Distance measurement system based on artificial intelligence

Abstract

The invention discloses a distance measurement system based on artificial intelligence, and the system can achieve the automatic grabbing of size parameters of glasses. According to the system, a camera is adopted to capture a glasses image, a gyroscope and an artificial intelligence algorithm are adopted to recognize the spatial position of glasses, spatial coordinates of feature points are obtained, the pixel scale # imgabs0 # of the image is obtained subsequently based on the physical geometrical relationship of the glasses and the real pupil distance of a user, and finally the size parameters of the glasses are obtained. The glasses parameters comprise a glasses width, a pupil distance, a front inclination angle, a surface bending angle and a glasses-eye distance; the camera is an apple panel built-in camera; the gyroscope can obtain the pitch angle # imgabs1 #, the deflection angle # imgabs2 # and the inclination angle # imgabs3 #, so that the head posture of a glasses wearer is obtained, and the size measurement precision of the glasses is greatly improved. According to the system, automatic measurement of glasses parameters is realized, the working intensity of manual glasses matching is greatly reduced, and digital upgrading of high-end customized glasses is realized.

Description

Technical Field

[0001] The present invention relates to the field of machine vision, and in particular to an artificial intelligence-based distance measurement system. Background Art

[0002] The size parameters of glasses are key factors affecting wearing comfort, optical performance and visual health. For example: and high Determine the fit of the frame to the face shape, ensuring stability and field of view coverage; pupil distance Precisely locate the optical center of the lens to avoid prism effect and visual fatigue; forward tilt angle Describes the angle between the lens and the visual axis to improve dynamic visual clarity; face curvature angle Enhance the fit between the frame and the facial contour, improve stability and wearing comfort; It directly affects the effective diopter, which is especially important for high-power lenses. When configuring high-end glasses, customers can obtain the width of the glasses according to the product specification sheet after selecting the glasses frame. and high Two physical dimensions, however, pupil distance , forward tilt angle , lens eye distance Kneading the dough corner It is closely related to the user's wearing habits and needs to be obtained on-site by the optician.

[0003] Currently, the eyewear industry primarily relies on manual measurement to obtain fitting parameters. Operators use rulers and protractors, combined with experience and judgment, to collect these parameters. However, this method suffers from low efficiency (a single measurement takes 5-10 minutes), strong subjectivity (for example, pupil distance measurement is easily affected by head posture, and anteversion and face curvature angle measurements are prone to introducing systematic errors), and insufficient adaptability (the measurement error rate for complex frames or unique facial features can exceed 15%). This leads to increased lens processing and fitting failure rates and increased rework costs, hindering the large-scale development of high-end customized eyewear. There is an urgent need to develop automated, high-precision measurement methods to overcome technical bottlenecks in the industry.

[0004] Currently, there are some dimensional measurement-related technologies on the market. For example, Chinese patent CN202510119348.0 discloses a machine vision-based rebar dimensional detection method. Through an image preprocessing algorithm, denoising and bimodal threshold segmentation are used to separate rebar from complex backgrounds. This method can complete rebar dimensional measurement in a short time, effectively overcoming the errors caused by improper operation in manual measurement. However, this image preprocessing algorithm can only separate the eyeglass image and cannot correct image distortion caused by changes in the wearer's head posture during the photo shoot. Therefore, it cannot obtain accurate dimensional parameters and is not suitable for the eyewear industry.

[0005] Chinese patent CN202510118529.1 discloses a real-time 3D human body perception system and method based on monocular RGB images. The system includes a data acquisition and preprocessing module, a feature extraction and matching module, a 3D reconstruction module, a dimensional measurement module, a posture estimation module, and a result display module. The system uses feature matching and 3D reconstruction to obtain the dimensional parameters of key human body parts. However, 3D reconstruction struggles to balance accuracy and speed, and cannot meet the high-precision and real-time requirements of the eyewear industry.

[0006] To solve the above problems, we have developed an AI-based distance measurement system. We use a camera to capture images of glasses, and then use a gyroscope + AI algorithm to identify the spatial position of the glasses. Based on the physical geometric relationship of the glasses, we obtain the spatial coordinates of the feature points and finally obtain the size parameters of the glasses. Summary of the Invention

[0007] To address the above issues, the present invention provides an artificial intelligence-based distance measurement system for automatically capturing glasses parameters. The system uses a camera to capture images of the glasses, then uses a gyroscope and artificial intelligence algorithm to identify the spatial position of the glasses. Based on the physical geometric relationships of the glasses, the spatial coordinates of the feature points are obtained, and ultimately the size parameters of the glasses are obtained. Specifically,

[0008] The glasses parameters are expressed as: glasses width , glasses high , pupil distance , forward tilt angle 、Surface bend angle and eye distance ;

[0009] The camera uses the built-in camera of the Apple tablet, which can capture the image of the glasses, including the front view , side view and top view ;

[0010] The feature point is expressed as the left hinge , left eye , Left glasses holder , right hinge , right eye and right glasses holder ;

[0011] The gyroscope can obtain the head posture changes of the glasses wearer, including: pitch angle , deflection angle and tilt angle The change in head posture of the wearer of glasses will cause image distortion, which will change the coordinates of the feature points and seriously affect the measurement accuracy of the size. The gyroscope can correct the image distortion and obtain the spatial coordinates of the feature points under the standard viewing angle. The mathematical description of the feature point change process is expressed as:

[0012] in

[0013] In the formula is the spatial coordinate of the feature point under the standard viewing angle, expressed as:

[0014]

[0015]

[0016]

[0017]

[0018]

[0019]

[0020] In the formula is the spatial coordinate of the feature point under the real perspective, expressed as:

[0021]

[0022]

[0023]

[0024]

[0025]

[0026]

[0027] In the formula is the coordinate transformation matrix between the standard perspective and the real perspective, expressed as:

[0028]

[0029] In the formula is the pitch angle rotation matrix, is the deflection angle rotation matrix, is the tilt angle rotation matrix, specifically:

[0030] {R}_{x}\left ( {\theta} \right )=\left [ {1, 0, 0; 0, cos\left ( {\theta} \right ), -sin\left ( {\theta} \right ); 0, sin\left ( {\theta} \right ), cos\left ( {\theta} \right )} \right ]

[0031] {R}_{y}\left ( {\phi} \right )=\left [ {cos\left ( {\phi} \right ), 0, sin\left ( {\phi} \right ); 0, 1, 0; -sin\left ( {\phi} \right ), 0, cos\left ( {\phi} \right )} \right ]

[0032] {R}_{z}\left ( {\psi} \right )=\left [ {cos\left ( {\psi} \right ), -sin\left ( {\psi} \right ), 0; sin\left ( {\psi} \right ), cos\left ( {\psi} \right ), 0; 0, 0, 1} \right ]

[0033] The distance measurement system based on artificial intelligence is characterized in that the artificial intelligence algorithm recognition expression is a deep convolutional neural network, which can obtain the pixel coordinates of the feature points of the glasses under the real viewing angle. ; Feature points under standard viewing angle In the same plane, that is:

[0034]

[0035] Glasses have symmetry, that is:

[0036]

[0037]

[0038]

[0039] Based on the above equation, the actual pupil distance input by the user , calibrate the pixel size of the glasses image and obtain the image scale , specifically:

[0040]

[0041] Based on image scale , front view Able to obtain glasses width , glasses high and pupil distance ; Side view Able to obtain the forward tilt angle and eye distance ; Top view Able to obtain face bending angle ;

[0042] The width of the glasses is:

[0043]

[0044] The advantages of the present invention are:

[0045] 1. High accuracy: The system uses a camera to capture glasses images and then uses a gyroscope to correct image distortion, ultimately obtaining highly accurate dimensional parameters. This method does not require complex image preprocessing and uses artificial intelligence algorithms to identify glasses' feature points for dimensional measurement.

[0046] 2. High integration: Integrating with Apple's ecosystem, the glasses size measurement technology is packaged into an app, making it easier for glasses stores to promote and apply it;

[0047] 3. Strong real-time performance: Compared with dimensional measurement achieved through 3D reconstruction technology, this solution has superior speed and can meet the needs of opticians for real-time interaction;

[0048] 4. Low cost: This solution uses image recognition and gyroscopes to achieve dimensional measurement. The implementation process does not require expensive lidar equipment, greatly saving the hardware investment cost of fitting glasses. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation of the embodiments of the present invention.

[0050] Figure 1 :A distance measurement system based on artificial intelligence - front view And the corresponding glasses features.

[0051] Figure 2:A distance measurement system based on artificial intelligence - side view And the corresponding glasses features.

[0052] Figure 3 :A distance measurement system based on artificial intelligence - top view And the corresponding glasses features.

[0053] Figure 4 : Macroscopic display of the locations of the characteristic points of the glasses.

[0054] Figure 5 :Application scenarios of image recognition algorithms.

[0055] Figure 6 : Pupillary distance , glasses width and glasses high pixel dimensions.

[0056] Figure 7 : Forward tilt angle and eye distance pixel dimensions.

[0057] Figure 8 : Face bending angle pixel dimensions.

[0058] Figure 9 : Image features and corresponding coordinate systems of standard and real perspectives.

[0059] Figure 10 : Experiment 1 and results of accuracy verification of glasses size parameter measurement system.

[0060] Figure 11 : Experiment 2 and results of accuracy verification of glasses size parameter measurement system.

[0061] Figure 12 : Experiment 3 and results of accuracy verification of glasses size parameter measurement system. DETAILED DESCRIPTION

[0062] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0063] Example 1

[0064] Attachment Figure 1-4A schematic diagram of the size parameters and feature points of an AI-based distance measurement system under a standard viewing angle is shown. An AI-based distance measurement system is used for automatically capturing glasses parameters. The system uses a camera to capture images of the glasses, then uses a gyroscope and an AI algorithm to identify the spatial position of the glasses. Based on the physical geometric relationship of the glasses, the spatial coordinates of the feature points are obtained, ultimately obtaining the size parameters of the glasses. Specifically,

[0065] The glasses parameters are expressed as: glasses width , glasses high , pupil distance , forward tilt angle 、Surface bend angle and eye distance ;

[0066] The camera uses the built-in camera of the Apple tablet, which can capture the image of the glasses, including the front view , side view and top view ;

[0067] The feature point is expressed as the left hinge , left eye , Left glasses holder , right hinge , right eye and right glasses holder ;

[0068] The gyroscope can obtain the head posture changes of the glasses wearer, including: pitch angle , deflection angle and tilt angle The change in head posture of the wearer of glasses will cause image distortion, which will change the coordinates of the feature points and seriously affect the measurement accuracy of the size. The gyroscope can correct the image distortion and obtain the spatial coordinates of the feature points under the standard viewing angle. The mathematical description of the feature point change process is expressed as:

[0069] in

[0070] In the formula is the spatial coordinate of the feature point under the standard viewing angle, expressed as:

[0071]

[0072]

[0073]

[0074]

[0075]

[0076]

[0077] In the formula is the spatial coordinate of the feature point under the real perspective, expressed as:

[0078]

[0079]

[0080]

[0081]

[0082]

[0083]

[0084] In the formula is the coordinate transformation matrix between the standard perspective and the real perspective, expressed as:

[0085]

[0086] In the formula is the pitch angle rotation matrix, is the deflection angle rotation matrix, is the tilt angle rotation matrix, specifically:

[0087] {R}_{x}\left ( {\theta} \right )=\left [ {1, 0, 0; 0, cos\left ( {\theta} \right ), -sin\left ( {\theta} \right ); 0, sin\left ( {\theta} \right ), cos\left ( {\theta} \right )} \right ]

[0088] {R}_{y}\left ( {\phi} \right )=\left [ {cos\left ( {\phi} \right ), 0, sin\left ( {\phi} \right ); 0, 1, 0; -sin\left ( {\phi} \right ), 0, cos\left ( {\phi} \right )} \right ]

[0089] {R}_{z}\left ( {\psi} \right )=\left [ {cos\left ( {\psi} \right ), -sin\left ( {\psi} \right ), 0; sin\left ( {\psi} \right ), cos\left ( {\psi} \right ), 0; 0, 0, 1} \right ]

[0090] The distance measurement system based on artificial intelligence is characterized in that the artificial intelligence algorithm recognition expression is a deep convolutional neural network, which can obtain the pixel coordinates of the feature points of the glasses under the real viewing angle. ; Feature points under standard viewing angle In the same plane, that is:

[0091]

[0092] Glasses have symmetry, that is:

[0093]

[0094]

[0095]

[0096] Based on the above equation, the actual pupil distance input by the user , calibrate the pixel size of the glasses image and obtain the image scale , specifically:

[0097]

[0098] Based on image scale , front view Able to obtain glasses width , glasses high and pupil distance ; Side view Able to obtain the forward tilt angle and eye distance ; Top view Able to obtain face bending angle ;

[0099] The width of the glasses is:

[0100] .

[0101] Example 2

[0102] Attachment Figure 5 This paper demonstrates the application of image recognition algorithms. This system uses PP-YOLO-R to identify the pixel coordinates of eyeglass feature points. This algorithm supports detection of rotated rectangular frames and outputs detection frames with rotation angles, enabling more accurate positioning of eyeglass frames, hinges, and nose pads, improving feature point recognition accuracy.

[0103] PP-YOLO-R is open sourced under the Apache License 2.0, allowing commercial use and modification while requiring the preservation of the original copyright notice. This agreement ensures the algorithm's scalability and facilitates subsequent optimization and industry application.

[0104] During the system implementation, the AI ​​inference algorithm was deployed on the Alibaba Cloud ECS server. The specific configuration is: 4 cores and 8GB RAM, 120GB ESSD storage space, and Ubuntu system. The project uses Python code. The specific environment deployment process is as follows:

[0105] 1. Miniconda installation and configuration

[0106] # Download the Miniconda installation script

[0107] wget https: / / repo.anaconda.com / miniconda / Miniconda3-latest-Linux-x86_64.sh

[0108] # Add execute permission

[0109] chmod +x Miniconda3-latest-Linux-x86_64.sh

[0110] # Execute the installation script

[0111] . / Miniconda3-latest-Linux-x86_64.sh

[0112] # After the installation is complete, reload the bash configuration

[0113] source ~ / .bashrc

[0114] 2. Create a Python 3.10 environment

[0115] # Create a Python 3.10 environment named glass

[0116] conda create -n glass python=3.10

[0117] # Activate the environment

[0118] conda activate haoya.

[0119] The client submits a feature point recognition request through the RESTful API interface, transmitting the JPEG image and gyroscope raw data in the multipart / form-data format. After receiving the POST request, the server first performs a data integrity check. The specific code is as follows:

[0120] @app.route(' / api / v1 / front / status', methods=['POST'])

[0121] def get_front_image_dimensions():

[0122] # Check file upload fields

[0123] if 'file' not in request.files:

[0124] return jsonify({

[0125] "code": 400,

[0126] "msg": "Missing file field"

[0127] }), 400

[0128] file = request.files['file']

[0129] # Verify file existence

[0130] if file.filename == '':

[0131] return jsonify({

[0132] "code": 400,

[0133] "msg": "Empty file uploaded"

[0134] }), 400

[0135] # Verify file type

[0136] if not file.content_type.startswith('image / '):

[0137] return jsonify({

[0138] "code": 400,

[0139] "msg": "Only image files are allowed"

[0140] }), 400

[0141] The server establishes a comprehensive data persistence solution: All incoming images are renamed and stored directly on the Alibaba Cloud server. The raw data and file names sent by the gyroscope are stored in a MySQL database. To ensure file name uniqueness, a UUID+image format is used as the file name. The specific Python code is as follows:

[0142] # Generate a unique UUID as the file name

[0143] unique_filename = f"{uuid.uuid4()}.{image.format.lower()}"

[0144] Example 3

[0145] Attachment Figure 6 Demonstrated standards Under the viewing angle, pupil distance , glasses width and glasses high The pixel size calculation process is as follows. The PP-YOLO-R algorithm shown in Example 2 is used to identify the positions of the pupil and glasses in the image. In the accompanying figure, point1, point2, point3, and point4 represent the coordinates of the frame position, and point5 and point6 represent the left and right pupils, respectively. Therefore, the pixel size parameters of the glasses are:

[0146] def front_view_calculate(feature_points):

[0147] # Extract the coordinates of point1 and point3

[0148] point1 = feature_points["point1"]

[0149] point3 = feature_points["point3"]

[0150] # Calculate the horizontal coordinate increment

[0151] x_increment = point3["x"] - point1["x"]

[0152] y_increment = point3["y"] - point1["y"]

[0153] # Extract the coordinates of point5 and point6

[0154] Point5 = feature_points["point5"]

[0155] Point6 = feature_points["point6"]

[0156] # Calculate the distance from point5 to point6

[0157] dx = point5["x"] – point6["x"]

[0158] dy = point5["y"] – point6["y"]

[0159] distance = math.sqrt(dx ** 2 + dy ** 2)

[0160] return x_increment, distance, y_increment

[0161] Attachment Figure 7 Demonstrated standards Viewing angle, forward tilt and eye distance The pixel size calculation process of the glasses is shown in the figure. Point 1 and point 2 represent the position coordinates of the lenses, and the pixel size parameters of the glasses are:

[0162] def side_view_calculate(feature_points):

[0163] # Extract the coordinates of two points

[0164] point1 = feature_points["point1"]

[0165] point2 = feature_points["point2"]

[0166] x1, y1 = point1["x"], point1["y"]

[0167] x2, y2 = point2["x"], point2["y"]

[0168] # Calculate distance L

[0169] L = math.sqrt((x2 - x1) ** 2 + (y2 - y1) ** 2)

[0170] # Calculate the angle xita (radians)

[0171] dx = abs(x2 - x1)

[0172] dy = abs(y2 - y1)

[0173] xita_radians = math.atan2(dx, dy) # Use atan2 to avoid division by zero errors

[0174] xita_degrees = math.degrees(xita_radians)

[0175] return L, xita_degrees

[0176] Attachment Figure 8 Demonstrated standards Under the viewing angle, the surface is curved In the figure, point 1 and point 2 represent the positions of the left hinge and the left mirror support, respectively; point 3 and point 4 represent the positions of the right hinge and the right mirror support, respectively. The pixel size parameters of the glasses are:

[0177] def top_view_calculate(feature_points):

[0178] # Extract the two points of the first line

[0179] point1 = feature_points["point1"]

[0180] point2 = feature_points["point2"]

[0181] # Extract the two points of the second line

[0182] point3 = feature_points["point3"]

[0183] point4 = feature_points["point4"]

[0184] # Calculate the slope m1 of the first line

[0185] dx1 = point2["x"] - point1["x"]

[0186] dy1 = point2["y"] - point1["y"]

[0187] if dx1 == 0: # Prevent division by zero error and handle vertical lines

[0188] m1 = float('inf') # infinite slope

[0189] else:

[0190] m1 = dy1 / dx1

[0191] # Calculate the slope m2 of the second line

[0192] dx2 = point4["x"] - point3["x"]

[0193] dy2 = point4["y"] - point3["y"]

[0194] if dx2 == 0: # Prevent division by zero error and handle vertical lines

[0195] m2 = float('inf') # infinite slope

[0196] else:

[0197] m2 = dy2 / dx2

[0198] # Calculate the angle between two lines

[0199] if m1 == float('inf') and m2 == float('inf'):

[0200] # Both lines are vertical, and the angle between them is 0 degrees

[0201] theta = 0.0

[0202] elif m1 == float('inf'):

[0203] # The first line is a vertical line, the second line is a normal line

[0204] theta = 90.0 - math.degrees(math.atan(abs(m2)))

[0205] elif m2 == float('inf'):

[0206] # The second line is a vertical line, the first line is a normal line

[0207] theta = 90.0 - math.degrees(math.atan(abs(m1)))

[0208] else:

[0209] # In normal cases, use the formula to calculate the angle

[0210] tan_theta = abs((m2 - m1) / (1 + m1 * m2))

[0211] theta = 0.5 * math.degrees(math.atan(tan_theta))

[0212] return theta.

[0213] Example 4

[0214] Attachment Figure 9 The image features and corresponding coordinate systems of the standard and real perspectives are shown. This section shows how to use gyroscopes to correct the size measurement errors caused by image distortion, thereby promoting the large-scale development of high-end customized glasses. The process of obtaining spatial coordinates is shown in Example 3, and its pixel coordinates can be expressed as:

[0215]

[0216]

[0217]

[0218]

[0219]

[0220]

[0221] Pitch angle obtained by the gyroscope , deflection angle and tilt angle ,therefore,

[0222] {R}_{x}\left ( {\theta} \right )=\left [ {1, 0, 0; 0, cos\left ( {\theta} \right ), -sin\left ( {\theta} \right ); 0, sin\left ( {\theta} \right ), cos\left ( {\theta} \right )} \right ]

[0223] {R}_{y}\left ( {\phi} \right )=\left [ {cos\left ( {\phi} \right ), 0, sin\left ( {\phi} \right ); 0, 1, 0; -sin\left ( {\phi} \right ), 0, cos\left ( {\phi} \right )} \right ]

[0224] {R}_{z}\left ( {\psi} \right )=\left [ {cos\left ( {\psi} \right ), -sin\left ( {\psi} \right ), 0; sin\left ( {\psi} \right ), cos\left ( {\psi} \right ), 0; 0, 0, 1} \right ] .

[0225] Substituting the gyroscope data into the above formula, we get:

[0226]

[0227] \left [ {1, 0, 0; 0, 0.9986, 0.0523; 0, -0.0523, 0.9986} \right ]

[0228]

[0229] \left [ {0.9962, 0, -0.0872; 0, 1, 0; 0, 0.0872, 0.9962} \right ]

[0230]

[0231] \left [ {0.9998, 0.0175, 0; -0.0175, 0.9998, 0; 0, 0, 1} \right ]

[0232] Coordinate transformation matrix between standard perspective and real perspective for:

[0233]

[0234] \left [ {0.996, 0.01743, -0.0872; -0.0129, 0.9985, 0.0521; 0.088, -0.0508, 0.9948} \right ]

[0235] Feature points under standard viewing angle In the same plane, that is:

[0236]

[0237] Substituting into the above formula, we can know:

[0238]

[0239]

[0240]

[0241]

[0242]

[0243]

[0244] Glasses have symmetry, that is: ,

[0245]

[0246]

[0247] By combining the above equations, we can obtain the coordinates in the real coordinate system. , specifically:

[0248]

[0249]

[0250]

[0251]

[0252]

[0253]

[0254] Coordinates in standard coordinates , specifically:

[0255]

[0256]

[0257]

[0258]

[0259]

[0260]

[0261] Example 5

[0262] Attachment Figure 10-12 The team demonstrated the accuracy verification experiments and results of the eyeglass size parameter measurement system. Before the project was put into use in Japan, the team conducted multiple experiments to verify the accuracy of the eyeglass size measurement system. The experimental results show that the measurement results are relatively stable, with slightly better accuracy than manual measurement and meeting industry standards.

[0263] The specific implementation method described above provides a detailed description of the objectives, technical solutions and beneficial effects of the present invention. It should be understood that the above is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An artificial intelligence-based distance measurement system, applied to automatic capture of eyeglass parameters, characterized in that: The system uses a camera to capture images of the glasses, and then uses a gyroscope + artificial intelligence algorithm to identify the spatial position and size parameters of the glasses. Specifically: The glasses parameters are expressed as: glasses width , glasses high , pupil distance , forward tilt angle 、Surface bend angle and eye distance ; The camera uses the built-in camera of the Apple tablet, which can capture the image of the glasses, including the front view , side view and top view ; The feature point is expressed as the left hinge , left eye , Left glasses holder , right hinge , right eye and right glasses holder ; The gyroscope can obtain the head posture changes of the glasses wearer, including: pitch angle , deflection angle and tilt angle The change in head posture of the wearer of glasses will cause image distortion, which will change the coordinates of the feature points and seriously affect the measurement accuracy of the size. The gyroscope can correct the image distortion and obtain the spatial coordinates of the feature points under the standard viewing angle. The mathematical description of the feature point change process is expressed as: in In the formula is the spatial coordinate of the feature point under the standard viewing angle, expressed as: In the formula is the spatial coordinate of the feature point under the real perspective, expressed as: In the formula is the coordinate transformation matrix between the standard perspective and the real perspective, expressed as: In the formula is the pitch angle rotation matrix, is the deflection angle rotation matrix, is the tilt angle rotation matrix, specifically: 。 2. The distance measurement system based on artificial intelligence according to claim 1, characterized in that: The artificial intelligence algorithm recognition expression is a deep convolutional neural network, which can obtain the pixel coordinates of the feature points of the glasses under real viewing angles. ; Feature points under standard viewing angle In the same plane, that is: Glasses have symmetry, that is: Based on the above equation, the actual pupil distance input by the user , calibrate the pixel size of the glasses image and obtain the image scale , specifically: Based on image scale , front view Able to obtain glasses width , glasses high and pupil distance ; Side view Able to obtain the forward tilt angle and eye distance ; Top view Able to obtain face bending angle ; The width of the glasses is: 。

Images