Baggage identification measurement method and device based on deep learning

Through deep learning and binocular camera technology, combined with lidar, the problems of high cost and large errors in luggage measurement equipment are solved, the accurate acquisition of luggage size information is achieved, and the airport transportation efficiency and service quality are improved.

CN115063472BInactive Publication Date: 2024-11-22CIVIL AVIATION UNIV OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210760186.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-29
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing luggage measurement equipment is costly and the traditional three-dimensional matching algorithm has large errors. It is impossible to effectively identify and measure multiple pieces of luggage of different categories and shapes, which affects airport transportation efficiency and service satisfaction.

Method used

Using a deep learning-based luggage recognition and measurement method, the ACVNet network model is used to combine binocular cameras and lidars to obtain accurate luggage depth maps and point cloud data through transfer learning and triangulation principles, and realize the visual classification of luggage size information.

Benefits of technology

It improves the accuracy and stability of luggage measurement, reduces external noise interference, and provides a more accurate and efficient intelligent luggage identification and handling solution, suitable for complex lighting environments and weak texture areas.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure SMS_1
    Figure SMS_1
  • Figure SMS_2
    Figure SMS_2
  • Figure SMS_3
    Figure SMS_3
Patent Text Reader

Abstract

The present invention relates to a baggage identification and measurement method based on deep learning, comprising the following steps: constructing an ACVNet network model, acquiring a training set and a baggage stereo matching data set, training the ACVNet network model by using the training set, performing transfer learning of the trained ACVNet network model by using the baggage stereo matching data set, and obtaining a final trained ACVNet network model; acquiring a baggage image by using a binocular camera, preprocessing the baggage image, inputting the preprocessed baggage image into the trained ACVNet network model to obtain a disparity map of the baggage, converting the disparity map into a depth map by using the triangulation principle, combining the internal and external parameters obtained by binocular camera calibration, and then obtaining point cloud data of a binocular camera shooting picture; visually classifying the point cloud data and obtaining the size information of the baggage; the present invention can ensure the accuracy and efficiency of baggage size measurement.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent luggage handling, identification and measurement, and in particular to a luggage identification and measurement method and device based on deep learning. Background Art

[0002] At present, the post-processing of checked baggage at domestic and foreign airports is generally done by ground service personnel who visually inspect and manually evaluate the specifications of the bags and manually move the baggage to baggage transport containers or aircraft cargo holds.

[0003] Manual luggage handling requires high labor costs and is prone to damage during the handling process, affecting passengers' flight experience and airport operating efficiency. Intelligent luggage handling requires obtaining information about the size and position of luggage in order to work with the robotic arm to place the luggage in the appropriate position, thereby improving handling efficiency and cargo hold space utilization.

[0004] Currently, luggage is generally measured by traditional grating measurement or line laser scanning. Grating measurement easily causes large errors in luggage size, and the cost of a complete set of laser scanning equipment is relatively high. Therefore, it cannot meet the requirements of identifying and measuring multiple pieces of luggage of different categories and shapes.

[0005] Therefore, in order to solve the above problems, the present invention proposes a non-contact baggage measurement method and device based on deep learning. Summary of the invention

[0006] The purpose of the present invention is to provide a method and device for baggage identification and measurement based on deep learning, so as to solve the problems of high cost of existing baggage measurement equipment and large error of traditional stereo matching algorithm. Compared with traditional algorithms, better size measurement accuracy can be obtained, and the problem of complex and changeable lighting environment in airport transportation environment and poor matching effect of weak texture areas on the baggage surface, which affects the baggage detection and handling efficiency and airport service satisfaction, can be solved.

[0007] The present invention solves the technical problem by adopting the following technical solutions:

[0008] The baggage identification measurement method based on deep learning includes the following steps:

[0009] Construct an ACVNet network model, obtain a training set and a luggage stereo matching data set, use the training set to train the ACVNet network model, and use the luggage stereo matching data set to perform transfer learning on the trained ACVNet network model to obtain the final trained ACVNet network model;

[0010] Use a binocular camera to obtain luggage images and preprocess them. Input the preprocessed luggage images into the trained ACVNet network model to obtain the disparity map of the luggage. Use the triangulation principle to convert the disparity map into a depth map. Combined with the internal and external parameters obtained by the binocular camera calibration, the point cloud data of the binocular camera shooting picture is obtained.

[0011] Visualize and classify point cloud data and obtain luggage size information.

[0012] Furthermore, the method for obtaining the luggage stereo matching dataset is:

[0013] a. Obtain the luggage RGB image and luggage disparity map through the binocular camera, process the acquired luggage RGB image to obtain the left and right eye images of the luggage, and fuse the luggage disparity map with the external participant of the lidar and radar calibration to form a complete disparity map;

[0014] b. Adjust the position and posture of the luggage in the field of view of the camera and lidar, repeat step a, and obtain multiple sets of image pairs and disparity maps, thereby obtaining a luggage stereo matching dataset.

[0015] Furthermore, the method of training the ACVNet network model using the training set is:

[0016] Use the SceneFlow dataset to pre-train the ACVNet network model and save the training weights;

[0017] The ACVNet network model is pre-trained using the real driving scene KITTI2012 and KITTI2015 datasets, and the training weights are saved.

[0018] Furthermore, the method for preprocessing the baggage image is: performing image epipolar correction, image enhancement processing, and image segmentation processing on the baggage image in sequence to obtain a preprocessed baggage image.

[0019] Furthermore, the disparity map is converted into a depth map through the principle of triangulation, and combined with the internal and external parameters obtained by the binocular camera calibration, the method of obtaining the point cloud data of the binocular camera shooting picture is as follows:

[0020] The disparity map is converted into a depth map using the triangulation principle:

[0021]

[0022] Where b is the distance between the left and right optical axes of the binocular camera, f is the focal length of the camera, z is the depth of the object to be measured from the camera, XL is the horizontal coordinate of the point in the left camera image in the left camera image coordinate system, and XR is the horizontal coordinate of the point in the right camera image in the right camera image coordinate system;

[0023] Get the three-dimensional coordinates of each point of the luggage:

[0024]

[0025] Where (x, y, z) is the three-dimensional coordinate information of the luggage, d is the disparity map of the left and right images of the luggage, (X a , Y a ) is the image coordinate value after image preprocessing;

[0026] The obtained luggage depth map is converted into a point cloud map.

[0027] Furthermore, the method for performing epipolar correction on the luggage image is as follows:

[0028] The camera is calibrated using Zhang Zhengyou's calibration method to obtain the internal and external parameters of the binocular camera. The Bouguet image correction algorithm is used to perform epipolar correction on the left and right luggage images obtained by the binocular camera so that the images meet the epipolar constraints of stereo matching.

[0029] Furthermore, the histogram equalization image grayscale transformation enhancement algorithm is used to enhance the luggage image after epipolar correction; and the Hough transform algorithm is used to perform luggage image segmentation on the enhanced image.

[0030] Baggage identification measurement device based on deep learning, including:

[0031] ACVNet network model acquisition module, used to build the ACVNet network model, and obtain the training set and luggage stereo matching data set, use the training set to train the ACVNet network model, use the luggage stereo matching data set to perform transfer learning on the trained ACVNet network model, and obtain the final trained ACVNet network model;

[0032] The baggage image point cloud data acquisition module is used to acquire baggage images using a binocular camera, preprocess the baggage images, input the preprocessed baggage images into the trained ACVNet network model to obtain the disparity map of the baggage, convert the disparity map into a depth map through the principle of triangulation, and combine it with the internal and external parameters obtained by the binocular camera calibration to obtain the point cloud data of the binocular camera shooting picture;

[0033] The baggage size information acquisition module is used to visualize and classify the point cloud data and obtain the size information of the baggage.

[0034] An electronic device comprises: at least one processor; and a memory in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the above-mentioned baggage identification measurement method based on deep learning.

[0035] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the above-mentioned baggage identification and measurement method based on deep learning.

[0036] The advantages and positive effects of the present invention are:

[0037] The present invention uses a binocular camera to collect luggage images, pre-processes the luggage images, obtains relatively accurate luggage images, and reduces the interference of external noise; then uses a pre-trained ACVNet network model to perform fine-tuning on a real luggage data set, adjusts the weight parameters of stereo matching, and uses a depth map obtained by a solid-state laser radar as a supplement to achieve more accurate luggage depth map estimation and obtain three-dimensional information of the luggage. This process improves the accuracy and stability of the luggage visual measurement solution; finally, the output depth map is converted into a luggage point cloud, and the size information of the luggage is obtained through visualization and classification of the point cloud data; therefore, the present invention can provide a more convenient and intelligent luggage size measurement method while ensuring accuracy and efficiency, and provide new ideas for the construction of smart airports and industrial structural reform. DETAILED DESCRIPTION

[0038] First, it should be noted that the specific structure, features and advantages of the present invention will be specifically described below by way of example, but all descriptions are only for illustration and should not be understood as limiting the present invention in any way. In addition, any single technical feature described or implied in the embodiments mentioned herein can still be combined or deleted between these technical features (or their equivalents) to obtain more other embodiments of the present invention that may not be directly mentioned herein.

[0039] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application may be combined with each other.

[0040] The baggage identification and measurement method based on deep learning provided in this embodiment includes the following steps:

[0041] Construct an ACVNet network model, obtain a training set and a luggage stereo matching data set, use the training set to train the ACVNet network model, and use the luggage stereo matching data set to perform transfer learning on the trained ACVNet network model to obtain the final trained ACVNet network model;

[0042] Use a binocular camera to obtain luggage images and preprocess them. Input the preprocessed luggage images into the trained ACVNet network model to obtain the disparity map of the luggage. Use the triangulation principle to convert the disparity map into a depth map. Combined with the internal and external parameters obtained by the binocular camera calibration, the point cloud data of the binocular camera shooting picture is obtained.

[0043] Visualize and classify point cloud data and obtain luggage size information.

[0044] The method for obtaining the luggage stereo matching dataset is as follows: a. obtaining the luggage RGB image and the luggage disparity map through a binocular camera, processing the obtained luggage RGB image to obtain the left and right eye images of the luggage, and fusing the luggage disparity map through a laser radar combined with an external participant calibrated by the radar to form a complete disparity map; b. adjusting the placement position and posture of the luggage under the field of view of the camera and the laser radar, repeating step a, obtaining multiple sets of image pairs and disparity maps, and thus obtaining the luggage stereo matching dataset.

[0045] Specifically, in this embodiment, the method for obtaining the luggage stereo matching data set is:

[0046] Step 2-1-1: Use the MYNT S1030-IR standard version binocular camera and sick tim 561 lidar to create a luggage stereo matching dataset, and fix the camera and lidar;

[0047] Step 2-1-2: Use the MYNT S1030-IR binocular camera to obtain the luggage RGB image and save it;

[0048] Step 2-1-3: Use the MYNT S1030-IR binocular camera to obtain the luggage disparity map and save it;

[0049] Step 2-1-4: Use sick tim561 lidar combined with radar calibrated external parameters to fuse the baggage disparity map obtained by the camera to form a complete disparity map;

[0050] Step 2-1-5: Process the binocular luggage image obtained in step 2-1-2 and save the image as a pair of left and right luggage images;

[0051] Step 2-1-6: Process the binocular baggage disparity map obtained in step 2-1-3 to make it meet the disparity map requirements of the image pair obtained in step 2-1-5, and save it;

[0052] Step 2-1-7: Adjust the position and posture of the luggage on the workbench under the camera and radar field of view, repeat steps 2-1-1, 2-1-2, 2-1-3, 2-1-4, 2-1-5 and 2-1-6, and take multiple sets of image pairs and disparity maps to complete the preparation of the luggage stereo matching data set;

[0053] Specifically, the method of training the ACVNet network model using the training set is:

[0054] Use the SceneFlow dataset to pre-train the ACVNet network model and save the training weights;

[0055] The ACVNet network model is pre-trained using the real driving scene KITTI2012 and KITTI2015 datasets, and the training weights are saved.

[0056] It should be noted that the SceneFlow dataset is a large-scale synthetic dataset that provides more than 30,000 pairs of training images, allowing deep neural networks to be fully trained. ACVNet first performs a model pre-training on this dataset and saves the training weights; the KITTI dataset is divided into two versions, 2012 and 2015. As a real dataset, it has a certain influence in the industry. ACVNet performs a pre-training on KITTI2012 and KITTI2015 respectively, and saves the training weights.

[0057] Furthermore, the method for preprocessing the baggage image is: performing image epipolar correction, image enhancement processing, and image segmentation processing on the baggage image in sequence to obtain a preprocessed baggage image.

[0058] Furthermore, the disparity map is converted into a depth map through the principle of triangulation, and combined with the internal and external parameters obtained by the binocular camera calibration, the method of obtaining the point cloud data of the binocular camera shooting picture is as follows:

[0059] The disparity map is converted into a depth map using the triangulation principle:

[0060]

[0061] Where b is the distance between the left and right optical axes of the binocular camera, f is the focal length of the camera, z is the depth of the object to be measured from the camera, XL is the horizontal coordinate of the point in the left camera image in the left camera image coordinate system, and XR is the horizontal coordinate of the point in the right camera image in the right camera image coordinate system;

[0062] Get the three-dimensional coordinates of each point of the luggage:

[0063]

[0064] Where (x, y, z) is the three-dimensional coordinate information of the luggage, d is the disparity map of the left and right images of the luggage, (X a , Y a ) is the image coordinate value after image preprocessing;

[0065] The obtained luggage depth map is converted into a point cloud map.

[0066] Furthermore, the method for performing epipolar correction on the luggage image is as follows:

[0067] The camera is calibrated by Zhang Zhengyou calibration method to obtain the internal and external parameters of the binocular camera. The Bouguet image correction algorithm is used to perform epipolar correction on the left and right luggage images obtained by the binocular camera so that the images meet the epipolar constraints of stereo matching. The histogram equalization image grayscale transformation enhancement algorithm is used to enhance the luggage image after epipolar correction. The Hough transform algorithm is used to segment the enhanced image.

[0068] Specifically, an experimental platform was built and the binocular camera was fixed. The camera was calibrated using the Zhang Zhengyou calibration method to obtain the internal and external parameters of the binocular camera. Based on the obtained camera parameters, including the camera's distortion parameter K, rotation matrix R, translation matrix T, etc., the Bouguet image correction algorithm was used to perform epipolar correction on the left and right luggage images obtained by the binocular camera, so that the images met the epipolar constraints of stereo matching.

[0069] The histogram equalization image grayscale transformation enhancement algorithm is used to enhance the corrected baggage image. The specific steps are as follows:

[0070] First, the grayscale histogram of the corrected luggage image is obtained, and the probability density function corresponding to each grayscale is obtained by statistics. The formula used is:

[0071]

[0072] In the formula, P(r k ) is the probability density function of the pixel with gray level k, n k is the number of pixels with gray level k, M×N is the total number of image pixels, and L represents the image gray level;

[0073] The cumulative distribution function of each gray level is obtained from the probability density function. The formula used is:

[0074]

[0075] In the formula, S k is the cumulative probability value of the pixel with gray level k in the entire image;

[0076] Round off and calculate to determine the pixel mapping relationship. The formula used is:

[0077] f(k)=(L-1)×S(k)

[0078] Where f(k) is the new pixel value corresponding to the pixel with gray level k after rounding and expansion according to the mapping relationship;

[0079] Calculate a new balanced histogram based on the mapping relationship;

[0080] The Hough transform algorithm is used to perform luggage image segmentation on the enhanced image. The specific steps are as follows:

[0081] Detect the edge of the luggage according to the Canny edge detection algorithm;

[0082] The obtained luggage edge image is transformed into a linear Hough space from the image Cartesian coordinate system to the image polar coordinate Hough space system. The formula used is:

[0083] r=x·cos(θ)+y·sin(θ)

[0084] Where θ is the angle between r and the x-axis, r is the geometric vertical distance to the straight line, x is the horizontal coordinate of the pixel point, and y is the vertical coordinate of the pixel point;

[0085] According to the image polar coordinate Hough space system, set the threshold and output the maximum Hough transform value;

[0086] According to the maximum Hough transform value obtained, the luggage image is segmented, and the formula used is:

[0087]

[0088] In the formula, G is the pixel value of the luggage grayscale image after transformation, g is the pixel value of the luggage grayscale image before transformation, c is the column number of the image pixel, c left To obtain the column number of the left Hough transform value of the image, c right To obtain the column number of the right Hough transform value of the image;

[0089] In this embodiment, the method of visually classifying the point cloud data and obtaining the size information of the luggage is as follows:

[0090] Use the Xiaomi S1030-IR binocular camera SDK and run the program on Visual Studio 2017 to output the binocular image of the luggage and the disparity map obtained by the network, and mark the size information of the luggage to be tested on the left image of the luggage;

[0091] The baggage weight information is obtained through the baggage conveyor device, and the conveyor belt speed is known;

[0092] The length and width data of the luggage can be obtained by the minimum circumscribed rectangle. For point cloud data with a flat surface, the average value of the luggage surface point cloud is taken as the height value to obtain the luggage volume.

[0093] For irregular point clouds, point cloud slicing is used to project them onto the XOY plane. Points on the plane have different depth values, and the volume of the irregular luggage is obtained by integration.

[0094] Example 2

[0095] Baggage identification measurement device based on deep learning, including:

[0096] ACVNet network model acquisition module, used to build the ACVNet network model, and obtain the training set and luggage stereo matching data set, use the training set to train the ACVNet network model, use the luggage stereo matching data set to perform transfer learning on the trained ACVNet network model, and obtain the final trained ACVNet network model;

[0097] The baggage image point cloud data acquisition module is used to acquire baggage images using a binocular camera, preprocess the baggage images, input the preprocessed baggage images into the trained ACVNet network model to obtain the disparity map of the baggage, convert the disparity map into a depth map through the principle of triangulation, and combine it with the internal and external parameters obtained by the binocular camera calibration to obtain the point cloud data of the binocular camera shooting picture;

[0098] The baggage size information acquisition module is used to visualize and classify the point cloud data and obtain the size information of the baggage.

[0099] In addition, this embodiment further provides a computing device, including:

[0100] one or more processing units;

[0101] A storage unit for storing one or more programs,

[0102] Wherein, when the one or more programs are executed by the one or more processing units, the one or more processing units execute the above-mentioned baggage identification and measurement method based on deep learning; it should be noted that the computing device may include but is not limited to a processing unit and a storage unit; those skilled in the art may understand that the computing device including a processing unit and a storage unit does not constitute a limitation on the computing device, and may include more components, or a combination of certain components, or different components, for example, the computing device may also include an input and output device, a network access device, a bus, etc.

[0103] A computer-readable storage medium having a non-volatile program code executable by a processor is also provided, and when the computer program is executed by the processor, the steps of the baggage identification measurement method based on deep learning are implemented; it should be noted that the readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above; the program contained on the readable medium can be transmitted by any appropriate medium, including, but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above. For example, the program code for performing the operation of the present invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and also conventional procedural programming languages ​​such as C language or similar programming languages. The program code can be executed entirely on the user computing device, partially on the user device, as an independent software package, or entirely on a remote computing device or server. In cases involving a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0104] The above embodiments describe the present invention in detail, but the contents are only preferred embodiments of the present invention and cannot be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A baggage identification measurement method based on deep learning, characterized in that: The steps include: Construct an ACVNet network model, obtain a training set and a luggage stereo matching data set, use the training set to train the ACVNet network model, and use the luggage stereo matching data set to perform transfer learning on the trained ACVNet network model to obtain the final trained ACVNet network model; Use a binocular camera to obtain luggage images and preprocess them. Input the preprocessed luggage images into the trained ACVNet network model to obtain the disparity map of the luggage. Use the triangulation principle to convert the disparity map into a depth map. Combined with the internal and external parameters obtained by the binocular camera calibration, the point cloud data of the binocular camera shooting picture is obtained. Visualize and classify point cloud data and obtain luggage size information; The method of converting the disparity map into a depth map by using the triangulation principle and combining it with the internal and external parameters obtained by the binocular camera calibration to obtain the point cloud data of the binocular camera shot is as follows: The disparity map is converted into a depth map using the triangulation principle: Where b is the distance between the left and right optical axes of the binocular camera, f is the focal length of the camera, z is the depth of the object to be measured from the camera, XL is the horizontal coordinate of the point in the left camera image in the left camera image coordinate system, and XR is the horizontal coordinate of the point in the right camera image in the right camera image coordinate system; Get the three-dimensional coordinates of each point of the luggage: Where (x, y, z) is the three-dimensional coordinate information of the luggage, d is the disparity map of the left and right images of the luggage, (X a , Y a ) is the image coordinate value after image preprocessing; The obtained luggage depth map is converted into a point cloud map.

2. The baggage identification and measurement method based on deep learning according to claim 1, characterized in that: The method for obtaining the luggage stereo matching data set is: a. Obtain the luggage RGB image and luggage disparity map through the binocular camera, process the acquired luggage RGB image to obtain the left and right eye images of the luggage, and fuse the luggage disparity map with the external participant of the lidar and radar calibration to form a complete disparity map; b. Adjust the position and posture of the luggage in the field of view of the camera and lidar, repeat step a, and obtain multiple sets of image pairs and disparity maps, thereby obtaining a luggage stereo matching dataset.

3. The baggage identification and measurement method based on deep learning according to claim 2, characterized in that: The method of training the ACVNet network model using the training set is: Use the SceneFlow dataset to pre-train the ACVNet network model and save the training weights; The ACVNet network model is pre-trained using the real driving scene KITTI2012 and KITTI2015 datasets, and the training weights are saved.

4. The baggage identification and measurement method based on deep learning according to claim 3, characterized in that: The method for preprocessing the baggage image is: performing image epipolar correction, image enhancement processing, and image segmentation processing on the baggage image in sequence to obtain the preprocessed baggage image.

5. The baggage identification and measurement method based on deep learning according to claim 4, characterized in that: The method for performing epipolar correction on baggage images is: The camera is calibrated using Zhang Zhengyou's calibration method to obtain the internal and external parameters of the binocular camera. The Bouguet image correction algorithm is used to perform epipolar correction on the left and right luggage images obtained by the binocular camera so that the images meet the epipolar constraints of stereo matching.

6. The baggage identification and measurement method based on deep learning according to claim 5, characterized in that: The histogram equalization image grayscale transformation enhancement algorithm is used to enhance the luggage image after epipolar correction; the Hough transform algorithm is used to segment the enhanced image.

7. A baggage identification and measurement device based on deep learning, characterized in that: include: ACVNet network model acquisition module, used to build the ACVNet network model, and obtain the training set and luggage stereo matching data set, use the training set to train the ACVNet network model, use the luggage stereo matching data set to perform transfer learning on the trained ACVNet network model, and obtain the final trained ACVNet network model; The baggage image point cloud data acquisition module is used to acquire baggage images using a binocular camera, preprocess the baggage images, input the preprocessed baggage images into the trained ACVNet network model to obtain the disparity map of the baggage, convert the disparity map into a depth map through the principle of triangulation, and combine it with the internal and external parameters obtained by the binocular camera calibration to obtain the point cloud data of the binocular camera shooting picture; The method of converting the disparity map into a depth map by using the triangulation principle and combining it with the internal and external parameters obtained by the binocular camera calibration to obtain the point cloud data of the binocular camera shot is as follows: The disparity map is converted into a depth map using the triangulation principle: Where b is the distance between the left and right optical axes of the binocular camera, f is the focal length of the camera, z is the depth of the object to be measured from the camera, XL is the horizontal coordinate of the point in the left camera image in the left camera image coordinate system, and XR is the horizontal coordinate of the point in the right camera image in the right camera image coordinate system; Get the three-dimensional coordinates of each point of the luggage: Where (x, y, z) is the three-dimensional coordinate information of the luggage, d is the disparity map of the left and right images of the luggage, (X a , Y a ) is the image coordinate value after image preprocessing; Convert the obtained luggage depth map into a point cloud map; The baggage size information acquisition module is used to visualize and classify the point cloud data and obtain the size information of the baggage.

8. An electronic device, characterized in that: include: at least one processor; And, a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 6.