Image depth acquisition device with optical zoom RGB lens and indication system
Through the use of an image depth of the image with optical zoom RGB lens in the two -eyed depth camera, combined with the feature extraction and in -depth calculation of fixed focus and zoom images, it solves the difficulty of obtaining depth information caused by the inability to adjust the lens in the existing technology. Real -time and accurate depth acquisition and instructions.
Patent Information
- Application Number
- CN202411369401.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2025-01-28
AI Technical Summary
The fixed-focus RGB lens of existing binocular depth cameras cannot be adjusted according to the application scenario, and the image cannot be enlarged or reduced, resulting in the inability to obtain the depth information of pixel points on the image, and it cannot be applied to real-time indications of dynamically changing scene conversion.
Using an image depth obtaining device with an image with optical zoom RGB lens, this device includes a fixed focus RGB lens image depth acquisition unit and optical zoom RGB lens. And combined with the characteristics of the zoom image, obtain the depth information of the zoom image in real time.
A real -time accurate image depth of the image is accurately obtained in real time, and the depth of the deepening device is accurately accurately obtained in real time.
Smart Images
Figure CN119379760A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of three-dimensional space indication technology, and in particular to an image depth acquisition device and an indication system with an optical zoom RGB lens. Background Art
[0002] Depth cameras, also known as 3D cameras, have color image acquisition and three-dimensional space depth measurement functions. They calibrate the depth information of each pixel on the color image through a certain algorithm. They can be widely used in mobile phones, robots, drones, logistics, VR / AR, smart homes, security, and car driving assistance.
[0003] The existing methods for obtaining three-dimensional space depth information through image recognition mainly include the following:
[0004] 1) Binocular stereo vision: two cameras are used to capture images from different angles and use parallax to calculate distance. This method works well under natural lighting and is low cost, but has poor matching effect on textureless surfaces and requires a lot of calculations.
[0005] 2) Structured light: Depth is calculated by projecting light of a known pattern and capturing its deformation. This method has high accuracy for close-range measurements, but the accuracy decreases at long distances and is greatly affected by ambient light.
[0006] 3) Time of Flight (ToF): This method determines the distance by measuring the flight time of a light pulse. This method measures long distances and is less affected by surface characteristics, but is more affected by ambient light intensity and reflective surfaces.
[0007] 4) Monocular depth estimation: Using a monocular camera combined with a deep learning algorithm to learn depth information from a single image. This method does not require additional hardware, but requires a large amount of training data and has low ability to distinguish between textureless surfaces and similar objects.
[0008] 5) 3D reconstruction based on deep learning: 3D images are processed through deep convolutional neural networks (DCNN) and 3D convolution operations. This method is suitable for medical image processing, 3D shape recognition and other fields.
[0009] In the three-dimensional space indication, binocular stereo vision method is usually used. The binocular depth camera set based on this method includes two fixed-focus stereo cameras and one fixed-focus color camera. Generally, the focus of the three cameras is set on the same horizontal line, and the focal length of the two stereo cameras is the same. The specific setting method is as follows: Figure 1 shown.
[0010] Since the color RGB lens for acquiring images is fixed-focus, this setting method cannot be adjusted according to the application scenario, that is, it is impossible to zoom in or out the image of the viewed object and obtain the depth information of the pixels on the image based on the adjusted image. Therefore, it is impossible to obtain the depth information of the pixels on the image for positioning and indication while viewing finer, farther, and clearer images.
[0011] Based on the existing binocular depth camera settings, there is a technical solution to replace the fixed-focus RGB lens with an optical zoom RGB lens, but this solution requires setting up a specific calibration plate or inspection field in the application scene, or building an image feature point-depth database in a certain space offline based on SLAM technology to replace the calibration plate. However, the above method cannot work in the case of scene conversion, that is, when the new application indication scene does not have a calibration plate or historical image database, which makes the indication mobility and generalization worse, and cannot be applied to real-time indication of dynamically changing scene conversion.
[0012] At the same time, when a new object to be indicated is added to the field of view of the target to be indicated based on the offline construction of the image feature point-depth database in a certain space based on the SLAM technology, the depth acquisition of the real-time target indication image of the zoom camera cannot be completed due to the absence of the corresponding new image-depth data of the indicated target, which makes the real-time and adaptability of the indication worse. Summary of the invention
[0013] Based on this, an object of the present invention is to provide an image depth acquisition device with an optical zoom RGB lens.
[0014] An image depth acquisition device with an optical zoom RGB lens, comprising:
[0015] A fixed-focus RGB lens image depth acquisition unit, used to acquire the fixed-focus image and the depth of each pixel of the fixed-focus image;
[0016] an optical zoom RGB lens, wherein the optical zoom RGB lens has the same field of view direction as the RGB lens in the fixed-focus RGB lens image depth acquisition unit and the axis is parallel, and the maximum field of view range does not exceed that of the RGB lens;
[0017] a processor, the processor being electrically and / or communicatively connected to the fixed-focus RGB lens image depth acquisition unit and the optical zoom RGB lens, the processor acquiring the depth of the optical zoom image in the following manner:
[0018] S1 acquires a color image captured by a fixed-focus RGB lens image depth acquisition unit and an optical zoom RGB lens, and obtains fixed-focus image data I.D1 and zoom image data I.D2;
[0019] S2 extracts each pixel point or pixel block p in the fixed-focus image data I.D1 i The fixed-focus image feature g i , get the fixed focus feature value set G; extract a pixel point or pixel block q in the zoom image data I.D2 j The zoom image feature F qj ;
[0020] S3 determines the zoom image feature F through feature value matching calculation qj The fixed-focus image feature g with the highest matching degree in the fixed-focus feature value set G j ; and in the fixed focus feature-depth map set S(g i ,d i ) to determine the fixed-focus image feature g j The corresponding depth d j ; and based on the depth d j Get a pixel point or pixel block q in the zoom image data I.D2 j Depth.
[0021] Compared with the prior art, the present invention adds a fixed-focus RGB depth acquisition unit on the basis of the existing zoom depth camera, combines the set zoom image depth acquisition method, performs feature extraction and depth calculation on the fixed-focus image taken by the fixed-focus RGB depth acquisition unit, obtains a fixed-focus feature-depth mapping set, performs feature extraction on the image acquired by the optical zoom RGB lens to obtain zoom features, combines the fixed-focus feature-depth mapping set obtained in real time to obtain the depth information of the zoom image, and can realize real-time and accurate acquisition of the changed image depth when zooming in or out on a specific object for change observation, thereby improving the adaptability, mobility and generalization of the depth acquisition device, and having a wide range of application scenarios.
[0022] Furthermore, the fixed-focus RGB lens image depth acquisition unit includes a fixed-focus RGB lens and two depth lenses; the fixed-focus RGB lens is arranged between the two depth lenses, and it and the two depth lenses are arranged in sequence on a straight line, and the focus of the two depth lenses is on the same straight line, and the focal lengths of the two depth lenses are fixed and the same.
[0023] Furthermore, the fixed-focus RGB lens image depth acquisition unit is composed of the focal length of the optical zoom RGB lens and two depth lenses under a specific known focal length; the spatial positions of the optical zoom RGB lens and the two depth lenses are relatively fixed and known, the field of view direction is the same and the axes are parallel, and the focal lengths of the two depth lenses are fixed and the same.
[0024] Furthermore, the optical zoom RGB lens is an RGB lens driven by an electric motor to achieve continuous zoom; or is composed of a plurality of fixed-focus RGB lenses with different focal lengths combined in a specific manner, and a discontinuous zoom RGB lens is achieved by selecting RGB lenses with different focal lengths.
[0025] Furthermore, the fixed focus feature-depth map set S(g i ,d i ) is obtained by the following method:
[0026] SA1 extracts each pixel point or pixel block p in the fixed-focus image data I.D1 i (x pi ,y pi )’s fixed-focus image features g i , get each pixel point or pixel block p i The corresponding fixed focus feature set G = {(p1,g1),(p2,g2),...,(p i ,g i ),...,(p n ,g n )};
[0027] SA2 is based on each pixel or pixel block p i (x pi ,y pi ) image coordinates, fixed focal length FL1, calculate each pixel point or pixel block p i (x pi ,y pi ) corresponds to the depth d i , get each pixel point or pixel block p i The corresponding depth set D = {(p1,d1),(p2,d2),...,(p i ,d i ),...,(p n ,d n )};
[0028] SA3 obtains each pixel point or pixel block p according to the fixed focus feature set G and the depth set D. i The corresponding fixed-focus feature-depth map set S(g i ,d i )={(g1,d1),(g2,d2),...,(g i ,d i ),...,(g n ,d n )}.
[0029] Furthermore, the fixed focus feature-depth map set S(g i ,d i) is obtained by the following method:
[0030] SC1 converts the input fixed-focus image data I.D1 into a graph structure, where N pixels or pixel blocks are N nodes, and each pixel or pixel block p i The image coordinates, color RGB values, texture features, geometric features, edge features and depth are the D features of each node, and the N nodes and the D features of each node form an N×D-dimensional node feature matrix X; the spatial relationship of each pixel or pixel block is formed into an N×N-dimensional symmetric adjacency matrix A;
[0031] SC2 uses a graph convolution network GCN with random initialization parameters or Graph SAGE derived from GCN to perform graph convolution operations on the graph structure to obtain each pixel or pixel block p i The corresponding fixed-focus feature-depth map set S(g i ,d i ).
[0032] Furthermore, scale-invariant feature transform SIFT, or histogram of oriented gradients HOG, or accelerated robust feature SURF, or oriented fast detection and rotation invariance (Oriented FAST and Rotated BRIEF), or local binary pattern LBP are used to extract features from the fixed-focus image data I.D1 and the zoom image data I.D2.
[0033] Furthermore, Brute-Force Matcher or FLANN (Fast Library for Approximate Nearest Neighbors) matching is used to determine the best matching pair by calculating the Euclidean distance or Hamming distance between the feature values.
[0034] On the other hand, the present invention provides an indication system using the above-mentioned image depth acquisition device with optical zoom RGB lens, the indication system also includes an indication unit, the indication unit according to a certain pixel point or pixel block q in the zoom RGB image transmitted by the image depth acquisition device with optical zoom RGB lens j The depth information is used to indicate the target indicator in the three-dimensional space.
[0035] Compared with the prior art, the beneficial effects of the indication system of the present invention are the same as those of the image depth acquisition device with an optical zoom RGB lens, which will not be described in detail here.
[0036] For better understanding and implementation, the present invention is described in detail below with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1It is a schematic diagram of the structure of an existing fixed-focus binocular depth camera;
[0038] Figure 2 Schematic diagram of the structure of a zoom binocular depth camera according to Embodiment 1 of the present invention;
[0039] Figure 3 The present invention is a flowchart of a depth acquisition method of a zoom binocular depth camera. DETAILED DESCRIPTION
[0040] The technical solution of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention.
[0041] The present application provides an image depth acquisition device with an optical zoom RGB lens, the device includes a fixed-focus RGB lens image depth acquisition unit and an optical zoom RGB lens, and extracts image features from a fixed-focus image taken by a fixed-focus GRB lens depth camera and combines the depth of the pixel points to obtain the corresponding depth of the fixed-focus image feature points, and obtains a fixed-focus image feature-depth mapping set; extracts image features from a zoom image taken by the optical zoom RGB lens to obtain a zoom image feature set; performs image feature matching on the fixed-focus image features and the zoom image features, and combines the fixed-focus image feature-depth mapping set to obtain a zoom image feature-depth mapping set, and obtains the depth information of the pixel points of the zoom image. The device can achieve real-time acquisition of the depth information of the observed object when the target object is zoomed in or out, that is, real-time acquisition of the depth of each pixel point of the zoom image, improves the adaptability, mobility and generalization of the depth acquisition device, and has a wide range of application scenarios.
[0042] Example 1
[0043] See also Figure 2 The present invention proposes an image depth acquisition device with an optical zoom RGB lens, comprising a fixed-focus RGB lens 10, an optical zoom RGB lens 20, two depth lenses 30, and a processor 40 electrically and / or communicatively connected to the fixed-focus RGB lens 10, the optical zoom RGB lens 20, and the two depth lenses 30. The fixed-focus RGB lens 10, the zoom RGB lens 20, and the two depth lenses 30 have the same field of view and parallel axes, and the fixed-focus RGB lens 10 and the two depth lenses 30 form a fixed-focus RGB lens image depth acquisition unit.
[0044] Specifically, the resolution and field of view of the fixed-focus RGB lens 10 are determined according to the requirements of the application scenario. For application scenarios with high precision requirements and clear images, a high-pixel, high-resolution fixed-focus RGB camera is selected.
[0045] The optical zoom RGB lens 20 is relatively fixed and closely adjacent to the fixed-focus RGB lens 10 in spatial setting position, and has the same field of view direction as the fixed-focus RGB lens 10 and an axis parallel to that of the fixed-focus RGB lens 10. When setting the spatially adjacent positions of the fixed-focus RGB lens 10 and the optical zoom RGB lens 20, the two RGB lenses are set side by side and closely adjacent to each other as much as possible according to the selected lens model. The optical zoom multiple, resolution, and field of view of the optical zoom RGB lens 20 are jointly determined according to the requirements of the application scenario and the selected fixed-focus RGB lens 10. During implementation, the optical zoom RGB lens 20 can use an RGB lens driven by an electric motor to achieve continuous zoom; it can also be selected to be composed of a plurality of fixed-focus RGB lenses with different focal lengths combined in a specific manner, and a discontinuous zoom RGB lens can be achieved by selecting RGB lenses with different focal lengths.
[0046] The two depth lenses 30 are respectively a first depth lens 31 and a second depth lens 32. The first depth lens 31, the fixed-focus RGB lens 10, and the second depth lens 32 are arranged in a straight line in sequence, and the focal lengths of the first depth lens 31 and the second depth lens 32 are fixed and the same. The first depth lens 31, the second depth lens 32, and the fixed-focus RGB lens 10 constitute a fixed-focus RGB lens image depth acquisition unit.
[0047] The fixed-focus RGB lens 10 , the optical zoom RGB lens 20 , and the two depth lenses 30 are in the same three-dimensional coordinate system, and the origin O of the three-dimensional coordinate system coincides with the focus of the fixed-focus RGB lens 10 , and the Z axis is perpendicular to the imaging plane of the fixed-focus RGB lens 10 .
[0048] The processor 40 obtains the parameters and image data of the above lens, and performs depth calculation based on the parameters and image data. Commonly used central processing unit CPU, graphics processing unit GPU, image signal processor ISP, etc. can be used, and this application does not limit it.
[0049] The image depth acquisition device with an optical zoom RGB lens is installed on a pan-tilt head or a robotic arm, and can change the field of view to be indicated as the pan-tilt head or the robotic arm rotates.
[0050] See also Figure 3 The present invention provides a zoom image depth acquisition method based on the above-mentioned optical zoom RGB lens image depth acquisition device, and the method includes the following steps.
[0051] S1 obtains the color images captured by the fixed-focus RGB lens and the zoom RGB lens at time t, and obtains the fixed-focus image data I.D1 and the zoom image data I.D2.
[0052] S2 extracts each pixel point or pixel block p in the fixed-focus image data I.D1 i (x pi ,ypi )’s fixed-focus image features g i , get the fixed focus feature value set G; extract a pixel point or pixel block q in the zoom image data I.D2 j (x qj ,y qj )’s zoom image feature F qj .
[0053] A pixel point or pixel block q in the zoom image data I.D2 j (x qj ,y qj ) is the indicated point specified by the user in the zoom image.
[0054] The image features include but are not limited to the image coordinates of each pixel point or pixel block, color RGB value, texture feature, geometric feature and edge feature.
[0055] The fixed focus feature value set G satisfies:
[0056] G={(p1,g1),(p2,g2),...,(p i ,g i ),...,(p n ,g n )}.
[0057] Existing image feature extraction modes such as scale-invariant feature transform SIFT, histogram of oriented gradients HOG, accelerated robust feature SURF, oriented fast detection and rotation invariance (Oriented FAST and Rotated BRIEF), local binary pattern LBP, etc. can all be used for feature extraction of fixed-focus image data I.D1 and zoom image data I.D2, and this application does not impose any restrictions.
[0058] S3 determines the zoom image feature F through feature value matching calculation qj The fixed-focus image feature g with the highest matching degree in the fixed-focus feature value set G j ; and in the fixed focus feature-depth map set S(g i ,d i ) to determine the fixed-focus image feature g j The corresponding depth d j ; and based on the depth d j Get a pixel point or pixel block q in the zoom image data I.D2 j Depth.
[0059] In specific implementation, an eigenvalue matching algorithm can be used to find the value with the highest matching degree, such as Brute-Force Matcher or FLANN (Fast Library for Approximate Nearest Neighbors). The best matching pair is determined by calculating the distance between the eigenvalues. Commonly used distance metrics include Euclidean distance and Hamming distance.
[0060] In a specific implementation, the fixed focus feature-depth map set S(g i ,d i ) can be obtained by the following method.
[0061] SA1 extracts each pixel point or pixel block p in the fixed-focus image data I.D1 i (x pi ,y pi )’s fixed-focus image features g i , get each pixel point or pixel block p i The corresponding fixed focus feature set G = {(p1,g1),(p2,g2),...,(p i ,g i ),...,(p n ,g n )};
[0062] SA2 is based on each pixel or pixel block p i (x pi ,y pi ) image coordinates, fixed focal length FL1, calculate each pixel point or pixel block p i (x pi ,y pi ) corresponds to the depth d i , get each pixel point or pixel block p i The corresponding depth set D = {(p1,d1),(p2,d2),...,(p i ,d i ),...,(p n ,d n )};
[0063] SA3 obtains each pixel point or pixel block p according to the fixed focus feature set G and the depth set D. i The corresponding fixed-focus feature-depth map set S(g i ,d i )={(g1,d1),(g2,d2),...,(g i ,d i ),...,(g n ,d n )}.
[0064] In another embodiment, the fixed focus feature-depth map set S(g i ,d i ) can also be obtained by the following method.
[0065] SC1 converts the input fixed-focus image data I.D1 into a graph structure, where N pixels or pixel blocks are N nodes, and each pixel or pixel block p i The graph coordinates, color RGB values, texture features, geometric features, edge features and depth are the D features of each node, and the N nodes and the D features of each node form an N×D-dimensional node feature matrix X; the spatial relationship of each pixel point or pixel block is formed into an N×N-dimensional symmetric adjacency matrix A.
[0066] SC2 uses a graph convolution network GCN with random initialization parameters or Graph SAGE derived from GCN to perform graph convolution operations on the graph structure to obtain each pixel or pixel block p i The corresponding fixed-focus feature-depth map set S(g i ,d i ).
[0067] Example 2
[0068] This embodiment is basically the same as Embodiment 1, except that the fixed-focus RGB lens image depth acquisition unit is different.
[0069] The image depth acquisition device with an optical zoom RGB lens of this embodiment includes an optical zoom RGB lens, two depth lenses and a processor electrically connected and / or communicatively connected to the optical zoom RGB lens and the two depth lenses. The optical zoom RGB lens is arranged between the two depth lenses; the optical zoom RGB lens and the two depth lenses are arranged in sequence on a straight line, and the focal lengths of the two depth lenses are fixed and the same. When the focal length of the optical zoom RGB lens is adjusted to a specific known focal length, it forms a fixed-focus RGB lens image depth acquisition unit with the two depth cameras.
[0070] The zoom image depth acquisition method based on the image depth acquisition device with an optical zoom RGB lens is different from the method of Embodiment 1 in step S1. Step S'1 of this embodiment is specifically as follows:
[0071] S`1 obtains a color image taken by the zoom RGB lens at a specific known focal length at time t, and a color image taken by the zoom RGB lens under zoom at time t+1, and obtains fixed-focus image data I.D1 and zoom image data I.D2. The position and posture of the image depth acquisition device with an optical zoom RGB lens at time t and time t+1 remain unchanged.
[0072] The fixed-focus image data I.D1 is the image data corresponding to the color image captured by the zoom RGB camera at a specific known focal length f0. Further, the specific known focal length f0 may be the minimum focal length f min .
[0073] At the same time, based on the above-mentioned image depth acquisition device with optical zoom RGB lens, the present invention also provides an indication system using the image depth acquisition device with optical zoom RGB lens, the indication system comprising an indication unit, the indication unit indicating a certain pixel point or pixel block q in the zoom image transmitted by the image depth acquisition device with optical zoom RGB lens j The depth information of the target object in the three-dimensional space is indicated by red or green light.
[0074] Furthermore, the indicating unit is also arranged on a gimbal or a mechanical arm, and rotates synchronously with the image depth acquisition device having an optical zoom RGB lens.
[0075] The terms used in the embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the embodiments of the present application. The singular forms of "a" and "the" used in the embodiments of the present application and the appended claims are intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that, unless otherwise specified, "multiple" and "several" refer to two or more; "and / or" refers to and includes any or all possible combinations of one or more associated listed items; "first", "second", "third", etc. are only used to distinguish, rather than to describe a specific order or sequence, and cannot be understood as indicating or implying relative importance.
[0076] The above embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present invention, and the present invention is also intended to include these modifications and modifications.
Claims
1. An image depth acquisition device with an optical zoom RGB lens, characterized in that: include: A fixed-focus RGB lens image depth acquisition unit, used to acquire the fixed-focus image and the depth of each pixel of the fixed-focus image; an optical zoom RGB lens, wherein the optical zoom RGB lens has the same field of view direction as the RGB lens in the fixed-focus RGB lens image depth acquisition unit and the axis is parallel, and the maximum field of view range does not exceed that of the RGB lens; a processor, the processor being electrically and / or communicatively connected to the fixed-focus RGB lens image depth acquisition unit and the optical zoom RGB lens, the processor acquiring the depth of the optical zoom image in the following manner: S1 acquires a color image captured by a fixed-focus RGB lens image depth acquisition unit and an optical zoom RGB lens, and obtains fixed-focus image data I.D1 and zoom image data I.D2; S2 extracts each pixel point or pixel block p in the fixed-focus image data I.D1 i The fixed-focus image feature g i , get the fixed focus feature value set G; Extract a pixel point or pixel block q in the zoom image data I.D2 j The zoom image feature F qj ; S3 determines the zoom image feature F through feature value matching calculation qj The fixed-focus image feature g with the highest matching degree in the fixed-focus feature value set G j ; and in the fixed focus feature-depth map set S(g i ,d i ) to determine the fixed-focus image feature g j The corresponding depth d j ; and based on the depth d j Get a pixel point or pixel block q in the zoom image data I.D2 j Depth.
2. The image depth acquisition device with an optical zoom RGB lens according to claim 1, characterized in that: The fixed-focus RGB lens image depth acquisition unit includes a fixed-focus RGB lens and two depth lenses; the fixed-focus RGB lens is arranged between the two depth lenses, and the fixed-focus RGB lens and the two depth lenses are arranged in sequence on a straight line, and the focus of the fixed-focus RGB lens and the two depth lenses are on the same straight line, and the focal lengths of the two depth lenses are fixed and the same.
3. The image depth acquisition device with an optical zoom RGB lens according to claim 1, characterized in that: The fixed-focus RGB lens image depth acquisition unit is composed of the focal length of the optical zoom RGB lens and two depth lenses under the condition of a specific known focal length; the spatial positions of the optical zoom RGB lens and the two depth lenses are relatively fixed and known, the field of view directions are the same and the axes are parallel, and the focal lengths of the two depth lenses are fixed and the same.
4. The image depth acquisition device with an optical zoom RGB lens according to claim 2 or 3, characterized in that: The optical zoom RGB lens is an RGB lens driven by an electric motor to achieve continuous zoom; or it is composed of multiple fixed-focus RGB lenses with different focal lengths combined in a specific way, and a discontinuous zoom RGB lens is achieved by selecting RGB lenses with different focal lengths.
5. The image depth acquisition device with an optical zoom RGB lens according to claim 4, characterized in that: The fixed focus feature-depth map set S(g i ,d i ) is obtained by the following method: SA1 extracts each pixel point or pixel block p in the fixed-focus image data I.D1 i (x pi ,y pi )’s fixed-focus image features g i , get each pixel point or pixel block p i The corresponding fixed focus feature set G = {(p1,g1),(p2,g2),...,(p i ,g i ),...,(p n ,g n )}; SA2 is based on each pixel or pixel block p i (x pi ,y pi ) image coordinates, fixed focal length FL1, calculate each pixel point or pixel block p i (x pi ,y pi ) corresponds to the depth d i , get each pixel point or pixel block p i The corresponding depth set D = {(p1,d1),(p2,d2),...,(p i ,d i ),...,(p n ,d n )}; SA3 obtains each pixel point or pixel block p according to the fixed focus feature set G and the depth set D. i The corresponding fixed-focus feature-depth map set S(g i ,d i )={(g1,d1),(g2,d2),...,(g i ,d i ),...,(g n ,d n )}.
6. The image depth acquisition device with an optical zoom RGB lens according to claim 4, characterized in that: The fixed focus feature-depth map set S(g i ,d i ) is obtained by the following method: SC1 converts the input fixed-focus image data I.D1 into a graph structure, where N pixels or pixel blocks are N nodes, and each pixel or pixel block p i The image coordinates, color RGB values, texture features, geometric features, edge features and depth are the D features of each node, and the N nodes and the D features of each node form an N×D-dimensional node feature matrix X; the spatial relationship of each pixel or pixel block is formed into an N×N-dimensional symmetric adjacency matrix A; SC2 uses a graph convolution network GCN with random initialization parameters or Graph SAGE derived from GCN to perform graph convolution operations on the graph structure to obtain each pixel or pixel block p i The corresponding fixed-focus feature-depth map set S(g i ,d i ).
7. The image depth acquisition device with an optical zoom RGB lens according to claim 5 or 6, characterized in that: The fixed-focus image data I.D1 and the zoom image data I.D2 are feature extracted using scale-invariant feature transform SIFT, or histogram of oriented gradients HOG, or accelerated robust feature SURF, or oriented fast detection and rotation invariance (Oriented FAST and Rotated BRIEF), or local binary pattern LBP.
8. The image depth acquisition device with an optical zoom RGB lens according to claim 7, characterized in that: Brute-Force Matcher or FLANN (Fast Library for Approximate Nearest Neighbors) matching is used to determine the best matching pair by calculating the Euclidean distance or Hamming distance between the feature values.
9. An indication system, characterized in that: The invention comprises an indication unit and an image depth acquisition device with an optical zoom RGB lens as claimed in any one of claims 1 to 8, wherein the indication unit indicates a certain pixel point or pixel block q in a zoom RGB image transmitted by the image depth acquisition device with an optical zoom RGB lens. j The depth information is used to indicate the target indicator in the three-dimensional space.
10. The indicator system according to claim 9, characterized in that: The indicating unit and the image depth acquisition device with an optical zoom RGB lens are both arranged on a pan-tilt platform or a mechanical arm, and the indicating unit and the image depth acquisition device with an optical zoom RGB lens rotate synchronously.