Neural network application system for image data processing

Through the neural network application system, combined with the convolutional neural network model, dynamic selection of smoothing algorithms has been solved, and the problem of poor processing of different picture content in the existing technology has been achieved, targeted optimization of image smoothing processing, and improved the image of road monitoring Processing effect.

CN119006327AActive Publication Date: 2024-11-22BEIJING LIUSHEN DATA TECH CO LTD
View PDF 12 Cites 0 Cited by

Patent Information

Application Number
CN202411184244.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-01-24
Filing Date
2024-08-27
Publication Date
2024-11-22
Estimated Expiration
2044-08-27

AI Technical Summary

Technical Problem

When the prior art performs image smoothing processing on different screen contents, the same smoothing algorithm cannot meet the needs of different screen contents, resulting in poor processing effects, especially in road surface state monitoring.

Method used

Design a neural network application system to dynamically select the optimal smoothing algorithm to achieve targeted optimization processing, including road surface capture equipment and point image Recovery equipment, signal filtering equipment, data interpolation equipment, on-site analysis mechanism, distribution identification mechanism and intelligent identification mechanism, select target smoothing algorithm based on the maximum signal-to-noise ratio value for processing.

Benefits of technology

It improves the intelligence level of image smoothing processing, ensures the best smoothing effect, is suitable for on-site monitoring of pavement contents of different screens, and improves the targetedness and effect of processing.

✦ Generated by Eureka AI based on patent content.
Patent Text Reader

Abstract

The invention relates to a neural network application system for image data processing, and the system comprises a road surface capturing device which employs a high-angle shooting mode to achieve the real-time image signal capturing of a set road surface region, so as to obtain a real-time road surface image; and the matching processing mechanism is used for acquiring respective corresponding signal-to-noise ratios based on the algorithm numerical values corresponding to the various smoothing algorithms, and the smoothing algorithm corresponding to the maximum signal-to-noise ratio is used as a target smoothing algorithm of the nearest neighbor interpolation image so as to perform smoothing processing on the nearest neighbor interpolation image. The neural network application system for image data processing is stable in operation and simple and convenient to operate. As the respective corresponding signal-to-noise ratios can be obtained based on the algorithm values corresponding to various smoothing algorithms for the optimized picture of the high-angle shot image in the same path area, the smoothing algorithm corresponding to the maximum value of the signal-to-noise ratio is used as the target smoothing algorithm of the optimized picture so as to perform smoothing processing on the optimized picture; therefore, the intelligent level of smoothing processing is improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The 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 fact that some areas with large brightness changes or some bright spots (also called noise) will appear on the remote sensing image due to the influence of factors such as sensors and atmosphere. This processing method to suppress noise and make the image brightness tend to be flat is image smoothing. Image smoothing is actually low-pass filtering, and the smoothing process will cause the image edges to become blurred. Image smoothing methods include: interpolation method, linear smoothing method, convolution method, etc. This processing method is smoothed according to the different image noises. For example, salt and pepper noise uses linear smoothing method. Patent application CN117078530A filed by Fujian Institute of Research on the Structure of Matter, Chinese Academy of Sciences, provides an image smoothing method, device, electronic device and storage medium, which converts the initial smoothed image obtained by smoothing the image to be smoothed into a binary image, calculates the connected areas in the binary image and determines the parameter information of the connected areas, and determines the initial isolated noise in the binary image; if the pixel value of the foreground area of ​​the initial smoothed image is higher than the pixel value of the background area, the white pixel in the binary image is used as the structural pixel, and if there is no structural pixel around the initial isolated noise or there is a structural pixel in the neighborhood in only one direction, the initial isolated noise is determined to be the target isolated noise; the binary image is converted into an RGB image, and based on an adaptive window, the target isolated noise is merged with the surrounding area of ​​the target isolated noise to obtain a target smoothed image. The method can accurately determine the target isolated noise that needs to be processed and prevent the misprocessing of non-isolated noise.

[0003] In the prior art, in order to continuously monitor the road 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 different frame contents. Summary of the invention

[0004] Different from the prior art, 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 ratios of various smoothing algorithms based on the algorithm numerical values ​​corresponding to each, 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 surface 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] A neural network application system for image data processing proposed by the present invention comprises: A road surface capturing 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, which is arranged at the road management server and connected to the road surface capture device, is used to perform single or multiple point image restoration processing 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, for performing 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, which is arranged at the road management server and connected to the data interpolation device, is used to detect 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 output the hue component value, brightness component value and saturation component value of each pixel in the nearest neighbor interpolation image as the picture content data of the nearest neighbor interpolation image; a distribution identification mechanism, used for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, wherein 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; 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 the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm, wherein the convolutional neural network model is a convolutional neural network after multiple learning; A matching processing mechanism, connected to the intelligent identification mechanism, for obtaining the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and using the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm of the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image to obtain a corresponding matching processing image; Among them, various smoothing algorithms include a non-scaling transformation fuzzy algorithm, a median fuzzy algorithm, a Gaussian fuzzy algorithm, and a bilateral smoothing fuzzy algorithm. DETAILED DESCRIPTION

[0006] The following is a detailed description of an embodiment of the neural network application system for image data processing of the present invention.

[0007] First embodiment

[0008] According to a first embodiment of the present invention, a neural network application system for image data processing is shown, the system comprising: A road surface capturing 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; For example, a road surface capture device is located above a set road surface area and is used to capture a real-time image signal of the set road surface area in a bird's-eye view mode to obtain and output a real-time road surface image, including: the road surface capture device includes a photoelectric sensor and a bird's-eye view imaging lens, which cooperate to capture a real-time image signal 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, which is arranged at the road management server and connected to the road surface capture device, is used to perform single or multiple point image restoration processing 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, for performing 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, which is arranged at the road management server and connected to the data interpolation device, is used to detect 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 output the hue component value, brightness component value and saturation component value of each pixel in the nearest neighbor interpolation image as the picture content data of the nearest neighbor interpolation image; a distribution identification mechanism, used for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, wherein 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; 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 the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm, wherein the convolutional neural network model is a convolutional neural network after multiple learning; A matching processing mechanism, connected to the intelligent identification mechanism, for obtaining the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and using the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm of the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image to obtain a corresponding matching processing image; Among them, various smoothing algorithms include non-scaling transformation fuzzy algorithm, median fuzzy algorithm, Gaussian fuzzy algorithm and bilateral smoothing fuzzy algorithm; 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. The road surface capture device 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.

[0009] Second embodiment

[0010] A second embodiment of the present invention shows a neural network application system for image data processing.

[0011] The neural network application system for image data processing shown in the second embodiment of the present invention may include: A road surface capturing 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, which is arranged at the road management server and connected to the road surface capture device, is used to perform single or multiple point image restoration processing 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, for performing 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, which is arranged at the road management server and connected to the data interpolation device, is used to detect 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 output the hue component value, brightness component value and saturation component value of each pixel in the nearest neighbor interpolation image as the picture content data of the nearest neighbor interpolation image; a distribution identification mechanism, used for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, wherein 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; 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 the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm, wherein the convolutional neural network model is a convolutional neural network after multiple learning; A matching processing mechanism, connected to the intelligent identification mechanism, for obtaining the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and using the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm of the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image to obtain a corresponding matching processing image; 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 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, including: the plurality of pressure detection units respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism and the distribution identification mechanism are a plurality of contact pressure sensors; 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: the internal structures of the 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 are the same; Wherein, 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 non-contact pressure sensors respectively used by the point image restoration device, the signal filtering device, the data interpolation device, the field analysis mechanism and the distribution identification mechanism have the same upper pressure measurement threshold and lower pressure measurement threshold; 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.

[0012] Third embodiment

[0013] A third embodiment of the present invention shows a neural network application system for image data processing.

[0014] The neural network application system for image data processing shown in the third embodiment of the present invention may include: A road surface capturing 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, which is arranged at the road management server and connected to the road surface capture device, is used to perform single or multiple point image restoration processing 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, for performing 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, which is arranged at the road management server and connected to the data interpolation device, is used to detect 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 output the hue component value, brightness component value and saturation component value of each pixel in the nearest neighbor interpolation image as the picture content data of the nearest neighbor interpolation image; a distribution identification mechanism, used for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, wherein 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; 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 the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm, wherein the convolutional neural network model is a convolutional neural network after multiple learning; A matching processing mechanism, connected to the intelligent identification mechanism, for obtaining the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and using the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm of the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image to obtain a corresponding matching processing image; 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 field 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 field analysis mechanism and the distribution identification mechanism; Wherein, the real-time warning mechanism is respectively connected with a plurality of contact pressure sensors respectively 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 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 field analysis mechanism and the distribution identification mechanism, including: the real-time warning mechanism includes a timing analysis unit and a warning operation unit; 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 field 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 field analysis mechanism and the distribution identification mechanism, and also includes: within the real-time warning mechanism, the timing analysis unit is connected to the warning operation unit.

[0015] 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 action for 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.

[0016] It can be seen that the present invention mainly has the following four significant technical effects: Firstly: 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 the picture content data of the nearest neighbor interpolation image; Secondly: detecting noise data corresponding to each type of noise in each type of noise appearing in the nearest neighbor interpolation image, wherein 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; Again: 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; Finally: based on the algorithm values ​​corresponding to various smoothing algorithms, the corresponding signal-to-noise ratios are obtained respectively, and the smoothing algorithm corresponding to the maximum signal-to-noise ratio is used as the target smoothing algorithm of the nearest neighbor interpolation image after targeted optimization of the road scene picture to perform smoothing processing on the nearest neighbor interpolation image to obtain the corresponding matching processing image, thereby improving the intelligent level of smoothing processing and ensuring the best smoothing processing effect.

[0017] The neural network application system for image data processing of the present invention is stable in operation and easy to operate. Since the corresponding signal-to-noise ratios of the optimized pictures of the overhead images of the same path area can be obtained based on the algorithm values ​​corresponding to various smoothing algorithms, the smoothing algorithm corresponding to the maximum signal-to-noise ratio is used as the target smoothing algorithm of the optimized picture to perform smoothing processing on the optimized picture, thereby improving the intelligent level of smoothing processing.

[0018] 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, It is characterized in that The system comprises: A road surface capturing 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, which is arranged at the road management server and connected to the road surface capture device, is used to perform single or multiple point image restoration processing 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, for performing 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, which is arranged at the road management server and connected to the data interpolation device, is used to detect 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 output the hue component value, brightness component value and saturation component value of each pixel in the nearest neighbor interpolation image as the picture content data of the nearest neighbor interpolation image; a distribution identification mechanism, used for detecting noise data corresponding to each type of noise among various types of noise appearing in the nearest neighbor interpolation image, wherein 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; 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 the signal-to-noise ratio of the image content obtained after the nearest neighbor interpolation image uses the smoothing algorithm, wherein the convolutional neural network model is a convolutional neural network after multiple learning; A matching processing mechanism, connected to the intelligent identification mechanism, for obtaining the signal-to-noise ratios corresponding to the various smoothing algorithms based on the algorithm values ​​corresponding to the various smoothing algorithms, and using the smoothing algorithm corresponding to the maximum signal-to-noise ratio as the target smoothing algorithm of the nearest neighbor interpolation image to perform smoothing processing on the nearest neighbor interpolation image to obtain a corresponding matching processing image; Among them, various smoothing algorithms include a non-scaling transformation fuzzy algorithm, a median fuzzy algorithm, a Gaussian fuzzy algorithm, and a bilateral smoothing fuzzy algorithm.

2. The neural network application system for image data processing according to claim 1, Features: 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. The road surface capture device has a built-in bird's-eye view lens and an image sensor, 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.

3. The neural network application system for image data processing according to claim 2, It is characterized in that 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; 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, including: the multiple pressure detection units 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 are multiple contact pressure sensors.

4. The neural network application system for image data processing according to claim 3, Features: 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: the internal structures of 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 are the same.

5. The neural network application system for image data processing according to claim 3, Features: 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 non-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 have the same pressure measurement upper limit threshold and pressure measurement lower limit threshold.

6. The neural network application system for image data processing according to claim 5, Features: 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.

7. A neural network application system for image data processing as claimed in any one of claims 3 to 6, It is characterized in that The system further comprises: The real-time warning mechanism is respectively connected to 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, and is used to perform corresponding pressure warning operations based on the pressure measurement results of the 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.

8. The neural network application system for image data processing according to claim 7, Features: The real-time warning mechanism is respectively connected to 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, and is used to perform corresponding pressure warning operations based on the pressure measurement results of the 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. The real-time warning mechanism includes a timing analysis unit and a warning operation unit.

9. The neural network application system for image data processing according to claim 8, Features: The real-time warning mechanism is respectively connected to 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, and is used to perform corresponding pressure warning operations based on the pressure measurement results of the 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. It also includes: within the real-time warning mechanism, the timing analysis unit is connected to the warning operation unit.

Citation Information

Patent Citations

  • Image smoothing method and device, electronic equipment and storage medium

    CN117078530A

  • Remote intelligent monitoring method and system

    CN108012113A

  • RAW domain image denoising method, computer device and computer readable storage medium

    CN111784603A

  • Intelligent blind restoration method for low-signal-to-noise-ratio fuzzy star map under high dynamic condition

    CN117237203A

  • Image data intelligent optimization transmission system

    CN117354467A