Neural network application system for image data processing

Abstract

The present invention relates to a neural network application system for image data processing, the system comprising: a road surface capture device, which uses an overhead mode to realize a real-time image signal capture action for a set road surface area to obtain a real-time road surface image; a matching processing mechanism, which obtains the corresponding signal-to-noise ratio based on the algorithm values ​​corresponding to various smoothing algorithms, and uses the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image. The neural network application system for image data processing of the present invention is stable in operation and easy to operate. Since it is possible to obtain the corresponding signal-to-noise ratio based on the algorithm values ​​corresponding to various smoothing algorithms for the optimized picture of the overhead image of the same path area, and use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the optimized picture to perform smoothing processing on the optimized picture, the intelligent level of smoothing processing is improved.

Description

Technical Field

[0001] The present invention relates to the field of smoothing processing, and in particular to a neural network application system for image data processing. Background Art

[0002] Image smoothing refers to the process of suppressing noise and smoothing image brightness, which can occur in remote sensing images due to factors such as the sensor and the atmosphere. This process involves applying low-pass filtering, which can blur image edges. Image smoothing methods include interpolation, linear smoothing, and convolution. These methods are tailored to the specific type of image noise; for example, linear smoothing is used for salt and pepper noise. Patent application CN117078530A, filed by the Fujian Institute of Research on the Structure of Matter, Chinese Academy of Sciences, provides an image smoothing method, apparatus, electronic device, and storage medium. This method converts an initial smoothed image obtained by smoothing an image to be smoothed into a binary image, calculates connected regions in the binary image, determines parameter information for the connected regions, and identifies initial isolated noise in the binary image. If the pixel values ​​in the foreground region of the initial smoothed image are higher than those in the background region, white pixels in the binary image are used as structural pixels. If there are no structural pixels surrounding the initial isolated noise or structural pixels exist only in one directional neighborhood, the initial isolated noise is identified as target isolated noise. The binary image is converted into an RGB image, and the target isolated noise is fused with its surrounding region based on an adaptive window to obtain a target smoothed image. This method accurately identifies the target isolated noise to be processed, preventing misprocessing of non-isolated noise.

[0003] In the prior art, in order to continuously monitor the road surface condition of a set road area, it is necessary to perform continuous image capture of the set road area. Before displaying each captured frame, in order to improve the image quality of each frame, image smoothing processing using the same smoothing algorithm is performed on various frames. Obviously, due to the different contents of various frames, there is an irreconcilable contradiction between the same smoothing algorithm and the different frame contents. Summary of the Invention

[0004] Different from the existing technology, the present invention proposes a neural network application system for image data processing, which can synchronously input the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used into a convolutional neural network model to obtain the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm, obtain the corresponding signal-to-noise ratio based on the algorithm numerical values ​​corresponding to various smoothing algorithms, and use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm of the nearest neighbor interpolation image after targeted optimization of the road scene to perform smoothing processing on the nearest neighbor interpolation image to obtain the corresponding matching processed image, thereby improving the intelligence level of the smoothing processing and ensuring the best smoothing processing effect.

[0005] The present invention proposes a neural network application system for image data processing, comprising:

[0006] A road surface capture device is located above a set road surface area and is used to capture real-time image signals of the set road surface area in a bird's-eye view mode to obtain and output a real-time road surface image;

[0007] a point image restoration device, provided at the road management server and connected to the road surface capture device, for performing single or multiple point image restoration processes on the received real-time road surface image to obtain and output a corresponding signal restoration image;

[0008] a signal filtering device connected to the point image restoration device, configured to perform FRANGI filtering on the received signal restoration image to obtain and output a corresponding instant filtered image;

[0009] a data interpolation device, connected to the signal filtering device, for performing nearest neighbor interpolation processing on the received instant filtered image to obtain and output a corresponding nearest neighbor interpolation image;

[0010] an on-site analysis mechanism, provided at the road management server and connected to the data interpolation device, for detecting the hue component value, brightness component value, and saturation component value of each pixel in the received nearest neighbor interpolation image in the HSL space, and outputting the hue component value, brightness component value, and saturation component value of each pixel in the nearest neighbor interpolation image as picture content data of the nearest neighbor interpolation image;

[0011] a distribution identification mechanism for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, the noise data corresponding to each type of noise being the noise amplitude corresponding to each type of noise and the number of associated pixels, and outputting the noise data of each type of noise in the nearest neighbor interpolation image as distribution information of the nearest neighbor interpolation image;

[0012] an intelligent identification mechanism, connected to the on-site analysis mechanism and the distribution identification mechanism, respectively, for synchronously inputting into a convolutional neural network model, based on the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used, to obtain a signal-to-noise ratio of the image content obtained by applying the smoothing algorithm to the nearest neighbor interpolation image, wherein the convolutional neural network model is a convolutional neural network that has been learned multiple times;

[0013] a matching processing mechanism connected to the intelligent identification mechanism, configured to obtain the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and to use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image, thereby obtaining a corresponding matching processed image;

[0014] Among them, various smoothing algorithms include non-scaling transformation blur algorithm, median blur algorithm, Gaussian blur algorithm and bilateral smoothing blur algorithm. DETAILED DESCRIPTION

[0015] The following is a detailed description of an embodiment of the neural network application system for image data processing of the present invention.

[0016] First embodiment

[0017] According to a first embodiment of the present invention, a neural network application system for image data processing is shown, the system comprising:

[0018] A road surface capture device is located above a set road surface area and is used to capture real-time image signals of the set road surface area in a bird's-eye view mode to obtain and output a real-time road surface image;

[0019] For example, a road surface capture device is located above a predetermined road surface area and is configured to capture real-time image signals of the predetermined road surface area in a top-down shooting mode to obtain and output a real-time road surface image. The road surface capture device includes a photoelectric sensor and a top-down shooting imaging lens, which cooperate to capture real-time image signals of the predetermined road surface area in a top-down shooting mode to obtain and output a real-time road surface image.

[0020] a point image restoration device, provided at the road management server and connected to the road surface capture device, for performing single or multiple point image restoration processes on the received real-time road surface image to obtain and output a corresponding signal restoration image;

[0021] a signal filtering device connected to the point image restoration device, configured to perform FRANGI filtering on the received signal restoration image to obtain and output a corresponding instant filtered image;

[0022] a data interpolation device, connected to the signal filtering device, for performing nearest neighbor interpolation processing on the received instant filtered image to obtain and output a corresponding nearest neighbor interpolation image;

[0023] an on-site analysis mechanism, provided at the road management server and connected to the data interpolation device, for detecting the hue component value, brightness component value, and saturation component value of each pixel in the received nearest neighbor interpolation image in the HSL space, and outputting the hue component value, brightness component value, and saturation component value of each pixel in the nearest neighbor interpolation image as picture content data of the nearest neighbor interpolation image;

[0024] a distribution identification mechanism for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, the noise data corresponding to each type of noise being the noise amplitude corresponding to each type of noise and the number of associated pixels, and outputting the noise data of each type of noise in the nearest neighbor interpolation image as distribution information of the nearest neighbor interpolation image;

[0025] an intelligent identification mechanism, connected to the on-site analysis mechanism and the distribution identification mechanism, respectively, for synchronously inputting into a convolutional neural network model, based on the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used, to obtain a signal-to-noise ratio of the image content obtained by applying the smoothing algorithm to the nearest neighbor interpolation image, wherein the convolutional neural network model is a convolutional neural network that has been learned multiple times;

[0026] a matching processing mechanism connected to the intelligent identification mechanism, configured to obtain the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and to use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image, thereby obtaining a corresponding matching processed image;

[0027] Among them, various smoothing algorithms include non-scaling transformation blur algorithm, median blur algorithm, Gaussian blur algorithm and bilateral smoothing blur algorithm;

[0028] Among them, the road surface capture device is located above the set road surface area, and is used to adopt a bird's-eye view mode to realize the real-time image signal capture action of the set road surface area, so as to obtain and output the real-time road surface image. It includes: the road surface capture device is built with a bird's-eye view lens and an image sensor, and is used to adopt a bird's-eye view mode to realize the real-time image signal capture action of the set road surface area, so as to obtain and output the real-time road surface image.

[0029] Second embodiment

[0030] According to the second embodiment of the present invention, a neural network application system for image data processing is shown.

[0031] According to the second embodiment of the present invention, a neural network application system for image data processing may include:

[0032] A road surface capture device is located above a set road surface area and is used to capture real-time image signals of the set road surface area in a bird's-eye view mode to obtain and output a real-time road surface image;

[0033] a point image restoration device, provided at the road management server and connected to the road surface capture device, for performing single or multiple point image restoration processes on the received real-time road surface image to obtain and output a corresponding signal restoration image;

[0034] a signal filtering device connected to the point image restoration device, configured to perform FRANGI filtering on the received signal restoration image to obtain and output a corresponding instant filtered image;

[0035] a data interpolation device, connected to the signal filtering device, for performing nearest neighbor interpolation processing on the received instant filtered image to obtain and output a corresponding nearest neighbor interpolation image;

[0036] an on-site analysis mechanism, provided at the road management server and connected to the data interpolation device, for detecting the hue component value, brightness component value, and saturation component value of each pixel in the received nearest neighbor interpolation image in the HSL space, and outputting the hue component value, brightness component value, and saturation component value of each pixel in the nearest neighbor interpolation image as picture content data of the nearest neighbor interpolation image;

[0037] a distribution identification mechanism for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, the noise data corresponding to each type of noise being the noise amplitude corresponding to each type of noise and the number of associated pixels, and outputting the noise data of each type of noise in the nearest neighbor interpolation image as distribution information of the nearest neighbor interpolation image;

[0038] an intelligent identification mechanism, connected to the on-site analysis mechanism and the distribution identification mechanism, respectively, for synchronously inputting into a convolutional neural network model, based on the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used, to obtain a signal-to-noise ratio of the image content obtained by applying the smoothing algorithm to the nearest neighbor interpolation image, wherein the convolutional neural network model is a convolutional neural network that has been learned multiple times;

[0039] a matching processing mechanism connected to the intelligent identification mechanism, configured to obtain the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and to use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image, thereby obtaining a corresponding matching processed image;

[0040] a pressure detection mechanism, comprising a plurality of pressure detection units, for respectively detecting current surface pressure values ​​of the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism, and the distribution identification mechanism;

[0041] The pressure detection mechanism includes a plurality of pressure detection units for respectively detecting current surface pressure values ​​of the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism, wherein the plurality of pressure detection units respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism are a plurality of contact pressure sensors;

[0042] The pressure detection mechanism includes a plurality of pressure detection units for respectively detecting the current surface pressure values ​​of the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism, and the distribution identification mechanism, and further includes: a plurality of contact pressure sensors respectively used for the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism, and the distribution identification mechanism having the same internal structure;

[0043] The pressure detection mechanism includes a plurality of pressure detection units for respectively detecting the current surface pressure values ​​of the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism, and further includes: a plurality of non-contact pressure sensors respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism have the same upper and lower pressure measurement thresholds;

[0044] Among them, the pressure detection mechanism includes multiple pressure detection units, which are used to respectively detect the current surface pressure values ​​of the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism and the distribution identification mechanism. It also includes: multiple contact pressure sensors respectively used for the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism and the distribution identification mechanism all perform corresponding current surface pressure value detection based on the difference between the air pressure and the internal pressure of the detected device.

[0045] Third embodiment

[0046] According to the third embodiment of the present invention, a neural network application system for image data processing is shown.

[0047] According to the third embodiment of the present invention, a neural network application system for image data processing may include:

[0048] A road surface capture device is located above a set road surface area and is used to capture real-time image signals of the set road surface area in a bird's-eye view mode to obtain and output a real-time road surface image;

[0049] a point image restoration device, provided at the road management server and connected to the road surface capture device, for performing single or multiple point image restoration processes on the received real-time road surface image to obtain and output a corresponding signal restoration image;

[0050] a signal filtering device connected to the point image restoration device, configured to perform FRANGI filtering on the received signal restoration image to obtain and output a corresponding instant filtered image;

[0051] a data interpolation device, connected to the signal filtering device, for performing nearest neighbor interpolation processing on the received instant filtered image to obtain and output a corresponding nearest neighbor interpolation image;

[0052] an on-site analysis mechanism, provided at the road management server and connected to the data interpolation device, for detecting the hue component value, brightness component value, and saturation component value of each pixel in the received nearest neighbor interpolation image in the HSL space, and outputting the hue component value, brightness component value, and saturation component value of each pixel in the nearest neighbor interpolation image as picture content data of the nearest neighbor interpolation image;

[0053] a distribution identification mechanism for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, the noise data corresponding to each type of noise being the noise amplitude corresponding to each type of noise and the number of associated pixels, and outputting the noise data of each type of noise in the nearest neighbor interpolation image as distribution information of the nearest neighbor interpolation image;

[0054] an intelligent identification mechanism, connected to the on-site analysis mechanism and the distribution identification mechanism, respectively, for synchronously inputting into a convolutional neural network model, based on the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used, to obtain a signal-to-noise ratio of the image content obtained by applying the smoothing algorithm to the nearest neighbor interpolation image, wherein the convolutional neural network model is a convolutional neural network that has been learned multiple times;

[0055] a matching processing mechanism connected to the intelligent identification mechanism, configured to obtain the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and to use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image, thereby obtaining a corresponding matching processed image;

[0056] a real-time warning mechanism, connected to the plurality of contact pressure sensors respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism, and configured to perform corresponding pressure warning operations based on pressure measurement results of the plurality of contact pressure sensors respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism;

[0057] The real-time warning mechanism is connected to the plurality of contact pressure sensors respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism, and is used to perform corresponding pressure warning operations based on the pressure measurement results of the plurality of contact pressure sensors respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism, and the distribution identification mechanism. The real-time warning mechanism includes a timing analysis unit and a warning operation unit.

[0058] Among them, the real-time warning mechanism is respectively connected to multiple contact pressure sensors used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism and the distribution identification mechanism, and is used to perform corresponding pressure warning operations based on the pressure measurement results of the multiple contact pressure sensors used by the point image restoration device, the signal filtering device, the data interpolation device, the on-site analysis mechanism and the distribution identification mechanism. It also includes: within the real-time warning mechanism, the timing analysis unit is connected to the warning operation unit.

[0059] In addition, in the neural network application system for image data processing, the road surface capture device is located above the set road surface area and is used to adopt a bird's-eye view mode to realize real-time image signal capture of the set road surface area to obtain and output a real-time road surface image. It also includes: the image sensor is a CMOS sensor or a CCD sensor, and the resolution of the real-time road surface image is proportional to the width of the set road surface area.

[0060] It can be seen that the present invention mainly has the following four significant technical effects:

[0061] First, the hue component value, brightness component value, and saturation component value of each pixel in the received nearest neighbor interpolation image in the HSL space are detected, and the hue component value, brightness component value, and saturation component value of each pixel in the nearest neighbor interpolation image are output as picture content data of the nearest neighbor interpolation image;

[0062] Secondly: detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, where the noise data corresponding to each type of noise is the noise amplitude corresponding to each type of noise and the number of associated pixels, and outputting the noise data of each type of noise in the nearest neighbor interpolation image as distribution information of the nearest neighbor interpolation image;

[0063] Thirdly: based on the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used, the data are synchronously input into the convolutional neural network model to obtain the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm. The convolutional neural network model is a convolutional neural network after multiple learning, thereby providing key information for the subsequent selection of the smoothing algorithm;

[0064] Finally, the corresponding signal-to-noise ratios of various smoothing algorithms are obtained based on their corresponding algorithm values. The smoothing algorithm corresponding to the maximum signal-to-noise ratio is used as the target smoothing algorithm for the nearest neighbor interpolation image after targeted optimization of the road scene image to perform smoothing processing on the nearest neighbor interpolation image and obtain the corresponding matching processing image, thereby improving the intelligence level of the smoothing processing and ensuring the best smoothing effect.

[0065] The neural network application system for image data processing of the present invention is stable and easy to operate. It can obtain the corresponding signal-to-noise ratios of various smoothing algorithms based on their corresponding algorithm values ​​for optimized images of aerial images of the same path area. The smoothing algorithm corresponding to the maximum signal-to-noise ratio is used as the target smoothing algorithm for performing smoothing processing on the optimized image, thereby improving the intelligent level of smoothing processing.

[0066] It should be understood that the embodiments and examples disclosed herein are exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than by the preceding description, and the equivalent concepts of the meaning of the claims and all changes within the meaning are intended to be included by the claims.

Claims

1. A neural network application system for image data processing, characterized in that: The system includes: A road surface capture device is located above a set road surface area and is used to capture real-time image signals of the set road surface area in a bird's-eye view mode to obtain and output a real-time road surface image; a point image restoration device, provided at the road management server and connected to the road surface capture device, for performing single or multiple point image restoration processes on the received real-time road surface image to obtain and output a corresponding signal restoration image; a signal filtering device connected to the point image restoration device, configured to perform FRANGI filtering on the received signal restoration image to obtain and output a corresponding instant filtered image; a data interpolation device, connected to the signal filtering device, for performing nearest neighbor interpolation processing on the received instant filtered image to obtain and output a corresponding nearest neighbor interpolation image; an on-site analysis mechanism, provided at the road management server and connected to the data interpolation device, for detecting the hue component value, brightness component value, and saturation component value of each pixel in the received nearest neighbor interpolation image in the HSL space, and outputting the hue component value, brightness component value, and saturation component value of each pixel in the nearest neighbor interpolation image as picture content data of the nearest neighbor interpolation image; a distribution identification mechanism for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, the noise data corresponding to each type of noise being the noise amplitude corresponding to each type of noise and the number of associated pixels, and outputting the noise data of each type of noise in the nearest neighbor interpolation image as distribution information of the nearest neighbor interpolation image; an intelligent identification mechanism, connected to the on-site analysis mechanism and the distribution identification mechanism, respectively, for synchronously inputting into a convolutional neural network model, based on the picture content data of the nearest neighbor interpolation image, the distribution information of the nearest neighbor interpolation image, and the algorithm numerical representation corresponding to the smoothing algorithm to be used, to obtain a signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image is smoothed using the algorithm, wherein the convolutional neural network model is a convolutional neural network that has been learned multiple times; a matching processing mechanism connected to the intelligent identification mechanism, configured to obtain the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and to use the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm for the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image, thereby obtaining a corresponding matching processing image; Among them, various smoothing algorithms include non-scaling transformation blur algorithm, median blur algorithm, Gaussian blur algorithm and bilateral smoothing blur algorithm; The road surface capture device is equipped with a built-in overhead camera and an image sensor, which is used to capture real-time image signals of a set road surface area in an overhead mode to obtain and output real-time road surface images. The image sensor is a CMOS sensor or a CCD sensor, and the resolution of the real-time road surface image is proportional to the width of the set road surface area.

2. The neural network application system for image data processing according to claim 1, wherein: The system further comprises: a pressure detection mechanism, comprising a plurality of pressure detection units, for respectively detecting current surface pressure values ​​of the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism, and the distribution identification mechanism; The plurality of pressure detection units are a plurality of contact pressure sensors.

3. The neural network application system for image data processing according to claim 2, wherein: The plurality of contact pressure sensors have the same internal structure.

4. The neural network application system for image data processing according to claim 2, wherein: The plurality of contact pressure sensors have the same pressure measurement upper limit threshold and pressure measurement lower limit threshold.

5. The neural network application system for image data processing according to claim 4, wherein: The plurality of contact pressure sensors all detect corresponding current surface pressure values ​​based on the difference between the air pressure and the internal pressure of the detected device.

6. The neural network application system for image data processing according to any one of claims 2 to 5, characterized in that: The system further comprises: The real-time warning mechanism is respectively connected to multiple contact pressure sensors used by the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism and the distribution identification mechanism, and is used to perform corresponding pressure warning operations for the pressure measurement results of the multiple contact pressure sensors.

7. The neural network application system for image data processing according to claim 6, wherein: The real-time warning mechanism includes a timing analysis unit and a warning operation unit.

8. The neural network application system for image data processing according to claim 7, wherein: In the real-time warning mechanism, the timing analysis unit is connected to the warning operation unit.