Method and system for measuring distance between two points based on real-time image and distance measurement information
Through the combination of real-time image and ranging information, the use of convolutional neural network and multi-layer perceptron for joint optimization, the accuracy and stability of two-point spacing measurement in complex environments is solved, and a high-precision, non-contact measurement method is realized , improves the performance and convenience of the system.
Patent Information
- Application Number
- CN202411329692.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-24
- Publication Date
- 2024-11-05
- Estimated Expiration
- 2044-09-24
AI Technical Summary
The prior art has problems of insufficient accuracy, poor stability and high computing resources in complex environments. Especially when ambient light changes and complex surface characteristics of objects, it is difficult to achieve high-precision and non-contact measurements.
Using a method based on real-time image and ranging information, by obtaining the azimuth difference value, spatial transformation amount and ranging distance, a convolutional neural network and multi-layer perceptron are used for joint optimization to achieve high-precision measurement of two-point spacing.
It improves the accuracy and stability of the two-point spacing measurement, reduces the computing resource requirements, and enhances the system's real-time response and operational convenience.
Smart Images

Figure CN118897293A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of distance measurement, and more particularly to a method and system for measuring the distance between two points based on real-time images and distance measurement information. Background Art
[0002] At present, traditional distance measurement methods usually use technologies such as laser ranging, ultrasonic ranging or mechanical measurement. However, these methods have certain limitations in certain application scenarios. For example, when the ambient light changes greatly or the surface characteristics of the measured object are complex, the measurement accuracy may be affected; in addition, some measurement methods may require contact with the measured object, which may not be applicable to some precision instruments or fragile materials.
[0003] With the development of computer vision technology, image-based measurement methods have gradually attracted attention; these methods can obtain distance information in a non-contact manner by analyzing feature points in images. However, traditional image measurement methods often rely on complex algorithms and a large amount of computing resources, and still have shortcomings in measurement accuracy and stability.
[0004] In the existing image-coordinated ranging module technology, the method of determining the distance between two points faces many challenges. First, the accuracy problem is particularly prominent. Due to the limitations of camera resolution and image processing algorithm accuracy, existing methods are difficult to meet the requirements of certain high-precision application scenarios. In addition, environmental factors such as changes in lighting and the complexity of object surface characteristics will also have a significant impact on the measurement results, resulting in reduced accuracy and reliability. At the same time, the existing technology is also insufficient in computing resources and algorithm complexity, which not only increases costs but may also affect the real-time response capability of the system. In a dynamic environment, it is particularly difficult to maintain the stability and accuracy of the measurement, and some systems are highly dependent on user interaction, which not only increases the complexity of the operation but also may introduce human errors.
[0005] Therefore, how to achieve efficient and accurate measurement of the distance between two targets in a non-contact manner is a problem that technical personnel in this field need to solve urgently. Summary of the invention
[0006] In view of this, the present invention provides a two-point distance measurement method and system based on real-time images and ranging information, which realizes efficient and accurate measurement of the distance between two targets in a non-contact manner and meets diverse high-precision measurement needs.
[0007] In order to achieve the above object, the present invention adopts the following technical solution:
[0008] A method for measuring the distance between two points based on real-time images and distance measurement information, comprising:
[0009] Based on the directions of the measuring point and the two targets to be measured, the azimuth difference is obtained;
[0010] Acquire two frames of images of the two targets to be measured, and input them into a spatial transformation estimation network to obtain a spatial transformation;
[0011] Acquire a first ranging distance and a second ranging distance between the measuring point and the two objects to be measured;
[0012] The azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance are spliced and input into a joint optimization network to obtain the distance between the two targets to be measured.
[0013] Preferably, the method for obtaining the azimuth difference is:
[0014] Based on the measurement points facing the two targets to be measured respectively, first posture information and second posture information are correspondingly obtained;
[0015] A posture change value based on a difference between the first posture information and the second posture information;
[0016] The azimuth angle difference is obtained by performing projection based on the posture change value.
[0017] Preferably, the first posture information and the second posture information are obtained by the same method, which are:
[0018] Obtaining the angular velocity, acceleration and magnetic field strength of the measuring point when it faces the target to be measured;
[0019] The angular velocity is used as a short-term prediction of the posture, and the acceleration and the magnetic field strength are used to correct the long-term posture deviation, and finally the posture information is obtained by fusion.
[0020] Preferably, after acquiring the two frames of images, image preprocessing is also included:
[0021] Noise filtering, interference filtering and automatic adjustment of image parameters are performed based on the two frames of images.
[0022] Preferably, the spatial transformation estimation network comprises: a first convolutional neural network, a second convolutional neural network and an iterative registration estimator;
[0023] The front frame image and the rear frame image of the front and rear two frames of images are respectively input into the first convolutional neural network and the second convolutional neural network for feature extraction, and the first extracted features and the second extracted features are correspondingly obtained;
[0024] Performing correlation calculation based on the first extracted feature and the second extracted feature to obtain feature related information;
[0025] The feature-related information is input into the iterative registration estimator to obtain the spatial transformation amount.
[0026] Preferably, the first convolutional neural network and the second convolutional neural network have the same structure, and both include a plurality of feature extraction modules with different input sizes;
[0027] The input size decreases from the first feature extraction module to the last feature extraction module;
[0028] The previous frame image and the next frame image are respectively input into the feature extraction module in sequence for feature extraction, and the first extracted feature and the second extracted feature are correspondingly outputted through the last feature extraction module.
[0029] Preferably, the feature extraction modules of different input sizes have the same structure, and all include: a 2D convolution module, a batch normalization module and a first Relu activation function layer;
[0030] The front frame image and the rear frame image are respectively input into the 2D convolution module, the batch normalization module and the first Relu activation function layer for feature extraction.
[0031] Preferably, the iterative registration estimator data processing process is:
[0032] The feature-related information and the initial unit transformation matrix are input into the iterative registration estimator to obtain an updated unit transformation matrix, which is used as the input for the next iteration. After the iteration threshold is reached, the final unit transformation matrix output by the last iteration is used as the spatial transformation amount.
[0033] Preferably, the joint optimization network includes: a first multi-layer perceptron, a second Relu activation function layer, a self-attention module and a second multi-layer perceptron;
[0034] The azimuth angle difference, the spatial transformation amount, the first ranging distance, and the second ranging distance are spliced and input into the first multi-layer perceptron for feature extraction to obtain fusion features;
[0035] The fused features are activated by the second Relu activation function layer and processed by the self-attention module to obtain refined features;
[0036] The refined features are input into the second multi-layer perceptron for processing to obtain the distance between the two objects to be measured.
[0037] A two-point distance measurement system based on real-time images and distance measurement information, comprising: an azimuth acquisition module, a transformation amount acquisition module, a distance measurement acquisition module and a result output module;
[0038] The azimuth acquisition module is used to obtain the azimuth difference based on the directions of the measurement point and the two targets to be measured;
[0039] The transformation amount acquisition module is used to acquire the previous and next two frames of images of the two targets to be measured, and input them into the spatial transformation amount estimation network to obtain the spatial transformation amount;
[0040] The ranging distance acquisition module is used to acquire the first ranging distance and the second ranging distance between the measuring point and the two targets to be measured;
[0041] The result output module is used to splice the azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance and input them into the joint optimization network to obtain the distance between the two targets to be measured.
[0042] It can be seen from the above technical solutions that, compared with the prior art, the present invention discloses a two-point distance measurement method and system based on real-time images and distance measurement information. The present invention uses distance measurement information, azimuth difference and image registration algorithm to achieve high-precision measurement of the distance between two targets in a complex environment. Compared with the solution that only relies on distance measurement information and azimuth difference, the present invention improves the accuracy, stability and ease of operation of the distance measurement between two targets. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0044] Figure 1 A flow chart of a method for measuring the distance between two points based on real-time images and distance measurement information provided by the present invention.
[0045] Figure 2 This is a schematic diagram of the spatial transformation estimation network structure provided by the present invention.
[0046] Figure 3 This is a schematic diagram of the joint optimization network structure provided by the present invention.
[0047] Figure 4 The present invention provides a schematic diagram of the structure of a two-point distance measurement system based on real-time images and distance measurement information.
[0048] Figure 5 The present invention provides a schematic structural diagram of a two-point distance measurement device based on real-time images and distance measurement information. DETAILED DESCRIPTION
[0049] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.
[0050] Example 1
[0051] like Figure 1 As shown, the embodiment of the present invention discloses a method for measuring the distance between two points based on real-time images and distance measurement information, including:
[0052] Based on the directions of the measuring point and the two targets to be measured, the azimuth difference is obtained;
[0053] Obtain the previous and next two frames of images of the two targets to be measured, and input them into the spatial transformation estimation network to obtain the spatial transformation;
[0054] Acquire a first ranging distance and a second ranging distance between a measuring point and two targets to be measured;
[0055] The azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance are spliced and input into the joint optimization network to obtain the distance between the two targets to be measured.
[0056] The present invention utilizes distance measurement information, azimuth difference and image registration algorithm to achieve high-precision measurement of the distance between two targets in a complex environment.
[0057] Example 2
[0058] The embodiment of the present invention discloses a method for measuring the distance between two points based on real-time images and distance measurement information, comprising:
[0059] Based on the directions of the measurement point and the two targets to be measured, the azimuth difference is obtained:
[0060] Preferably, the method for obtaining the azimuth difference is:
[0061] Based on the measurement points facing two targets to be measured, the first pose information and the second pose information are obtained accordingly;
[0062] A pose change value based on the difference between the first pose information and the second pose information;
[0063] The azimuth angle difference is obtained by projection based on the posture change value.
[0064] Preferably, the first pose information and the second pose information are both pose information of the measurement point.
[0065] Preferably, the first posture information and the second posture information are obtained in the same method, which are:
[0066] Obtain the angular velocity, acceleration and magnetic field strength when the measuring point faces the target to be measured;
[0067] The angular velocity is used as a short-term prediction of the attitude, and the acceleration and magnetic field strength are used to correct the long-term attitude deviation, and finally the posture information is fused.
[0068] Preferably, this embodiment uses a gyroscope, an accelerometer and a magnetometer to obtain angular velocity, acceleration and magnetic field strength respectively.
[0069] Preferably, after obtaining the angular velocity, acceleration and magnetic field strength, data preprocessing is performed respectively:
[0070] The angular velocity is sequentially processed: zero bias correction, temperature drift compensation and Kalman filter operation;
[0071] The acceleration is sequentially processed: removing the influence of gravity, pre-integration, Kalman filtering and coordinate system transformation operations;
[0072] The magnetic field strength is carried out in sequence: hard iron and soft iron correction, Kalman filtering, and coordinate system transformation.
[0073] Preferably, this embodiment uses the Madgwick algorithm to fuse angular velocity, acceleration and magnetic field strength to obtain accurate posture information.
[0074] Preferably, the Madgwick algorithm is a sensor fusion algorithm for attitude estimation, which is used to determine the orientation of an object in three-dimensional space in real time; the attitude at the next moment is predicted based on the angular velocity information provided by the gyroscope, and then the prediction is corrected using the measurements of the accelerometer and magnetometer to obtain a more accurate attitude estimate.
[0075] The Madgwick algorithm combines the data from the gyroscope, accelerometer, and magnetometer through an optimization process, using a weighting factor (usually adjustable) to balance the contributions between the different sensors to obtain a fused attitude estimate.
[0076] Get the two frames of images before and after the two targets to be measured, and input them into the spatial transformation estimation network to obtain the spatial transformation:
[0077] Preferably, the two frames of images before and after the two targets to be measured are: an image at a current moment including the two targets to be measured and an image at a next moment including the two targets to be measured.
[0078] Preferably, after acquiring the two frames of images, image preprocessing is also included:
[0079] Noise filtering, interference filtering and automatic adjustment of image parameters are performed based on the previous and next two frames of images.
[0080] Preferably, in this embodiment, a high-resolution camera is used to obtain the front and back two frames of images of the two targets to be measured, and the noise and interference in the image are filtered through the calibrated ISP (image signal processor) module of the smart chip, and the image parameters are automatically adjusted to adapt to various lighting conditions to obtain high-quality image data.
[0081] Preferably, Figure 2 As shown, the spatial transformation estimation network includes: a first convolutional neural network, a second convolutional neural network and an iterative registration estimator;
[0082] The front frame image and the rear frame image of the front and rear two frames of images are respectively input into the first convolutional neural network and the second convolutional neural network for feature extraction, and the first extracted feature FI and the second extracted feature FV are obtained correspondingly;
[0083] Perform correlation calculation based on the first extracted feature FI and the second extracted feature FV to obtain feature related information;
[0084] The feature-related information is input into the iterative registration estimator to obtain the spatial transformation.
[0085] Preferably, the two HxWxC features are correlated to obtain HxWxCxHxWxC feature-related information, the HxWxC features are regarded as N-dimensional vectors (N=HxWxC), and the two vectors are correlated to obtain NxN "feature-related information".
[0086] Preferably, the first convolutional neural network and the second convolutional neural network have the same structure, and both include a plurality of feature extraction modules with different input sizes;
[0087] The input size decreases from the first feature extraction module to the last feature extraction module;
[0088] The previous frame image and the next frame image are respectively input into the feature extraction module for feature extraction, and the last feature extraction module outputs the first extracted feature and the second extracted feature accordingly.
[0089] Preferably, the size of the feature extraction module output is half of the input (both the length and the width are halved).
[0090] Preferably, the feature extraction modules of different input sizes have the same structure, and all include: a 2D convolution module, a batch normalization module and a first Relu activation function layer;
[0091] The previous frame image and the next frame image are respectively input into the 2D convolution module, the batch normalization module and the first Relu activation function layer for feature extraction.
[0092] Preferably, the iterative registration estimator data processing process is:
[0093] The feature-related information and the initial unit transformation matrix are input into the iterative registration estimator to obtain the updated unit transformation matrix, which is used as the input for the next iteration. After the iteration threshold is reached, the final unit transformation matrix output by the last iteration is used as the spatial transformation amount.
[0094] Preferably, the initial unit transformation matrix is an initial default matrix built into the iterative registration estimator.
[0095] Preferably, the iterative registration estimator structure is the same as the first convolutional neural network and the second convolutional neural network structure, and is used to iteratively optimize the spatial transformation amount based on historical estimation results.
[0096] Get the first distance and the second distance between the measuring point and the two targets to be measured:
[0097] Preferably, this embodiment uses a laser ranging module to directly measure the distances between the measuring point and the two targets to be measured, respectively, and obtains a first ranging distance and a second ranging distance accordingly.
[0098] Preferably, this embodiment uses a laser ranging module of model LH-D2000AH for ranging. The laser ranging module will accurately measure the time interval from the emission of the laser pulse to the reception of the reflected light pulse. This time interval is the flight time of light traveling back and forth between the target object and the ranging module. Using the known speed of light and the measured flight time, the distance from the target object to the ranging module can be calculated.
[0099] The azimuth angle difference, spatial transformation, first ranging distance and second ranging distance are stitched together and input into the joint optimization network to obtain the distance between the two targets to be measured:
[0100] Preferably, Figure 3 As shown, the joint optimization network includes: a first multi-layer perceptron, a second Relu activation function layer, a self-attention module and a second multi-layer perceptron;
[0101] The azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance are spliced and input into the first multi-layer perceptron for feature extraction to obtain fusion features;
[0102] The fused features are activated through the second Relu activation function layer and processed through the self-attention module to obtain refined features;
[0103] The refined features are input into the second multi-layer perceptron for processing to obtain the distance between the two targets to be measured.
[0104] The present invention utilizes ranging information, azimuth difference and image registration algorithm to achieve high-precision measurement of the distance between two targets in a complex environment, thereby improving the accuracy, stability and operational convenience of the distance measurement between the two targets.
[0105] Example 3
[0106] like Figure 4 As shown, a two-point distance measurement system based on real-time images and ranging information includes: an azimuth acquisition module, a transformation amount acquisition module, a ranging distance acquisition module and a result output module;
[0107] An azimuth acquisition module is used to obtain an azimuth difference based on the directions of the measuring point and the two targets to be measured;
[0108] The transformation acquisition module is used to acquire the two frames of images before and after the two targets to be measured, and input them into the spatial transformation estimation network to obtain the spatial transformation;
[0109] A ranging distance acquisition module is used to acquire a first ranging distance and a second ranging distance between a measuring point and two targets to be measured;
[0110] The result output module is used to splice the azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance and input them into the joint optimization network to obtain the distance between the two targets to be measured.
[0111] Example 4
[0112] like Figure 5 As shown, a two-point distance measurement device based on real-time image and distance measurement information includes:
[0113] Smart chip, video image interface, peripheral circuit, distance measurement module and nine-axis sensor;
[0114] The intelligent chip is respectively connected to the video image input interface, the chip peripheral circuit, the ranging module and the nine-axis sensor, and has a built-in spatial transformation amount estimation network and a joint optimization network, which are used to receive the image of the target to be measured and process it to obtain the spatial transformation amount, to receive the first pose information and the second pose information and process them to obtain the azimuth difference, and to obtain the distance between the two targets to be measured based on the spatial transformation amount, the azimuth difference, the first ranging distance and the second ranging distance input into the joint optimization network.
[0115] Video image interface, supports commonly used MIPI, DVP and RGB interfaces, used to input the target image to be tested.
[0116] The distance measuring module is used to measure the distance between the device and two targets to be measured respectively, obtain a first distance measuring distance and a second distance measuring distance correspondingly, and input them into the smart chip.
[0117] The nine-axis sensor includes an accelerometer, a gyroscope and a magnetometer, which are used to respectively measure the first posture information and the second posture information of the device when facing two targets to be measured, and input the information into the smart chip.
[0118] Peripheral circuits include common DDR, EMMC, power supply, etc., which are used for smart chip computing and processing, data storage, power supply for each module, and maintenance of normal operation of the device.
[0119] Example 5
[0120] Based on the same inventive concept, the present invention also provides a computer device, including a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;
[0121] Memory, used to store computer programs;
[0122] The processor, when used to execute the program stored in the memory, can implement a two-point distance measurement method based on real-time image and distance measurement information as in embodiment 1 or 2.
[0123] The electronic device may include: a processor, a communications interface, a memory and a communications bus, wherein the processor, the communications interface and the memory communicate with each other via the communications bus.
[0124] The processor can call the logic instructions in the memory to execute a two-point distance measurement method based on real-time images and distance measurement information in embodiment 1 or 2.
[0125] In addition, the logic instructions in the above-mentioned memory can be implemented in the form of software functional units and can be stored in a computer-readable storage medium when sold or used as an independent product. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or the part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0126] It can be seen from the above technical solutions that the present invention discloses a method and system for measuring the distance between two points based on real-time images and distance measurement information. The present invention uses distance measurement information, azimuth difference and image registration algorithm to achieve high-precision measurement of the distance between two targets in a complex environment. Compared with the solution that only relies on distance measurement information and azimuth difference, the present invention improves the accuracy, stability and ease of operation of the distance measurement between two targets.
[0127] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.
[0128] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A two-point distance measurement method based on real-time image and distance measurement information, characterized in that: include: Based on the directions of the measuring point and the two targets to be measured, the azimuth difference is obtained; Acquire two frames of images of the two targets to be measured, and input them into a spatial transformation estimation network to obtain a spatial transformation; Acquire a first ranging distance and a second ranging distance between the measuring point and the two objects to be measured; The azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance are spliced and input into a joint optimization network to obtain the distance between the two targets to be measured.
2. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 1, characterized in that: The method for obtaining the azimuth difference is: Based on the measurement points facing the two targets to be measured respectively, first posture information and second posture information are correspondingly obtained; A posture change value based on a difference between the first posture information and the second posture information; The azimuth angle difference is obtained by performing projection based on the posture change value.
3. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 2, characterized in that: The first posture information and the second posture information are obtained in the same way, which are: Obtaining the angular velocity, acceleration and magnetic field strength of the measuring point when it faces the target to be measured; The angular velocity is used as a short-term prediction of the posture, and the acceleration and the magnetic field strength are used to correct the long-term posture deviation, and finally the posture information is obtained by fusion.
4. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 1, characterized in that: After obtaining the two frames of images, image preprocessing is also included: Noise filtering, interference filtering and automatic adjustment of image parameters are performed based on the two frames of images.
5. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 1, characterized in that: The spatial transformation estimation network includes: a first convolutional neural network, a second convolutional neural network and an iterative registration estimator; The front frame image and the rear frame image of the front and rear two frames of images are respectively input into the first convolutional neural network and the second convolutional neural network for feature extraction, and the first extracted features and the second extracted features are correspondingly obtained; Performing correlation calculation based on the first extracted feature and the second extracted feature to obtain feature related information; The feature-related information is input into the iterative registration estimator to obtain the spatial transformation amount.
6. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 5, characterized in that: The first convolutional neural network and the second convolutional neural network have the same structure, and both include a plurality of feature extraction modules with different input sizes; The input size decreases from the first feature extraction module to the last feature extraction module; The previous frame image and the next frame image are respectively input into the feature extraction module in sequence for feature extraction, and the first extracted feature and the second extracted feature are correspondingly outputted through the last feature extraction module.
7. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 6, characterized in that: The feature extraction module structures of different input sizes are the same, including: 2D convolution module, batch normalization module and the first Relu activation function layer; The front frame image and the rear frame image are respectively input into the 2D convolution module, the batch normalization module and the first Relu activation function layer for feature extraction.
8. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 5, characterized in that: The iterative registration estimator data processing process is: The feature-related information and the initial unit transformation matrix are input into the iterative registration estimator to obtain an updated unit transformation matrix, which is used as the input for the next iteration. After the iteration threshold is reached, the final unit transformation matrix output by the last iteration is used as the spatial transformation amount.
9. The method for measuring the distance between two points based on real-time images and distance measurement information according to claim 1, characterized in that: The joint optimization network includes: a first multi-layer perceptron, a second Relu activation function layer, a self-attention module and a second multi-layer perceptron; The azimuth angle difference, the spatial transformation amount, the first ranging distance, and the second ranging distance are spliced and input into the first multi-layer perceptron for feature extraction to obtain fusion features; The fused features are activated by the second Relu activation function layer and processed by the self-attention module to obtain refined features; The refined features are input into the second multi-layer perceptron for processing to obtain the distance between the two objects to be measured.
10. A two-point distance measurement system based on real-time image and distance measurement information, applied to a two-point distance measurement method based on real-time image and distance measurement information as claimed in any one of claims 1 to 9, characterized in that: include: Azimuth acquisition module, transformation acquisition module, ranging distance acquisition module and result output module; The azimuth acquisition module is used to obtain the azimuth difference based on the directions of the measurement point and the two targets to be measured; The transformation amount acquisition module is used to acquire the previous and next two frames of images of the two targets to be measured, and input them into the spatial transformation amount estimation network to obtain the spatial transformation amount; The ranging distance acquisition module is used to acquire the first ranging distance and the second ranging distance between the measuring point and the two targets to be measured; The result output module is used to splice the azimuth angle difference, the spatial transformation amount, the first ranging distance and the second ranging distance and input them into the joint optimization network to obtain the distance between the two targets to be measured.
Citation Information
Patent Citations
Distance measurement method and terminal equipment
CN110148167A
Method for controlling robot system to be in place based on target pose space
CN117656052A
Azimuth measuring device and azimuth measuring method
CN1666086A
Photographed image retrieving device and method
JP2005218135A
Position and orientation measurement apparatus, position and orientation measurement method, and storage medium
US20110211066A1
