[0064] x′ a =U⊙x a +V⊙x b (4)

[0065]

[0066] Assume x a and x bIt is a tactile graphic data sample input after being filled with the same type of data; U and V are defined to represent the corresponding binary masks, that is, to extract x. a x b For the two inputs, the positions to be deleted (weight reset to 0), retained (weight reset to 1), or weighted and combined (weights reset to p and 1-p) are given, where U∈{1,⊙} 32×32 V∈{0,1-p} 32×32 p as x a x b The coefficients for fusing the two input shearing portions are randomly sampled from the beta distribution; ⊙ represents element-wise multiplication, and ||·|| represents selecting the shape of 1 in the matrix. It indicates that the shapes are exactly the same.

[0067] This embodiment employs the concept of convex combination instead of directly splicing the cut-off portions of two tactile graphic data samples to prevent the tactile graphic data from rapidly degrading and deviating from the data manifold. This embodiment assumes the input x... a As an enhanced sample object, and from the input x a Extract a rectangle from the input x, and combine the original data within it with the data from the input x of the same type. b The data portions of the same shape randomly selected from the input x are combined in a convex manner, and then combined with the randomly selected input x of the same type. b Extract shapes of the same size from the input x and incorporate them into the input x. a Go to the middle.

[0068] The randomness in random transfer learning is mainly reflected in two aspects: first, in the original input x... a The size of the selected portion is random; secondly, although the cut-out parts maintain a consistent shape, the absolute positions of the selected areas between the two inputs are not the same. This embodiment achieves the offset effect by a small segment translation of the starting point. When input data of different parts of the same category are convexly combined, from input x... b The time-frequency information obtained from an image is randomly shifted to the input x. a In another image, this results in a frequency shift. This is a characteristic of the MSDA strategy, which involves sharing time-frequency features within the same class. As data of the same class become closer together, the inter-class distance is increased. Furthermore, when the dataset is expanded in this way, the diversity of the data is increased, thereby increasing the model's prediction variance for certain abstract transformations. This also helps the model control its own complexity, reducing its dependence on the training set data while generalizing to a wider range of tactile data.

[0069] This embodiment also provides an enhancement system for irregular tactile graphic data, including the following modules:

[0070] The preprocessing module is used to eliminate outlier data in the tactile graphic data and to standardize the tactile graphic data.

[0071] The data filling module is used to fill haptic graphic data with rectangular or cross-shaped fill.

[0072] Rectangular filling uses the pixel to be filled as the center to determine the receptive field of the rectangle. All valid pixel values ​​within the receptive field are weighted and summed, and then non-linearly projected. The projection result is used as the pixel value of the center pixel. After one filling is completed, the center pixel is regarded as a valid pixel, and the pixel kernel is moved to the next pixel to continue the next filling.

[0073] Cross-filling determines the receptive field centered on the pixel to be filled. The receptive field is all pixels in the corresponding row and column of the entire tactile graphics data. The weighted sum of all valid pixel values ​​within the receptive field is then non-linearly projected, and the projection result is used as the pixel value of the center pixel.

[0074] The random migration module performs random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data.

[0075] In this embodiment, each module in the enhanced system is used to implement steps S1-S3, and the detailed implementation process is described in the aforementioned enhancement method.

[0076] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. An enhancement method for irregular tactile graphic data, characterized in that, Includes the following steps: S1. Eliminate outlier data in the tactile graphic data and standardize the tactile graphic data. S2. Use rectangular fill or cross fill to fill the tactile graphic data; The rectangular filling method determines the receptive field of the rectangle with the pixel to be filled as the center. After weighted summation of all valid pixel values ​​within the receptive field, a non-linear projection is performed, and the projection result is used as the pixel value of the center pixel. Once a fill cycle is complete, the center pixel will be considered a valid pixel, and the pixel kernel will move to the next pixel to continue the next fill cycle. Cross-filling determines the receptive field centered on the pixel to be filled. The receptive field is all pixels in the corresponding row and column of the entire tactile graphics data. The weighted sum of all valid pixel values ​​within the receptive field is then non-linearly projected, and the projection result is used as the pixel value of the center pixel. S3. Perform random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data.

2. The enhancement method for irregular tactile graphic data according to claim 1, characterized in that, In step S2, the pixel value of the center pixel in the rectangle filling is calculated using the following formula: Where k is the preset receptive field size; valid represents the number of valid pixels within the receptive field. (i,j) represents the relative coordinates of a pixel within the receptive field; X(i,j) represents the pixel value of the center pixel obtained during filling; F(i,j) represents the pixel value corresponding to pixel (i,j) in the receptive field; M(i,j) indicates whether pixel (i,j) is a valid pixel, if valid, the pixel is "1", otherwise it is "0"; F represents the matrix of pixels within the receptive field; N represents the discrimination matrix for whether each pixel within the receptive field is a valid pixel; the sigmoid function represents projecting the summation value between 0 and 1.

3. The enhancement method for irregular tactile graphic data according to claim 1, characterized in that, In step S2, the pixel value of the center pixel in the cross-fill is calculated using the following formula: Where X(i,j) represents the pixel value of the center pixel obtained during filling; X(i,a) represents the pixel in the i-th row and a-th column within the receptive field, and X(b,j) represents the pixel in the b-th row and j-th column within the receptive field; the sigmoid function represents projecting the summation value between 0 and 1.

4. The enhancement method for irregular tactile graphic data according to claim 1, characterized in that, In step S3, the random migration processing uses the idea of ​​convex combination to splice the cut-off parts of the two tactile graphic data sample inputs. The first tactile graphic data sample input is used as the enhanced sample object, and a rectangle is cut out from the first tactile graphic data sample input. The original data in the rectangle and the data part of the same shape randomly selected from the second tactile graphic data sample input after the same type of data filling processing are convexly combined. The same size shape is cut out from the randomly selected second tactile graphic data sample input and incorporated into the first tactile graphic data sample input.

5. The enhancement method for irregular tactile graphic data according to claim 4, characterized in that, The random migration process in step S3 is represented as follows: x′ a =U⊙x a +V⊙x b Where x a and x b Input tactile graphic data samples after they have undergone the same type of data filling processing; define U and V to represent the corresponding binary masks, that is, to extract x. a x b The positions where U∈{1,p} are deleted, retained, or weighted and merged from two haptic graphic data sample inputs. 32×32 V∈{0,1-p} 32×32 p as x a x b The coefficients for fusing two tactile graphic data samples at the shearing site; ⊙ represents element-wise multiplication, ||·|| represents selecting the shape with 1 in the matrix. It indicates that the shapes are exactly the same.

6. The enhancement method for irregular tactile graphic data according to claim 1, characterized in that, The standardization process in step S1 is as follows: Where x norm For standardized data, x represents haptic graphic data; in each collection round, x ave x max and x min These represent the average, maximum, and minimum values ​​of the subset of tactile graphic data in this collection round, respectively. The Clip(·) function is used to restrict the value inside the parentheses to the range [0,1].

7. An enhancement system for irregular tactile graphic data, characterized in that, Includes the following modules: The preprocessing module is used to eliminate outlier data in the tactile graphic data and to standardize the tactile graphic data. The data filling module is used to fill haptic graphic data with rectangular or cross-shaped fill. The rectangular filling method determines the receptive field of the rectangle with the pixel to be filled as the center. After weighted summation of all valid pixel values ​​within the receptive field, a non-linear projection is performed, and the projection result is used as the pixel value of the center pixel. Once a fill cycle is complete, the center pixel will be considered a valid pixel, and the pixel kernel will move to the next pixel to continue the next fill cycle. Cross-filling determines the receptive field centered on the pixel to be filled. The receptive field is all pixels in the corresponding row and column of the entire tactile graphics data. The weighted sum of all valid pixel values ​​within the receptive field is then non-linearly projected, and the projection result is used as the pixel value of the center pixel. The random migration module performs random migration processing on the filled tactile graphic data to obtain enhanced tactile graphic data.

8. The enhancement system for irregular tactile graphic data according to claim 7, characterized in that, In the data filling module, the formula for calculating the pixel value of the center pixel for rectangular filling is: Where k is the preset receptive field size; valid represents the number of valid pixels within the receptive field. (i,j) represents the relative coordinates of a pixel within the receptive field; X(i,j) represents the pixel value of the center pixel obtained during filling; F(i,j) represents the pixel value corresponding to pixel (i,j) in the receptive field; M(i,j) indicates whether pixel (i,j) is a valid pixel, if it is valid, the pixel is "1", otherwise it is "0"; F represents the matrix of pixels within the receptive field; M represents the discrimination matrix for each pixel within the receptive field to determine whether it is a valid pixel; the sigmoid function represents projecting the summation value between 0 and 1.

9. The enhancement system for irregular tactile graphic data according to claim 7, characterized in that, In the data fill module, the calculation formula for the pixel value of the center pixel in cross fill is: Where X(i,j) represents the pixel value of the center pixel obtained during filling; X(i,a) represents the pixel in the i-th row and a-th column within the receptive field, and X(b,j) represents the pixel in the b-th row and j-th column within the receptive field; the sigmoid function represents projecting the summation value between 0 and 1.

10. The enhancement system for irregular tactile graphic data according to claim 7, characterized in that, In the random migration module, the idea of ​​convex combination is used to splice the cut-off parts of two tactile graphic data sample inputs. The first tactile graphic data sample input is used as the enhanced sample object, and a rectangle is cut out from the first tactile graphic data sample input. The original data in the rectangle and the data part of the same shape randomly selected from the second tactile graphic data sample input after being filled with the same type of data are convexly combined. The same size shape is cut out from the randomly selected second tactile graphic data sample input and incorporated into the first tactile graphic data sample input.

Images