Virtual reality VR image recognition method and system

Abstract

The present invention discloses a virtual reality (VR) image recognition method and system, comprising: in a scenario where a user accesses a virtual reality (VR) device, obtaining a virtual reality image based on a user's perspective through a virtual camera, marking it as an image to be recognized, obtaining pixel data of the image, performing Gaussian blur processing on the image to be recognized through a Gaussian blur algorithm to obtain a Gaussian blurred image of the image to be recognized, performing comparative analysis on the two based on pixel values ​​to obtain a blur representation value of the image to be recognized, judging the blur situation of the image to be recognized, marking the image to be recognized as a blurred image or a normal image based on the marked blurred image, obtaining the number of blurred images in the image to be recognized and the simulated motion period in the virtual camera shooting, analyzing to obtain a device impact value, and then analyzing and judging the cause of the blurred image, and importing the obtained normal image into a trained convolutional neural network (CNN) to complete the required image recognition task.

Description

Technical Field

[0001] The present invention relates to the field of virtual reality technology, and in particular to a virtual reality (VR) image recognition method and system. Background Art

[0002] With the continuous advancement of virtual reality (VR) technology, VR devices have been widely used in entertainment, education, medical treatment, industrial design and other fields. As an important part of VR technology, VR image recognition is of great significance for improving user experience and achieving accurate interaction. However, in the current VR image recognition technology, there are still some problems that need to be solved.

[0003] First, when VR devices acquire virtual reality images from the user's perspective, the acquired images may be blurred due to the influence of device performance and user motion factors. This blurring phenomenon not only affects the clarity of the image, but may also interfere with subsequent image recognition tasks and reduce recognition accuracy. Secondly, for the analysis of the causes of blurred images, current technologies often lack systematic methods. In VR systems, the causes of blurred images include the hardware and software devices of the virtual reality system and user actions. However, existing technologies can often only perform simple classification of blurred images, but cannot deeply explore their causes, thus failing to provide strong support for subsequent optimization and improvement to improve image recognition accuracy, generalization ability and recognition efficiency.

[0004] In view of the above problems, the present invention proposes a virtual reality (VR) image recognition method, which aims to perform blur judgment on images through Gaussian blur processing, analyze the causes of blurred images, and accurately identify normal images. Summary of the invention

[0005] The object of the present invention is to provide a virtual reality (VR) image recognition method and system to solve at least one of the above-mentioned prior art problems.

[0006] In a first aspect, a virtual reality (VR) image recognition method comprises the following steps:

[0007] When the user accesses a virtual reality VR device, a virtual reality image based on the user's perspective is obtained through a virtual camera, marked as an image to be identified, and pixel data of the image is obtained using MATLAB software, where the pixel data includes the pixel value of each pixel point;

[0008] The image to be identified is Gaussian blurred by using a Gaussian blur algorithm to obtain a Gaussian blurred image of the image to be identified, and the two are compared and analyzed based on pixel values ​​to obtain a blur representation value of the image to be identified;

[0009] The blur condition of the image to be identified is judged based on the obtained blur representation value, and the image to be identified is marked as a blurry image or a normal image;

[0010] The number of blurred images in the image to be identified and the simulated motion period in the virtual camera shooting are obtained, and the equipment impact value is obtained by analyzing the equipment impact value. Based on the equipment impact value analysis, it is determined whether the cause of the blurred image is a virtual reality system failure. If so, the equipment is repaired. If not, a motion impact signal is generated. The virtual reality system includes all the hardware and software equipment used to implement virtual reality during the development process.

[0011] Based on the generated motion influence signal, the blur performance values ​​of all the images to be identified during the simulated motion period and the simulated motion rate of the virtual camera at the time node of capturing the image to be identified are analyzed to determine whether the simulated motion rate of the virtual camera has an impact on the blur degree of the image to be identified. If so, the simulated motion rate of the virtual camera is adjusted.

[0012] The obtained normal image is imported into the trained convolutional neural network CNN to complete the required image recognition task, and the successfully recognized normal image is incorporated into the training database of the convolutional neural network to continuously train the convolutional neural network.

[0013] In a second aspect, the present invention provides a virtual reality (VR) image recognition system, comprising: an image acquisition module, a fuzzy analysis module, a fuzzy judgment module, an impact analysis module, a maintenance and control module, and an image recognition module.

[0014] Beneficial effects of the present invention:

[0015] In the scenario where the user accesses a virtual reality VR device, a virtual reality image based on the user's perspective is obtained through a virtual camera, marked as an image to be recognized, and pixel data of the image is obtained. The image to be recognized is Gaussian blurred by a Gaussian blur algorithm to obtain a Gaussian blurred image of the image to be recognized. The two are compared and analyzed based on pixel values ​​to obtain a blur performance value of the image to be recognized. The blur situation of the image to be recognized is judged, and the image to be recognized is marked as a blurred image or a normal image. The number of blurred images in the image to be recognized and the simulated motion period in the virtual camera shooting are obtained and analyzed to obtain the device impact value. It is analyzed and determined whether the cause of the blurred image is If there is a problem with the virtual reality system, if so, a device impact signal is generated and the equipment is repaired. If not, a motion impact signal is generated. Based on the motion impact signal, the motion impact data within the simulated motion period is obtained for analysis to determine whether the simulated motion rate of the virtual camera has an impact on the blurriness of the image to be recognized. If so, the simulated motion rate of the virtual camera is regulated. The obtained normal image is imported into the trained convolutional neural network to complete the required image recognition task, and the successfully recognized normal image is incorporated into the training database of the convolutional neural network. The convolutional neural network is continuously trained to improve image recognition accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 This is a flowchart of a virtual reality (VR) image recognition method provided by Embodiment 1 of the present invention;

[0018] Figure 2 It is a structural diagram of a virtual reality VR image recognition system provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0020] Embodiment 1

[0021] Figure 1 This is a flowchart of a virtual reality VR image recognition method provided in the first embodiment of the present invention. The embodiment of the present invention can be applied to the case of virtual reality image recognition. The virtual reality VR image recognition method can be executed by a virtual reality VR image recognition system.

[0022] like Figure 1 As shown, a virtual reality (VR) image recognition method provided by an embodiment of the present invention specifically includes the following steps:

[0023] S1. When a user accesses a virtual reality VR device, a virtual reality image based on the user's perspective is obtained through a virtual camera, marked as an image to be identified, and pixel data of the image is obtained using MATLAB software, wherein the pixel data includes the pixel value of each pixel point;

[0024] Specifically, in a virtual reality development environment, a virtual camera associated with a virtual reality device is configured. By simulating a tracking system, the virtual camera simulates the user's perspective and captures and shoots the virtual reality scene image seen by the user to obtain an image to be recognized. The tracking system is a system in the virtual reality device that accurately tracks the user's head and hand movements through built-in sensors.

[0025] In some embodiments, the pixel value of each pixel point of the image to be identified is obtained using MATLAB software. Specifically, the imread function is used to read the image file to be identified, and the size function is used to obtain the size of the image, which includes the number of rows, columns and channels of the image. Then, the pixel value of a specific pixel point is obtained by specifying the index of the row and column.

[0026] S2, performing Gaussian blur processing on the image to be identified by using a Gaussian blur algorithm to obtain a Gaussian blurred image of the image to be identified, and performing a comparative analysis between the two based on pixel values ​​to obtain a blur representation value of the image to be identified;

[0027] Specifically, the size of the Gaussian kernel (also called the Gaussian filter) is selected. The size of the Gaussian kernel determines the spatial range of the blur effect. It is usually an odd number × an odd number (such as 3×3, 5×5, 7×7, etc.) so that the center pixel point can be clearly determined; a Gaussian kernel matrix is ​​generated with the center pixel point as the origin, and the values ​​of other points are determined by the simplified two-dimensional Gaussian function formula The calculation results show that, where x and y represent the number of rows and columns in the spatial range affected by the pixel blur, Represents the standard deviation of the Gaussian kernel; apply the constructed Gaussian kernel matrix to the image. The specific steps are: cover the Gaussian kernel matrix on a certain pixel point of the image, calculate the weighted average of all pixel values ​​in the covered area, the weight of the weighted average is determined by the element value in the Gaussian kernel matrix, and then use the obtained value as the new value of the pixel point. Repeat the operation for each pixel point in the image until the entire image is processed, and obtain a Gaussian blurred image of the image to be identified;

[0028] The image to be recognized is evenly divided into several small image blocks, and the pixel values ​​of each pixel point in the area where each small image block is located before and after Gaussian blur processing are obtained. The pixel values ​​are summed and averaged to obtain the original image mean and the blurred image mean of the small image block. The difference between the two is processed and the absolute value is taken to obtain the blurred difference;

[0029] Based on any small image block, the blur difference value is compared with the blur difference threshold;

[0030] If the blur difference of the small image block is less than the blur difference threshold, it means that the image to be identified has a high blur degree, and the small image block is marked as a blurry image block;

[0031] If the blur difference of the small image block is greater than or equal to the blur difference threshold, it means that the blur level of the image to be identified is low and no processing is performed;

[0032] Obtain the number of all fuzzy image blocks in the image to be identified, and perform ratio processing on it with the total number of small image blocks in the image to be identified to obtain the fuzzy number ratio, which is marked as KX;

[0033] Based on any blurred image block, obtain the blurred difference of the blurred image block, perform difference processing on it and the blurred difference threshold and take the absolute value to obtain the blurred absolute difference, obtain the blurred absolute differences of all blurred image blocks in the image to be identified, sum and average them to obtain the blurred average difference of the image to be identified, perform ratio processing on the blurred average difference and the blurred difference threshold to obtain the blur degree ratio, marked as KY;

[0034] The obtained fuzzy number ratio KX and fuzzy degree ratio KY are processed and the formula The fuzzy expression value XY is calculated, where s1 and s2 are both preset proportional coefficients, s1=2.314, s2=3.427;

[0035] It should be noted that the meaning of the blur performance value XY is to judge the blur of the image to be identified by analyzing the similarity number and similarity degree of the small image blocks divided by the image to be identified before and after the Gaussian blur processing. The more similar the image to be identified is before and after the Gaussian blur processing, the larger the blur performance value XY is, indicating that the blur of the image to be identified is more serious;

[0036] For example, a gray image to be recognized with a pixel size of 1920×1080 is captured by a virtual camera, and the standard deviation is taken. The Gaussian kernel size is 1 and 7×7. The image to be identified is Gaussian blurred. The image to be processed is evenly divided into 144 small image blocks. The fuzzy difference of the small image blocks before and after Gaussian blurring is obtained through MATLAB software. The fuzzy difference values ​​are 19, 77, 102, 4, etc. They are not listed here. After comparing the fuzzy difference with the threshold, it is found that there are about 104 small image blocks that generate fuzzy signals. Three decimal places are retained. The fuzzy number ratio KX=0.722 and the fuzzy degree ratio KY=0.541 are obtained through analysis. The fuzzy performance value XY=0.614 is calculated by the formula.

[0037] S3, judging the blurriness of the image to be identified based on the obtained blur representation value, and marking the image to be identified as a blurry image or a normal image;

[0038] Specifically, the fuzzy representation value of the image to be identified is obtained, and the fuzzy representation value is compared and analyzed with the fuzzy representation threshold;

[0039] If the blur representation value of the image to be identified is less than or equal to the blur representation threshold, it means that the blur level of the image to be identified is low, and it is marked as a normal image;

[0040] If the blur representation value of the image to be identified is greater than the blur representation threshold, it means that the blur degree of the image to be identified is high, and it is marked as a blurry image;

[0041] It should be noted that the purpose of the above steps is to filter out images that cannot be accurately recognized due to high blurriness in the images to be recognized, so as to improve the accuracy and efficiency of subsequent image recognition;

[0042] S4, obtaining the number of blurred images in the image to be identified and the simulated motion period in the virtual camera shooting, analyzing them to obtain a device impact value, and judging whether the cause of the blurred image is a virtual reality system failure based on the device impact value analysis, if so, generating a device impact signal and repairing the device, if not, generating a motion impact signal, the virtual reality system includes all the software and hardware devices that implement virtual reality during the development process;

[0043] It should be noted that the virtual camera simulates the user's perspective by simulating the tracking system. Therefore, there is a situation where the image is captured and shot during the simulated action. In the virtual reality environment, the causes of blurred images are divided into simulated action blur and virtual reality system problems that cause image blur.

[0044] The period of simulated motion in the shooting of the virtual camera is obtained, the number of blurred images captured during the simulated motion period is obtained, and the ratio of the blurred images to the total number of blurred images captured is processed to obtain a motion blur ratio, which is marked as EC;

[0045] Obtain the number of all blurred images in the image to be identified, and perform a ratio process on it with the total number of captured images to be identified, to obtain a blurred image ratio, which is marked as HN;

[0046] The obtained motion blur ratio EC and blurred image ratio HN are processed and the formula The equipment impact value EH is calculated, where a1 and a2 are preset proportional coefficients, a1=1.374, a2=2.913;

[0047] For example, in one shooting, the virtual camera captured a total of 100 images to be identified. Using MATLAB software to judge and analyze them, it was found that there were 79 blurred images. Through background query to obtain the simulated motion period of the virtual camera, it was found that the number of blurred images captured during the simulated motion period was 21. After data processing, retaining three decimal places, the motion blur ratio EC=0.266, the blurred image ratio HN=0.790, and the equipment impact value HN=3.387;

[0048] Compare the obtained device impact value with the device impact threshold;

[0049] If the device impact value is less than or equal to the device impact threshold, it means that the cause of the blurred image is not a virtual reality system failure, and a motion impact signal is generated;

[0050] If the device impact value is greater than the device impact threshold, it indicates that the cause of the blurred image is a virtual reality system failure, and a device impact signal is generated;

[0051] It should be noted that, since the image to be identified is captured by a virtual camera in a virtual reality environment, the only causes of blurred images are simulated action blur and problems with the virtual reality system. The fact that the blurred image is not caused by a virtual reality system failure does not mean that there are no blurred images caused by problems with the virtual reality system, but that the number and proportion of such blurred images do not constitute a situation where the virtual reality system fails and needs to be repaired, and it can be judged that the main cause of the blurred image at this time is simulated action blur.

[0052] Arrange developers to inspect and repair the virtual reality system based on the device impact signal to prevent excessive blur in captured images due to virtual reality system failure, which would affect the subsequent image recognition accuracy and efficiency;

[0053] S4. Based on the generated motion influence signal, obtain motion influence data in the simulated motion period, wherein the motion influence data includes blur performance values ​​of all images to be identified in the simulated motion period and simulated motion rate of the virtual camera at the time node of capturing and shooting the images to be identified, analyze the motion influence data, determine whether the simulated motion rate of the virtual camera has an impact on the blur degree of the images to be identified, and if so, adjust the simulated motion rate of the virtual camera;

[0054] Specifically, based on the motion impact signal, the capture and shooting time nodes of all the images to be identified in the simulated motion period are obtained, and the time periods between adjacent time nodes are marked as interval time periods;

[0055] Based on any interval time period, the fuzzy expression values ​​of the image to be identified captured and photographed at the time nodes at the beginning and end of the interval time period are obtained, and the difference between the two is processed to obtain the fuzzy expression difference of the interval time period. Similarly, the simulated motion rate of the virtual camera at the time nodes at the beginning and end of the interval time period is obtained, and the difference between the two is processed to obtain the motion rate difference of the interval time period.

[0056] Based on any interval time period, the blur performance difference and motion rate difference of the interval time period are compared;

[0057] If the two are both positive, negative, and equal to 0, the interval time period is marked as the same direction time period;

[0058] If one is positive and the other is negative, the interval time period is marked as an out-of-direction time period;

[0059] Obtain the number of all same-direction time periods in the simulated motion period, and perform ratio processing on the number of all interval time periods in the simulated motion period to obtain the same-direction number ratio;

[0060] Compare the obtained same-direction number ratio with the same-direction number ratio threshold;

[0061] If the same-direction number ratio is less than or equal to the same-direction number ratio threshold, it is determined that the simulated motion rate of the virtual camera has no effect on the blur degree of the image to be identified, and the appearance of the blurred image in the simulated motion time period is random;

[0062] If the same-direction number ratio is greater than the same-direction number ratio threshold, it is determined that the simulated motion rate of the virtual camera has an impact on the blur degree of the image to be identified, and a rate impact signal is generated;

[0063] Based on the rate-affecting signal, the simulated motion rate of the virtual camera is adjusted to keep it at a motion rate that can capture and shoot normal images, so as to prevent the subsequent image recognition accuracy and efficiency from being affected by too many blurred images;

[0064] S5, importing the obtained normal image into the trained convolutional neural network CNN to complete the required image recognition task, and incorporating the successfully recognized normal image into the training database of the convolutional neural network, and continuously training the convolutional neural network to improve the image recognition accuracy;

[0065] The technical solution of the embodiment of the present invention is as follows: in a scenario where a user accesses a virtual reality VR device, a virtual reality image based on the user's perspective is obtained through a virtual camera, marked as an image to be recognized, pixel data of the image is obtained, Gaussian blur processing is performed on the image to be recognized through a Gaussian blur algorithm to obtain a Gaussian blurred image of the image to be recognized, the two are compared and analyzed based on pixel values ​​to obtain a blur performance value of the image to be recognized, the blur situation of the image to be recognized is judged, the image to be recognized is marked as a blurred image or a normal image, the number of blurred images in the image to be recognized and the simulated motion period in the virtual camera shooting are obtained, and they are analyzed to obtain a device impact value, and the blur image is analyzed and judged. Whether the cause of the image is a problem with the virtual reality system. If so, generate a device impact signal and repair the equipment. If not, generate a motion impact signal. Based on the motion impact signal, obtain the motion impact data during the simulated motion period for analysis to determine whether the simulated motion rate of the virtual camera has an impact on the blurriness of the image to be recognized. If so, adjust the simulated motion rate of the virtual camera. Import the obtained normal image into the trained convolutional neural network to complete the required image recognition task, and incorporate the successfully recognized normal image into the training database of the convolutional neural network. Continuously train the convolutional neural network to improve image recognition accuracy and efficiency.

[0066] Embodiment 2

[0067] like Figure 2 As shown, a virtual reality VR image recognition system provided by an embodiment of the present invention includes:

[0068] Image acquisition module: When the user accesses the virtual reality VR device, the virtual reality image based on the user's perspective is acquired through a virtual camera, marked as an image to be identified, and the pixel data of the image is acquired using MATLAB software, where the pixel data includes the pixel value of each pixel point;

[0069] Fuzzy analysis module: Gaussian blur processing is performed on the image to be identified through the Gaussian blur algorithm to obtain the Gaussian blurred image of the image to be identified, and the two are compared and analyzed based on the pixel value to obtain the fuzzy representation value of the image to be identified;

[0070] Fuzzy judgment module: judges the blurriness of the image to be identified based on the obtained blur representation value, and marks the image to be identified as a blurred image or a normal image;

[0071] Impact analysis module: obtains the number of blurred images in the image to be identified and the simulated motion period in the virtual camera shooting, analyzes them, obtains the device impact value, determines the cause of the blurred image based on the device impact value analysis, and generates a device impact signal or a motion impact signal according to the cause of the blurred image;

[0072] Inspection and control module: based on the generated equipment impact signal, the virtual reality system is inspected and repaired. Based on the generated motion impact signal, it is analyzed and determined whether the simulated motion rate of the virtual camera has an impact on the blurriness of the image to be recognized. If so, the simulated motion rate of the virtual camera is adjusted.

[0073] Image recognition module: The obtained normal image is imported into the trained convolutional neural network CNN to complete the required image recognition task, and the successfully recognized normal image is incorporated into the training database of the convolutional neural network to continuously train the convolutional neural network.

[0074] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0075] The above is a detailed description of an embodiment of the present invention, but the content is only a preferred embodiment of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A virtual reality (VR) image recognition method, characterized in that: The following steps are involved: When the user accesses a virtual reality VR device, a virtual reality image based on the user's perspective is obtained through a virtual camera, marked as an image to be identified, and pixel data of the image is obtained using MATLAB software, where the pixel data includes the pixel value of each pixel point; The image to be identified is Gaussian blurred by using a Gaussian blur algorithm to obtain a Gaussian blurred image of the image to be identified, and the two are compared and analyzed based on pixel values ​​to obtain a blur representation value of the image to be identified; The blur condition of the image to be identified is judged based on the obtained blur representation value, and the image to be identified is marked as a blurry image or a normal image; The number of blurred images in the image to be identified and the simulated motion period in the virtual camera shooting are obtained, and the equipment impact value is obtained by analyzing the equipment impact value. Based on the equipment impact value analysis, it is determined whether the cause of the blurred image is a virtual reality system failure. If so, the equipment is repaired. If not, a motion impact signal is generated. The virtual reality system includes all the hardware and software equipment used to implement virtual reality during the development process. Based on the generated motion influence signal, the blur performance values ​​of all the images to be identified during the simulated motion period and the simulated motion rate of the virtual camera at the time node of capturing the image to be identified are analyzed to determine whether the simulated motion rate of the virtual camera has an impact on the blur degree of the image to be identified. If so, the simulated motion rate of the virtual camera is adjusted. The obtained normal image is imported into the trained convolutional neural network CNN to complete the required image recognition task, and the successfully recognized normal image is incorporated into the training database of the convolutional neural network to continuously train the convolutional neural network.

2. A virtual reality (VR) image recognition method according to claim 1, characterized in that: The specific method of obtaining the fuzzy representation value XY is: The obtained fuzzy number ratio KX and fuzzy degree ratio KY are processed and the fuzzy performance value XY is calculated by a weighted formula.

3. A virtual reality (VR) image recognition method according to claim 2, characterized in that: The specific method of obtaining the fuzzy number ratio KX and the fuzzy degree ratio KY is as follows: Obtain the number of all fuzzy image blocks in the image to be identified, and perform ratio processing on it with the total number of small image blocks in the image to be identified to obtain the fuzzy number ratio, which is marked as KX; Based on any blurred image block, obtain the blurred difference of the blurred image block, perform difference processing on it and the blurred difference threshold and take the absolute value to obtain the blurred absolute difference, obtain the blurred absolute differences of all blurred image blocks in the image to be identified, sum and average them to obtain the blurred mean difference of the image to be identified, perform ratio processing on the blurred mean difference and the blurred difference threshold to obtain the blur degree ratio, which is marked as KY.

4. A virtual reality (VR) image recognition method according to claim 3, characterized in that: The specific method of obtaining the blurred image block is: The image to be recognized is evenly divided into several small image blocks, and the pixel values ​​of each pixel point in the area where each small image block is located before and after Gaussian blur processing are obtained. The pixel values ​​are summed and averaged to obtain the original image mean and the blurred image mean of the small image block. The difference between the two is processed and the absolute value is taken to obtain the blurred difference; Based on any small image block, the blur difference value is compared with the blur difference threshold; If the blur difference value of the small image block is less than the blur difference threshold, it means that the blur degree of the image to be identified is relatively high, and the small image block is marked as a blurred image block.

5. A virtual reality (VR) image recognition method according to claim 1, characterized in that: The specific method of judging the blur of the image to be recognized is: Obtain the fuzzy representation value of the image to be identified, and compare and analyze it with the fuzzy representation threshold; If the blur representation value of the image to be identified is less than or equal to the blur representation threshold, it means that the blur level of the image to be identified is low, and it is marked as a normal image; If the blur representation value of the image to be identified is greater than the blur representation threshold, it means that the blur degree of the image to be identified is high, and it is marked as a blurry image.

6. A virtual reality (VR) image recognition method according to claim 1, characterized in that: The specific method of analyzing and judging whether the cause of the blurred image is a virtual reality system failure is as follows: Compare the obtained device impact value with the device impact threshold; If the device impact value is less than or equal to the device impact threshold, it means that the cause of the blurred image is not a virtual reality system failure, and a motion impact signal is generated; If the device impact value is greater than the device impact threshold, it indicates that the cause of the blurred image is a virtual reality system failure, and a device impact signal is generated.

7. A virtual reality (VR) image recognition method according to claim 6, characterized in that: The specific method for obtaining the device impact value is as follows: The period of simulated motion in the shooting of the virtual camera is obtained, the number of blurred images captured during the simulated motion period is obtained, and the ratio of the blurred images to the total number of blurred images captured is processed to obtain a motion blur ratio, which is marked as EC; Obtain the number of all blurred images in the image to be identified, and perform a ratio process on it with the total number of captured images to be identified, to obtain a blurred image ratio, which is marked as HN; The obtained motion blur ratio EC and blurred image ratio HN are processed and the equipment influence value EH is calculated by the formula.

8. A virtual reality (VR) image recognition method according to claim 1, characterized in that: The specific method of determining whether the simulated motion rate of the virtual camera has an impact on the blur degree of the image to be recognized is: Obtain the number of all same-direction time periods in the simulated motion period, and perform ratio processing on the number of all interval time periods in the simulated motion period to obtain the same-direction number ratio; Compare the obtained same-direction number ratio with the same-direction number ratio threshold; If the same-direction number ratio is less than or equal to the same-direction number ratio threshold, it is determined that the simulated motion rate of the virtual camera has no effect on the blur degree of the image to be identified, and the appearance of the blurred image in the simulated motion time period is random; If the same-direction number ratio is greater than the same-direction number ratio threshold, it is determined that the simulated motion rate of the virtual camera has an impact on the blur degree of the image to be identified, and a rate impact signal is generated.

9. A virtual reality (VR) image recognition method according to claim 8, characterized in that: The specific method for obtaining the same-direction number ratio is as follows: Based on the motion impact signal, the capture and shooting time nodes of all the images to be identified in the simulated motion period are obtained, and the time periods between adjacent time nodes are marked as interval time periods; Based on any interval time period, the fuzzy expression values ​​of the image to be identified captured and photographed at the time nodes at the beginning and end of the interval time period are obtained, and the difference between the two is processed to obtain the fuzzy expression difference of the interval time period. Similarly, the simulated motion rate of the virtual camera at the time nodes at the beginning and end of the interval time period is obtained, and the difference between the two is processed to obtain the motion rate difference of the interval time period. Based on any interval time period, the blur performance difference and motion rate difference of the interval time period are compared; If the two are both positive, negative, and equal to 0, the interval time period is marked as the same direction time period; The number of all same-direction time periods in the simulated motion period is obtained, and the ratio is processed with the total number of all interval time periods in the simulated motion period to obtain the same-direction number ratio.

10. A virtual reality (VR) image recognition system, the system being used to implement the recognition method according to any one of claims 1 to 9, characterized in that: include: Image acquisition module: When the user accesses the virtual reality device, the virtual reality image based on the user's perspective is acquired through a virtual camera, marked as an image to be identified, and the pixel data of the image is acquired using MATLAB software, where the pixel data includes the pixel value of each pixel point; Fuzzy analysis module: Gaussian blur processing is performed on the image to be identified through the Gaussian blur algorithm to obtain the Gaussian blurred image of the image to be identified, and the two are compared and analyzed based on the pixel value to obtain the fuzzy representation value of the image to be identified; Fuzzy judgment module: judges the blurriness of the image to be identified based on the obtained blur representation value, and marks the image to be identified as a blurred image or a normal image; Impact analysis module: obtains the number of blurred images in the image to be identified and the simulated motion period in the virtual camera shooting, analyzes them, obtains the device impact value, determines the cause of the blurred image based on the device impact value analysis, and generates a device impact signal or a motion impact signal according to the cause of the blurred image; Inspection and control module: based on the generated equipment impact signal, the virtual reality system is inspected and repaired. Based on the generated motion impact signal, it is analyzed and determined whether the simulated motion rate of the virtual camera has an impact on the blurriness of the image to be recognized. If so, the simulated motion rate of the virtual camera is adjusted. Image recognition module: import the obtained normal image into the trained convolutional neural network to complete the required image recognition task, and incorporate the successfully recognized normal image into the training database of the convolutional neural network to continuously train the convolutional neural network.

Images