Weight updating iteration method and system of feature matching model and storage medium

Abstract

The invention discloses a weight updating iteration method of a feature matching model. The weight updating iteration method comprises the following steps: collecting image data containing position and attitude information of intelligent glasses in different operation environments; generating a training data set according to the collected image data, and uploading the training data set to cloud equipment; the cloud device performs incremental training on a feature matching model according to the training set to obtain a latest feature matching model and a corresponding model weight; when it is judged that the latest feature matching model result is superior to the current feature matching model, issuing a model weight corresponding to the latest feature matching model; and updating the current feature matching model used in the intelligent glasses according to the issued model weight. The invention further discloses a corresponding system and a storage medium. According to the invention, through interaction between the terminal device and the cloud device, automatic updating iteration of the weight of the feature matching model can be realized, the updating efficiency is improved, and the adaptability is high.

Description

Technical Field

[0001] The present invention relates to the technical field of updating and iterating deep learning models, and in particular to a weight updating and iterative method, system and storage medium for a feature matching model of smart glasses in a vehicle. Background Art

[0002] Smart glasses (such as VR glasses) are increasingly used in vehicles, and passengers wear smart glasses to play games or watch movies in a moving car. In order to ensure the correct display of smart content and the user's comfortable experience, the system needs to obtain the position coordinates of the smart glasses in the vehicle in real time. In the process of obtaining the position coordinates of the smart glasses in the vehicle, feature matching plays a vital role.

[0003] Generally, multiple cameras are installed in the vehicle to collect images containing smart glasses in real time. These images contain not only information about the smart glasses themselves, but also environmental information inside the vehicle. The system extracts features from the collected images, identifies key points in the images (such as corner points, edge points, etc.) and their descriptors (i.e., attribute information of feature points), and these feature points will serve as the basis for subsequent matching. The system matches the feature points in the current image with the feature points in the previous image or pre-built image. The feature matching algorithm based on deep learning can find matching points more accurately and maintain a high matching rate even under different lighting, weather or camera postures. Through the matched feature point pairs, the system can calculate the position and posture of the smart glasses at the current moment. This information will be used to adjust the display angle and depth of the smart content to ensure that users get the best viewing or gaming experience.

[0004] Feature matching can be divided into traditional matching algorithms and deep learning-based feature matching algorithms. The accuracy and generalization of deep learning-based feature matching algorithms are generally higher than those of traditional feature matching algorithms. In order to obtain a higher-quality feature matching model, it is necessary to collect a large amount of cabin image data to train the model, and then deploy it to the equipment used for visual positioning in the cabin. The constantly changing cabin environment, such as car models, different lighting, weather, different camera postures in the vehicle, occlusion, etc., will degrade the model performance.

[0005] Therefore, in the existing technology, there are solutions that use incremental learning to achieve model updates. However, most of the existing incremental learning technical solutions only consider fine-tuning the model through small sample data from the algorithm side, without comprehensively considering the resources required for the deployment of the actual model, and lack the steps of automated updating and iterating the model. The actual deployment of its model still relies on manual judgment of the new model and manual deployment to the device, which is inefficient and cannot adapt to the ever-changing needs and environment. Summary of the invention

[0006] The technical problem to be solved by the present invention is that the present invention proposes a weight update and iteration method, system and storage medium for a feature matching model, which can realize automatic update and iteration of the weights of the feature matching model through the interaction between the terminal device and the cloud device, thereby improving the update efficiency and having strong adaptability.

[0007] As one aspect of the present invention, a weight updating iterative method for a feature matching model is provided, which at least comprises the following steps:

[0008] Monitor the operating environment information of the smart glasses in the vehicle, and collect image data containing the position and posture information of the smart glasses in different operating environments;

[0009] Generate a training data set based on the collected image data, and upload the training data set to a cloud device;

[0010] The cloud device performs incremental training on the feature matching model according to the training set to obtain the latest feature matching model and the corresponding model weights;

[0011] Compare the latest feature matching model with the current feature matching model, and when the latest feature matching model result is better than the current feature matching model, issue the model weight corresponding to the latest feature matching model;

[0012] According to the model weights issued, the current feature matching model used in the smart glasses is updated.

[0013] Among them, further include:

[0014] Uploading the performance index of the current feature matching model of the smart glasses to the cloud device along with the training data set;

[0015] The performance indicator is associated with the current feature matching model; the performance indicator includes at least: inference time, CPU occupancy, and model memory / video memory occupancy information.

[0016] The step of generating a training data set based on the collected image data further includes:

[0017] A self-supervised optimization method is used to obtain image data corresponding to the current operating environment information, wherein the operating environment information includes: weather, lighting conditions and vehicle type;

[0018] According to the proportional relationship between different weather conditions, different lighting conditions and different vehicle types predetermined in the operating environment information, corresponding image data are selected, and after reaching a predetermined number, a training data set for incremental training of the feature matching model is formed.

[0019] Among them, further include:

[0020] After the training data set is uploaded to the cloud device, an incremental training request signal is sent to the cloud device.

[0021] Wherein, comparing the latest feature matching model with the current feature matching model further comprises:

[0022] The performance index of the latest feature matching model is compared with the performance index of the current feature matching model, where the performance index is a single index or a comprehensive index among accuracy, recall rate, model complexity, model inference time, and model occupancy rate.

[0023] Correspondingly, another aspect of the present invention further provides a weight update iteration system for a feature matching model, which at least includes a cloud device and a terminal device connected to smart glasses in a vehicle, wherein:

[0024] The terminal device is used to monitor the operating environment information of the smart glasses in the vehicle, and collect image data containing the position and posture information of the smart glasses under different operating environments; generate a training data set based on the collected image data, and upload the training data set to the cloud device; and receive the model weights sent by the cloud device to update the current feature matching model used in the smart glasses;

[0025] The cloud device is used to perform incremental training on the feature matching model according to the training set to obtain the latest feature matching model and the corresponding model weight; compare the latest feature matching model with the current feature matching model, and when the latest feature matching model result is better than the current feature matching model, send the model weight corresponding to the latest feature matching model to the terminal device.

[0026] Wherein, the terminal device further comprises:

[0027] A data acquisition module is used to monitor the operating environment information of the smart glasses in the vehicle and collect image data containing the position and posture information of the smart glasses under different operating environments; it is used to upload the training data set and the performance indicators of the current feature matching model to the cloud device; the operating environment information includes: weather, lighting conditions and vehicle type, and the performance indicators include at least: inference time, CPU usage, model memory / video memory usage information;

[0028] A model reasoning module, in which a current feature matching model is arranged, is used to update the current feature matching model used in the smart glasses after receiving the model weights sent by the cloud device;

[0029] The visual mapping and positioning module is used to form a training data set based on the image data collected by the data acquisition module;

[0030] The terminal update module is used to send an incremental training request signal to the cloud device after the data acquisition module completes uploading the training data set to the cloud device.

[0031] The cloud device includes:

[0032] A database module, used to store the training data set from the terminal device and the performance indicators of the current feature matching model;

[0033] A model training module, configured to, upon receiving an incremental training request signal from a terminal device, call the training data set in the database module, perform incremental training on the feature matching model, obtain the latest feature matching model and the corresponding model weight, and store them in the database module;

[0034] The cloud update module is used to compare the latest feature matching model with the current feature matching model, and when the latest feature matching model result is better than the current feature matching model, the model weight corresponding to the latest feature matching model is sent to the terminal device.

[0035] Among them, in the visual mapping and positioning module, according to the predetermined proportional relationship of different weather conditions, different lighting conditions and different vehicle types in the operating environment information, the corresponding image data are selected, and after reaching a predetermined number, a training data set for incremental training of the feature matching model is formed;

[0036] In the cloud update module, the performance index of the latest feature matching model is compared with the performance index of the current feature matching model, and the performance index is a single index or a comprehensive index among accuracy, recall rate, model complexity, model inference time, and model occupancy rate.

[0037] Correspondingly, another aspect of the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the aforementioned method when executed by a processor.

[0038] Implementing the embodiments of the present invention has the following beneficial effects:

[0039] The present invention proposes a weight update and iteration method, system and storage medium for a feature matching model. Through the interaction between a terminal device and a cloud device, the weight of the feature matching model can be automatically updated and iterated, thereby improving the update efficiency; and it can continuously adapt to the changing automobile environment, making the position and posture estimation of the smart glasses in the vehicle more accurate.

[0040] In the method provided by the present invention, the device-side and cloud-side applications are managed in the form of services, which makes it easier to achieve flexible deployment and expansion of applications; at the same time, the data collected by the terminal devices are uniformly managed through the cloud-side automatic update service database, which is convenient for maintenance and upgrading, while improving data security; at the same time, the deployment of model reasoning services on the device side improves the real-time positioning and rapid response performance of smart glasses in the cabin. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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 some embodiments of the present invention. For those skilled in the art, without paying creative labor, other drawings obtained based on these drawings still belong to the scope of the present invention.

[0042] Figure 1 A structural schematic diagram of an embodiment of a weight update iteration system for a feature matching model provided by the present invention;

[0043] Figure 2 A schematic diagram of the main process of an embodiment of a weight update iterative method for a feature matching model provided by the present invention. DETAILED DESCRIPTION

[0044] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings.

[0045] like Figure 1 As shown, it is a structural diagram of an embodiment of a weight update iteration system of a feature matching model provided by the present invention; in this embodiment, the weight update iteration system 1 of the feature matching model at least includes a cloud device 2 and a terminal device 1 connected to the smart glasses in the vehicle, wherein:

[0046] The terminal device 1 is used to monitor the operating environment information of the smart glasses in the vehicle, and collect image data containing the position and posture information of the smart glasses under different operating environments; generate a training data set based on the collected image data, and upload the training data set to the cloud device; and receive the model weights sent by the cloud device to update the current feature matching model used in the smart glasses;

[0047] Cloud device 2 is used to perform incremental training on the feature matching model according to the training set to obtain the latest feature matching model and the corresponding model weight; compare the latest feature matching model with the current feature matching model, and when the latest feature matching model result is better than the current feature matching model, send the model weight corresponding to the latest feature matching model to the terminal device.

[0048] More specifically, the terminal device 1 further comprises:

[0049] The data acquisition module 10 is used to monitor the operating environment information of the smart glasses in the vehicle, and collect image data containing the position and posture information of the smart glasses under different operating environments; and is used to upload the training data set and the performance indicators of the current feature matching model to the cloud device; the operating environment information includes: weather, lighting conditions, vehicle type and current data collection time, and the performance indicators include at least: inference time, CPU occupancy, model memory / video memory occupancy information;

[0050] In a specific example, the data acquisition module 10 can use a camera to capture real-time images inside the vehicle, including the position of the smart glasses, the environment inside the vehicle, etc.; an inertial measurement unit (IMU) can be used to measure the acceleration, angular velocity and other motion states of the device, thereby assisting in determining the position and posture of the smart glasses; environmental parameters inside the vehicle can be collected through environmental sensors (such as temperature sensors, humidity sensors, light sensors, etc.); automobile information system: the model, speed, direction and other information of the vehicle can also be obtained through the vehicle's bus system (such as CAN bus) to more accurately understand the motion state of the smart glasses in the vehicle.

[0051] A model reasoning module 11, in which a current feature matching model is arranged, is used to update the current feature matching model used in the smart glasses after receiving the model weights sent by the cloud device;

[0052] A visual mapping and positioning module 12, used to form a training data set based on the image data collected by the data acquisition module;

[0053] The terminal update module 13 is used for the data acquisition module to send an incremental training request signal to the cloud device after uploading the training data set to the cloud device.

[0054] It is understandable that, in a specific example, the model inference module on the device terminal can further deploy an image retrieval model, an image feature extraction model, and a feature matching model, which are used for visual mapping and positioning services of smart glasses. The feature matching model is used to match the feature points of the image and obtain the performance indicators of the current feature matching model, such as inference time, occupancy of hardware resources such as CPU and GPU, and model memory / video memory occupancy information.

[0055] Furthermore, the terminal device is usually an embedded low-power ARM64 architecture-based device installed in a car. Its computing power is higher than that of smart glasses but lower than that of cloud devices. It can be installed in locations such as the roof of the car, behind the windshield, etc.

[0056] More specifically, the cloud device 2 includes:

[0057] Database module 20, used to store the training data set from the terminal device and the performance index of the current feature matching model;

[0058] The model training module 21 is used to call the training data set in the database module upon receiving the incremental training request signal from the terminal device, perform incremental training on the feature matching model, obtain the latest feature matching model and the corresponding model weight, and store them in the database module;

[0059] Among them, the weights are parameter weights in the deep learning model. These weights are learned by the model during the training process and are used to indicate the degree of contribution of different features or inputs to the model output. In a specific example, the weights may be, for example, neural disconnected weights, convolutional layer weights, fully connected layer weights, and weights of other types of layers (such as pooling layers, batch normalization layers, etc.). When adjusting the weights, optimization algorithms (such as gradient descent, Adam, etc.) are usually used to iteratively update the weight values ​​to minimize the loss function (i.e., the difference between the model prediction results and the actual results). By continuously adjusting the weights, the model can perform better on the training data, and it is expected to maintain good performance on unseen test data.

[0060] The cloud update module 22 is used to compare the latest feature matching model with the current feature matching model, and when the latest feature matching model result is better than the current feature matching model, send the model weight corresponding to the latest feature matching model to the terminal device.

[0061] In a specific example, in the visual mapping and positioning module 12, corresponding image data are selected according to the ratio relationship of different weather conditions, different lighting conditions and different vehicle types predetermined in the operating environment information, and after reaching a predetermined number, a training data set for incremental training of the feature matching model is formed;

[0062] In the cloud update module 22, the performance index of the latest feature matching model is compared with the performance index of the current feature matching model, and the performance index is a single index or a comprehensive index among accuracy, recall rate, model complexity, model inference time, and model occupancy rate.

[0063] Furthermore, the cloud device is configured as a high-computing power platform, generally including accelerated neural network computing units such as GPU.

[0064] The multiple modules in the terminal device and cloud device involved in the present invention can be implemented in the form of services. Specifically, model reasoning service, data collection service, visual mapping and positioning service and model automatic update service are adopted in the terminal device; and training service and model automatic update service are adopted in the cloud.

[0065] The following will be combined Figure 2 The principle of the present invention is described in more detail.

[0066] like Figure 2 FIG. 1 is a schematic diagram showing a main flow chart of an embodiment of a weight update iteration method of a feature matching model provided by the present invention. In this embodiment, the method adopts Figure 2 More specifically, the method provided by the present invention comprises at least the following steps:

[0067] Step S10, monitoring the operating environment information of the smart glasses in the vehicle, and collecting image data containing position and posture information of the smart glasses in different operating environments;

[0068] Specifically, within a collection cycle, the data collection service starts and automatically detects environmental conditions in the vehicle, such as weather (sunny, rainy), light intensity, vehicle model, etc. Sensors such as cameras are used to capture image data under these conditions. The data collection service stores the captured image data and the corresponding environmental condition labels locally.

[0069] Step S11, generating a training data set based on the collected image data, and uploading the training data set to a cloud device;

[0070] In a specific example, in this step, generating a training data set based on the collected image data further includes:

[0071] A self-supervised optimization method is used to obtain image data corresponding to the current operating environment information, wherein the operating environment information includes: weather, lighting conditions and vehicle type;

[0072] According to the proportional relationship between different weather conditions, different lighting conditions and different vehicle types predetermined in the operating environment information, corresponding image data are selected, and after reaching a predetermined number, a training data set for incremental training of the feature matching model is formed.

[0073] In the specific example, the composition of the training data set is collected under the conditions shown in Table 1. The collection ratio under weather conditions is sunny day: cloudy day / hazy day / rainy day = 2:1:1, the collection ratio under lighting conditions is daytime: early morning: evening / night = 2:1:1, and the collection ratio of vehicle models is sedan: SUV = 1:1. A total of 10,000 matching pairs are generated by the self-supervised tuning method for constructing the training data set.

[0074] Table 1

[0075] weather illumination Vehicle Model sunny 5:00-7:00 AM Sedan Cloudy / hazy day Daytime 8:00-18:00 SUV rain Evening / night 18:00-21:00

[0076] In a specific example, the feature matching model with better performance in the model inference service is called to process the collected image data. Through the self-supervised optimization method, the correlation and spatial consistency between images are used to generate training data sets. These training data sets contain positive samples (well-matched image pairs) and negative samples (unmatched or low-similarity image pairs). For example, in one example, image data of sunny days, SUV models, and daytime lighting conditions are automatically obtained from the data collection service of the terminal device, and a training data set containing 1,000 matching pairs is generated through the self-supervised tuning method.

[0077] Further including:

[0078] The performance indicators of the current feature matching model of the smart glasses are uploaded to the cloud device together with the training data set; wherein the performance indicators are associated with the current feature matching model; the performance indicators at least include: inference time, CPU occupancy, model memory / video memory occupancy information. These performance indicator information are very important for evaluating model performance and optimizing resource allocation. With this information, the cloud can understand the operating status and performance of the current model of the terminal device, so as to make more reasonable decisions. For example, if a model occupies too many resources on the terminal device but the performance improvement is not significant, the cloud may choose not to update the model, or optimize the model to reduce resource usage.

[0079] After the training data set is uploaded to the cloud device, an incremental training request signal is sent to the cloud device.

[0080] Step S12, the cloud device performs incremental training on the feature matching model according to the training set to obtain the latest feature matching model and the corresponding model weight;

[0081] It can be understood that after the cloud device receives the incremental training request signal from the terminal device, it starts the training service and initializes it to the waiting mode, and calls the training data set in the database module for incremental training, obtains the latest feature matching model and the corresponding model weights, and evaluates the performance indicators of the new model (such as accuracy, recall rate, etc.); then the results of the latest feature matching model (such as indicators) and weights and other information are stored in the database module.

[0082] Step S13, comparing the latest feature matching model with the current feature matching model, and when the latest feature matching model result is better than the current feature matching model, issuing the model weight corresponding to the latest feature matching model;

[0083] Wherein, in this step, comparing the latest feature matching model with the current feature matching model further includes:

[0084] The performance index of the latest feature matching model is compared with the performance index of the current feature matching model, where the performance index is a single index or a comprehensive index among accuracy, recall rate, model complexity, model inference time, and model occupancy rate.

[0085] For example, in one example, when the cloud update module finds that the accuracy and inference time of the new model are better than the old model, it can be determined that the latest feature matching model result is better than the current feature matching model.

[0086] Step S14: updating the current feature matching model used in the smart glasses according to the issued model weights.

[0087] After the model inference module of the terminal device receives the model weights sent by the cloud device, it restarts the device-side inference service to load and use the new feature matching model, thereby completing the update and iteration of the feature matching model.

[0088] It is understandable that, in some embodiments, the latest model may also be updated periodically.

[0089] For more details, please refer to and combine the above Figure 2 The description is not repeated here.

[0090] It can be understood that in the method provided by the present invention, the automatic update and iteration of the feature matching model are managed by deploying multiple services in the terminal device and the cloud device. Model reasoning service, data acquisition service, visual mapping and positioning service and model automatic update service are deployed in the terminal device. In one acquisition cycle, the cabin environment under different conditions, such as weather, lighting, vehicle model, etc., is collected through the data acquisition service, and the collected images are processed by the visual mapping and positioning service in a self-supervised optimization method to generate a training data set. The visual mapping and positioning service calls the feature matching model with the best recent index deployed in the model reasoning service. After the generation of the training data set is completed, the training data set is uploaded to the model automatic update service database of the cloud device through the acquisition service, and the recent better performance indicators are also uploaded together. When the data upload on the device side is completed, the model automatic update service on the device side notifies the model automatic update service on the cloud side to start the training service on the cloud side. After the cloud training service is completed, the new model indicator results and weights are added to the cloud model automatic update service data and compared with the most recent indicator results of the data. If the new model indicator results are better than the most recent results, the cloud automatic update service sends the new model weights to the device-side inference service and restarts the device-side inference service to complete the update and iteration of the feature matching model.

[0091] As another aspect of the present invention, there is also provided a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the above-mentioned Figure 2 For more details, please refer to and combine with the above Figure 2 The description is not repeated here.

[0092] Implementing the embodiments of the present invention has the following beneficial effects:

[0093] The present invention proposes a weight update and iteration method, system and storage medium for a feature matching model. Through the interaction between a terminal device and a cloud device, the weight of the feature matching model can be automatically updated and iterated, thereby improving the update efficiency; and it can continuously adapt to the changing automobile environment, making the position and posture estimation of the smart glasses in the vehicle more accurate.

[0094] In the method provided by the present invention, the device-side and cloud-side applications are managed in the form of services, which makes it easier to achieve flexible deployment and expansion of applications; at the same time, the data collected by the terminal devices are uniformly managed through the cloud-side automatic update service database, which is convenient for maintenance and upgrading, while improving data security; at the same time, the deployment of model reasoning services on the device side improves the real-time positioning and rapid response performance of smart glasses in the cabin.

[0095] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0096] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0097] The above disclosure is only a preferred embodiment of the present invention, which certainly cannot be used to limit the scope of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope of the present invention.

Claims

1. A weight update iterative method for a feature matching model, characterized in that: At least the following steps are included: Monitor the operating environment information of the smart glasses in the vehicle, and collect image data containing the position and posture information of the smart glasses in different operating environments; Generate a training data set based on the collected image data and upload it to the cloud device; The cloud device performs incremental training on the feature matching model according to the training set to obtain the latest feature matching model and the corresponding model weights; When it is determined that the result of the latest feature matching model is better than the current feature matching model, the model weight corresponding to the latest feature matching model is issued; According to the model weights issued, the current feature matching model used in the smart glasses is updated.

2. The method according to claim 1, characterized in that Further including: Uploading the performance index of the current feature matching model of the smart glasses to the cloud device along with the training data set; The performance indicator is associated with the current feature matching model; the performance indicator includes at least: inference time, CPU occupancy, and model memory / video memory occupancy information.

3. The method according to claim 2, characterized in that Generating a training data set based on the collected image data further includes: A self-supervised optimization method is used to obtain image data corresponding to the current operating environment information, wherein the operating environment information includes: weather, lighting conditions and vehicle type; According to the proportional relationship between different weather conditions, different lighting conditions and different vehicle types predetermined in the operating environment information, corresponding image data are selected, and after reaching a predetermined number, a training data set for incremental training of the feature matching model is formed.

4. The method according to claim 3, characterized in that: Further including: After the training data set is uploaded to the cloud device, an incremental training request signal is sent to the cloud device.

5. The method according to claim 4, characterized in that Further including: The performance index of the latest feature matching model is compared with the performance index of the current feature matching model, where the performance index is a single index or a comprehensive index among accuracy, recall rate, model complexity, model inference time, and model occupancy rate.

6. A weight update iterative system for a feature matching model, characterized in that: At least includes a cloud device and a terminal device connected to the smart glasses in the vehicle, wherein: The terminal device is used to monitor the operating environment information of the smart glasses in the vehicle, and collect image data containing the position and posture information of the smart glasses in different operating environments; generate a training data set based on the collected image data and upload it to the cloud device; and receive the model weights sent by the cloud device to update the current feature matching model used in the smart glasses; The cloud device is used to perform incremental training on the feature matching model according to the training set to obtain the latest feature matching model and the corresponding model weights; when it is judged that the latest feature matching model result is better than the current feature matching model, the model weights corresponding to the latest feature matching model are sent to the terminal device.

7. The system according to claim 6, characterized in that The terminal device further comprises: A data acquisition module is used to monitor the operating environment information of the smart glasses in the vehicle and collect image data containing the position and posture information of the smart glasses under different operating environments; it is used to upload the training data set and the performance indicators of the current feature matching model to the cloud device; the operating environment information includes: weather, lighting conditions and vehicle type, and the performance indicators include at least: inference time, CPU usage, model memory / video memory usage information; A model reasoning module, in which a current feature matching model is arranged, is used to update the current feature matching model used in the smart glasses after receiving the model weights sent by the cloud device; The visual mapping and positioning module is used to form a training data set based on the image data collected by the data acquisition module; The terminal update module is used to send an incremental training request signal to the cloud device after the data acquisition module completes uploading the training data set to the cloud device.

8. The system according to claim 7, characterized in that The cloud device includes: A database module, used to store the training data set from the terminal device and the performance indicators of the current feature matching model; A model training module, configured to, upon receiving an incremental training request signal from a terminal device, call the training data set in the database module, perform incremental training on the feature matching model, obtain the latest feature matching model and the corresponding model weight, and store them in the database module; The cloud update module is used to compare the latest feature matching model with the current feature matching model, and when it is determined that the latest feature matching model result is better than the current feature matching model, the model weight corresponding to the latest feature matching model is sent to the terminal device.

9. The system according to claim 8, characterized in that in: In the visual mapping and positioning module, corresponding image data are selected according to the ratio relationship between different weather conditions, different lighting conditions and different vehicle types predetermined in the operating environment information, and after reaching a predetermined number, a training data set for incremental training of the feature matching model is formed; In the cloud update module, the performance index of the latest feature matching model is compared with the performance index of the current feature matching model, and the performance index is a single index or a comprehensive index among accuracy, recall rate, model complexity, model inference time, and model occupancy rate.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images