Image searching method and device, storage medium and computer equipment

Abstract

The invention discloses an image searching method and device, a storage medium and computer equipment, and the method comprises the steps: constructing a distributed database through employing sample images collected in the driving process of a vehicle; receiving a to-be-searched image for a target scene, and extracting a first image feature corresponding to the to-be-searched image through a preset feature extraction model; extracting a second image feature of each sample image in the distributed database through the feature extraction model; and based on the similarity between the first image feature and each second image feature, obtaining a target matching image corresponding to the target scene in the sample image. According to the method, new data does not need to be directionally collected for the target scene, the images collected by the vehicle at ordinary times are fully utilized to construct the distributed database, and parallel accurate image acquisition is performed in a form of searching images by images, so that the data acquisition cost is reduced, and the data acquisition speed and the data acquisition accuracy are improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to an image search method and device, a storage medium, and a computer device. Background Art

[0002] Current autonomous driving technology is extremely dependent on artificial intelligence algorithms driven by massive data. Massive sample data is the foundation of autonomous driving technology. Without massive data, data-driven algorithms and closed-loop automation built around data lack the foundation for iterative upgrades. To achieve fully autonomous driving, it is necessary to solve millions of Corner Cases (corner cases), but the cost of obtaining the data required to solve a problem is high.

[0003] Taking Lidar perception as an example, when optimizing various CornerCase problems such as false detection and missed detection, the traditional solution is to collect data in a targeted manner by defining filters that cover the target scene. This is costly, time-consuming, and has low accuracy. Summary of the invention

[0004] In view of this, the present application provides an image search method and apparatus, a storage medium, and a computer device, which help to reduce data acquisition costs and improve data acquisition speed and accuracy.

[0005] According to one aspect of the present application, an image search method is provided, the method comprising:

[0006] Using sample images collected by vehicles during driving, a distributed database is constructed;

[0007] Receiving an image to be searched for a target scene, and extracting a first image feature corresponding to the image to be searched through a preset feature extraction model;

[0008] Extracting the second image feature of each sample image in the distributed database through the feature extraction model;

[0009] Based on the similarity between the first image feature and each second image feature, a target matching image corresponding to the target scene is obtained in the sample image.

[0010] Optionally, the use of sample images collected by the vehicle during driving to construct a distributed database includes:

[0011] Acquire sample images collected by the vehicle during driving and sample annotation information of the sample images, and determine at least one storage database corresponding to the sample images based on the sample annotation information and database indexes corresponding to each of a plurality of distributed databases;

[0012] The sample images are stored in the corresponding database.

[0013] Optionally, the determining, based on the sample annotation information and database indexes corresponding to each of the plurality of distributed databases, at least one database corresponding to the sample image includes:

[0014] When the sample image is a visual image, based on the visual sample annotation information of the visual image and the visual database indexes corresponding to the plurality of visual databases, determining at least one visual database corresponding to the visual image, wherein the distributed database includes the visual database;

[0015] When the sample image is a point cloud image, at least one point cloud database corresponding to the point cloud image is determined based on the point cloud sample annotation information of the point cloud image and the point cloud database indexes corresponding to each of the multiple point cloud databases, wherein the distributed database includes the point cloud database.

[0016] Optionally, the visual sample annotation information includes at least one of a camera position, a data collection time, and a data collection scene, and the visual database index includes at least one of a camera position, a data collection time, and a data collection scene; the point cloud sample annotation information includes the data collection time and / or the data collection scene, and the point cloud database index includes the data collection time and / or the data collection scene.

[0017] Optionally, extracting the second image features of each sample image in the distributed database by using the feature extraction model includes:

[0018] Determining a target search database from a plurality of distributed databases according to the search image annotation information of the image to be searched;

[0019] The second image features of each sample image in the target search database are extracted respectively through the feature extraction model.

[0020] Optionally, the use of sample images collected by the vehicle during driving to construct a distributed database includes:

[0021] Perform dimensionality reduction processing on sample images collected during vehicle driving, and use the sample images after dimensionality reduction processing to build a distributed database;

[0022] Correspondingly, extracting the first image feature corresponding to the image to be searched by using a preset feature extraction model includes:

[0023] A dimensionality reduction process is performed on the image to be searched to obtain a reduced-dimensionality image, and a first image feature corresponding to the reduced-dimensionality image is extracted through a feature extraction model.

[0024] Optionally, acquiring a target matching image corresponding to the target scene in the sample image based on the similarity between the first image feature and each second image feature includes:

[0025] Based on the similarity between the first image feature and each second image feature, obtaining a candidate matching image corresponding to the target scene in the sample image;

[0026] Extracting a third image feature corresponding to the image to be searched by using the feature extraction model;

[0027] Performing dimension restoration on the candidate matching image, and extracting a fourth image feature of the candidate matching image after dimension restoration by using the feature extraction model;

[0028] Based on the similarity between the third image feature and the fourth image feature, a target matching image corresponding to the target scene is obtained from the candidate matching images after dimensionality restoration.

[0029] Optionally, before constructing the distributed database using sample images collected by the vehicle during driving, the method further includes:

[0030] Obtain the full amount of sample images collected in real time while the vehicle is driving;

[0031] The sample images are obtained by performing image screening on the full amount of sample images according to a preset sample screening ratio, wherein the preset sample screening ratio is used to constrain the number of images retained within a preset time length.

[0032] According to another aspect of the present application, an image search device is provided, the device comprising:

[0033] A database construction module is used to construct a distributed database using sample images collected by the vehicle during driving;

[0034] A feature extraction module, configured to receive an image to be searched for a target scene, and extract a first image feature corresponding to the image to be searched through a preset feature extraction model; and extract a second image feature of each sample image in the distributed database through the feature extraction model;

[0035] An image acquisition module is used to acquire a target matching image corresponding to the target scene in the sample image based on the similarity between the first image feature and each second image feature.

[0036] Optionally, the database construction module is further used to:

[0037] Acquire sample images collected by the vehicle during driving and sample annotation information of the sample images, and determine at least one storage database corresponding to the sample images based on the sample annotation information and database indexes corresponding to each of a plurality of distributed databases;

[0038] The sample images are stored in the corresponding database.

[0039] Optionally, the database construction module is further used to:

[0040] When the sample image is a visual image, based on the visual sample annotation information of the visual image and the visual database indexes corresponding to the plurality of visual databases, determining at least one visual database corresponding to the visual image, wherein the distributed database includes the visual database;

[0041] When the sample image is a point cloud image, at least one point cloud database corresponding to the point cloud image is determined based on the point cloud sample annotation information of the point cloud image and the point cloud database indexes corresponding to each of the multiple point cloud databases, wherein the distributed database includes the point cloud database.

[0042] Optionally, the visual sample annotation information includes at least one of a camera position, a data collection time, and a data collection scene, and the visual database index includes at least one of a camera position, a data collection time, and a data collection scene; the point cloud sample annotation information includes the data collection time and / or the data collection scene, and the point cloud database index includes the data collection time and / or the data collection scene.

[0043] Optionally, the feature extraction module is further used to:

[0044] Determining a target search database from a plurality of distributed databases according to the search image annotation information of the image to be searched;

[0045] The second image features of each sample image in the target search database are extracted respectively through the feature extraction model.

[0046] Optionally, the database construction module is further used to:

[0047] Perform dimensionality reduction processing on sample images collected during vehicle driving, and use the sample images after dimensionality reduction processing to build a distributed database;

[0048] Accordingly, the feature extraction module is also used for:

[0049] A dimensionality reduction process is performed on the image to be searched to obtain a reduced-dimensionality image, and a first image feature corresponding to the reduced-dimensionality image is extracted through a feature extraction model.

[0050] Optionally, the image acquisition module is further used to:

[0051] Based on the similarity between the first image feature and each second image feature, obtaining a candidate matching image corresponding to the target scene in the sample image;

[0052] Extracting a third image feature corresponding to the image to be searched by using the feature extraction model;

[0053] Performing dimension restoration on the candidate matching image, and extracting a fourth image feature of the candidate matching image after dimension restoration by using the feature extraction model;

[0054] Based on the similarity between the third image feature and the fourth image feature, a target matching image corresponding to the target scene is obtained from the candidate matching images after dimensionality restoration.

[0055] Optionally, the database construction module is further used to:

[0056] Obtain the full amount of sample images collected in real time while the vehicle is driving;

[0057] The sample images are obtained by performing image screening on the full amount of sample images according to a preset sample screening ratio, wherein the preset sample screening ratio is used to constrain the number of images retained within a preset time length.

[0058] According to another aspect of the present application, a storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above-mentioned image search method is implemented.

[0059] According to another aspect of the present application, a computer device is provided, including a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, wherein the processor implements the above-mentioned image search method when executing the program.

[0060] By means of the above technical solution, the present application provides an image search method and device, storage medium, and computer equipment, which use sample images collected by the vehicle during driving to build a distributed database, so as to obtain target matching images similar to the image to be searched of the target scene by searching the image in a distributed manner for a specific target scene. The present application does not need to collect new data for the target scene in a targeted manner, but makes full use of the images collected by the vehicle in normal times to build a distributed database, so that when it is necessary to obtain image cases for a specific scene, it can directly use the form of searching the image in a parallel manner in multiple databases to obtain accurate images, which reduces the cost of data acquisition and improves the speed and accuracy of data acquisition.

[0061] A distributed database is constructed using sample images collected by the vehicle during driving, so that for a specific target scene, a distributed search is performed on the database to obtain a target matching image that is similar to the image to be searched of the target scene by searching by image. This application does not need to collect new data for the target scene, and makes full use of the images collected by the vehicle in normal times to build a distributed database. When it is necessary to obtain image cases for a specific scene, accurate images can be directly acquired in parallel in multiple databases by searching by image, which reduces the cost of data acquisition and improves the speed and accuracy of data acquisition.

[0062] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0064] Figure 1 A schematic diagram of a process of an image search method provided in an embodiment of the present application is shown;

[0065] Figure 2 A schematic diagram showing a flow chart of another image search method provided in an embodiment of the present application is shown;

[0066] Figure 3 A schematic diagram showing a flow chart of another image search method provided in an embodiment of the present application is shown;

[0067] Figure 4 A schematic diagram showing a flow chart of another image search method provided in an embodiment of the present application is shown;

[0068] Figure 5 A schematic diagram of the structure of an image search device provided in an embodiment of the present application is shown;

[0069] Figure 6 A schematic diagram of the device structure of a computer device provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0070] The present application will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict.

[0071] In this embodiment, an image search method is provided, such as Figure 1 As shown, the method includes:

[0072] Step 101, constructing a distributed database using sample images collected during the driving process of the vehicle.

[0073] In the embodiment of the present application, the vehicle will collect data through sensors during the automatic driving process, such as collecting visual image data through a camera, collecting radar data through a radar sensor and using the radar data to generate point cloud image data. These collected data can be stored in a database so that when sample data needs to be obtained for a specific scene in the future, the historical collection data of the vehicle can be fully utilized, so that the required data can be directly found in the historical collection data during subsequent data acquisition, without the need to collect new data in a targeted manner, thereby reducing the cost of data acquisition and shortening the data acquisition cycle. In addition, in order to improve the subsequent data acquisition speed, the present application will store sample images in a distributed manner, build multiple distributed databases for general recall needs, so that when data searches are performed subsequently, distributed searches can be performed in multiple databases to improve search efficiency.

[0074] Optionally, before step 101, it also includes: obtaining a full amount of sample images collected in real time during the driving process of the vehicle; performing image screening on the full amount of sample images according to a preset sample screening ratio to obtain the sample images, wherein the preset sample screening ratio is used to constrain the number of images retained within a preset time length.

[0075] In this embodiment, in order to control the search cost, after obtaining the full amount of sample images collected, the full amount of sample images can be screened to a certain extent, and some of the images can be retained to control the scale of the sample images. Specifically, multiple sample images acquired in a continuous time are generally similar, and multiple sample images acquired in a short time can be screened, and only some of the images can be retained and stored in the database. For example, according to the ratio of 1 frame / second, one of the same type of images acquired by the same vehicle within 1 second is randomly retained as a sample image, and the image types can include visual images and point cloud images.

[0076] Step 102: receiving an image to be searched for a target scene, and extracting a first image feature corresponding to the image to be searched by using a preset feature extraction model.

[0077] Step 103: extracting the second image feature of each sample image in the distributed database through the feature extraction model.

[0078] Step 104: acquiring a target matching image corresponding to the target scene in the sample image based on the similarity between the first image feature and each second image feature.

[0079] In an embodiment of the present application, an image search method can be adopted to perform image search in a distributed database. When obtaining sample images for a target scene, a case image for the target scene can be first found as the image to be searched, and the distributed database can be synchronously searched for images using the image to be searched, thereby saving search time and improving search efficiency. Specifically, image features can be extracted from the image to be searched using a preset feature extraction model, and the features of the image to be searched can be vectorized to obtain a first image feature, and the feature extraction model can be used to extract image features from sample images in the distributed database to obtain a second image feature of each sample image. Finally, the similarity between the first image feature and each second image feature is calculated, and a sample image with a similarity higher than a preset value or several sample images with a high similarity are obtained as the final target matching image.

[0080] By applying the technical solution of this embodiment, a distributed database is constructed using sample images collected by the vehicle during driving, so that for a specific target scene, a distributed search is performed on the database to obtain a target matching image that is similar to the image to be searched of the target scene by searching by image. This application does not need to collect new data for the target scene, and makes full use of the images collected by the vehicle at ordinary times to build a distributed database, so that when it is necessary to obtain image cases for a specific scene, it can directly use the form of searching by image in multiple databases to perform parallel accurate image acquisition, which reduces the cost of data acquisition and improves the speed and accuracy of data acquisition.

[0081] Further, as a refinement and extension of the specific implementation of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another image search method is provided, such as Figure 2 As shown, the method includes:

[0082] Step 201, obtain sample images collected by the vehicle during driving and sample annotation information of the sample images, and determine at least one storage database corresponding to the sample images based on the sample annotation information and database indexes corresponding to multiple distributed databases.

[0083] Step 202: store the sample image in the corresponding database.

[0084] In the embodiment of the present application, when the vehicle collects sample images during driving, it also collects sample annotation information of the sample images. The sample annotation information may include the collection time of the sample images, the vehicle's automatic driving scene, etc. Therefore, when storing the sample images in the database, it is possible to first determine which database the sample images should be stored in based on the sample annotation information of the sample images, that is, determine the database where the sample images should be stored. Specifically, different distributed databases can be established, and database indexes can be configured for each database, such as Figure 3 As shown, the database index may include the acquisition time corresponding to the stored sample, the autonomous driving scene corresponding to the sample, etc. The database matching the sample annotation information is used as the database of the sample image.

[0085] Optionally, step 201 includes: when the sample image is a visual image, based on the visual sample annotation information of the visual image and the visual database indexes corresponding to each of the multiple visual databases, determining at least one stored point cloud database corresponding to the visual image, wherein the distributed database includes the visual database; when the sample image is a point cloud image, based on the point cloud sample annotation information of the point cloud image and the point cloud database indexes corresponding to each of the multiple point cloud databases, determining at least one stored point cloud database corresponding to the point cloud image, wherein the distributed database includes the point cloud database.

[0086] Among them, the visual sample annotation information includes at least one of the camera position, data collection time and data collection scene, and the visual database index includes at least one of the camera position, data collection time and data collection scene; the point cloud sample annotation information includes the data collection time and / or data collection scene, and the point cloud database index includes the data collection time and / or data collection scene.

[0087] In the above embodiment, the sample image may include a visual image and a point cloud image, and the distributed database may include a visual database and a point cloud database. The visual image is stored in the visual database, and the point cloud image is stored in the point cloud database. The visual sample annotation information corresponding to the visual image may include one or more of the camera acquisition position, data acquisition time, and data acquisition scene. The visual database index corresponding to the visual database generally only includes one of the camera position, data acquisition time, and data acquisition scene. A visual image may correspond to multiple stored visual databases. The point cloud sample annotation information corresponding to the point cloud image may include one or more of the data acquisition time and data acquisition scene. The visual database index corresponding to the point cloud database generally only includes one of the data acquisition time and data acquisition scene. A point cloud image may correspond to multiple stored point cloud databases. For example, if the annotation information of the visual sample image includes camera position 1, HNP (High Way Navigation Pilot) and January 12, then it can be determined that the stored visual database of the visual sample image includes the camera position 1 database, the HNP database, and the January database. For another example, if the annotation information of the point cloud sample image includes UNP (Urban Navigation Assisted Pilot) and February 12, then it can be determined that the database of the sample image includes the UNP database and the February database.

[0088] Step 203: receiving an image to be searched for a target scene, and extracting a first image feature corresponding to the image to be searched by using a preset feature extraction model.

[0089] Step 204: determining a target search database from a plurality of distributed databases according to the search image annotation information of the image to be searched.

[0090] Step 205: extract the second image features of each sample image in the target search database respectively through the feature extraction model.

[0091] Step 206: Based on the similarity between the first image feature and each second image feature, a target matching image corresponding to the target scene is obtained in the sample image.

[0092] In an embodiment of the present application, after receiving the image to be searched, the first image feature corresponding to the image to be searched is extracted using a feature extraction model, and the target search database is found based on the search image annotation information corresponding to the image to be searched and the database index of each distributed database, so as to search in the target search database, narrow the data search range, and improve the data search accuracy. For example, if the search image annotation information is a NHP scene, the database with a database index of HNP can be used as the target search database. Then, a distributed search is performed on each target search database, and the second image feature of the sample image in each target search database is extracted using a feature extraction model. Finally, based on the similarity between the first image feature and the second image feature, the target matching image is determined.

[0093] In addition, to further improve the search efficiency, the sample images in the database may be reduced in dimension to reduce the space occupied by the sample images. The present application embodiment provides another image search method, such as Figure 4 As shown, the method includes:

[0094] Step 301, performing dimensionality reduction processing on sample images collected during the driving process of the vehicle, and using the sample images after dimensionality reduction processing to build a distributed database;

[0095] Step 302, receiving an image to be searched for a target scene, and extracting a first image feature corresponding to the image to be searched by using a preset feature extraction model;

[0096] Step 303, performing dimensionality reduction processing on the image to be searched to obtain a reduced-dimensionality image, and extracting a first image feature corresponding to the reduced-dimensionality image through a feature extraction model;

[0097] Step 304, extracting the second image feature of each sample image in the distributed database through the feature extraction model;

[0098] Step 305, based on the similarity between the first image feature and each second image feature, obtaining a candidate matching image corresponding to the target scene in the sample image;

[0099] Step 306, extracting a third image feature corresponding to the image to be searched by using the feature extraction model;

[0100] Step 307, performing dimension restoration on the candidate matching image, and extracting a fourth image feature of the candidate matching image after dimension restoration by using the feature extraction model;

[0101] Step 308: Based on the similarity between the third image feature and the fourth image feature, a target matching image corresponding to the target scene is obtained from the candidate matching images after dimensionality restoration.

[0102] In the above embodiment, the sample image can be compressed by dimensionality reduction and then stored in the database, so that when performing image search, the image to be searched is first compressed by dimensionality reduction, and then the image to be searched after dimensionality reduction is subjected to feature extraction to obtain the first image feature. After the target search database is determined according to the image annotation information of the image to be searched, the sample image in the target search database is subjected to feature extraction to obtain the second image feature, so that the sample image is preliminarily screened according to the similarity between the first image feature and the second image feature to obtain the candidate matching image. Since the dimension of the image is compressed, the calculation speed will be improved when the image after dimensionality reduction and compression is extracted. For example, when 50-dimensional data is compressed into 10 dimensions, the feature extraction speed will be doubled, and a small number of candidate samples will be screened from a large number of database samples. After obtaining the candidate matching image, the candidate matching image is further screened, first the candidate matching image is restored to the original image dimension, and then feature extraction is performed to obtain the fourth image feature, and the fourth image feature of the initial image to be searched is extracted, and finally the candidate matching image is screened according to the similarity between the third image feature and each fourth image feature to obtain the final target matching image.

[0103] It should be noted that when the amount of data in the database exceeds a certain threshold, the sample images in the database can be stored in reduced dimensions to improve the search speed, while when the amount of data in the database is small, the original sample storage is maintained.

[0104] Further, as Figure 1 The specific implementation of the method, the embodiment of the present application provides an image search device, such as Figure 5 As shown, the device comprises:

[0105] A database construction module is used to construct a distributed database using sample images collected by the vehicle during driving;

[0106] A feature extraction module, configured to receive an image to be searched for a target scene, and extract a first image feature corresponding to the image to be searched through a preset feature extraction model; and extract a second image feature of each sample image in the distributed database through the feature extraction model;

[0107] An image acquisition module is used to acquire a target matching image corresponding to the target scene in the sample image based on the similarity between the first image feature and each second image feature.

[0108] Optionally, the database construction module is further used to:

[0109] Acquire sample images collected by the vehicle during driving and sample annotation information of the sample images, and determine at least one storage database corresponding to the sample images based on the sample annotation information and database indexes corresponding to each of a plurality of distributed databases;

[0110] The sample images are stored in the corresponding database.

[0111] Optionally, the database construction module is further used to:

[0112] When the sample image is a visual image, based on the visual sample annotation information of the visual image and the visual database indexes corresponding to the plurality of visual databases, determining at least one visual database corresponding to the visual image, wherein the distributed database includes the visual database;

[0113] When the sample image is a point cloud image, at least one point cloud database corresponding to the point cloud image is determined based on the point cloud sample annotation information of the point cloud image and the point cloud database indexes corresponding to each of the multiple point cloud databases, wherein the distributed database includes the point cloud database.

[0114] Optionally, the visual sample annotation information includes at least one of a camera position, a data collection time, and a data collection scene, and the visual database index includes at least one of a camera position, a data collection time, and a data collection scene; the point cloud sample annotation information includes the data collection time and / or the data collection scene, and the point cloud database index includes the data collection time and / or the data collection scene.

[0115] Optionally, the feature extraction module is further used to:

[0116] Determining a target search database from a plurality of distributed databases according to the search image annotation information of the image to be searched;

[0117] The second image features of each sample image in the target search database are extracted respectively through the feature extraction model.

[0118] Optionally, the database construction module is further used to:

[0119] Perform dimensionality reduction processing on sample images collected during vehicle driving, and use the sample images after dimensionality reduction processing to build a distributed database;

[0120] Accordingly, the feature extraction module is also used for:

[0121] A dimensionality reduction process is performed on the image to be searched to obtain a reduced-dimensionality image, and a first image feature corresponding to the reduced-dimensionality image is extracted through a feature extraction model.

[0122] Optionally, the image acquisition module is further used to:

[0123] Based on the similarity between the first image feature and each second image feature, obtaining a candidate matching image corresponding to the target scene in the sample image;

[0124] Extracting a third image feature corresponding to the image to be searched by using the feature extraction model;

[0125] Performing dimension restoration on the candidate matching image, and extracting a fourth image feature of the candidate matching image after dimension restoration by using the feature extraction model;

[0126] Based on the similarity between the third image feature and the fourth image feature, a target matching image corresponding to the target scene is obtained from the candidate matching images after dimensionality restoration.

[0127] Optionally, the database construction module is further used to:

[0128] Obtain the full amount of sample images collected in real time while the vehicle is driving;

[0129] The sample images are obtained by performing image screening on the full amount of sample images according to a preset sample screening ratio, wherein the preset sample screening ratio is used to constrain the number of images retained within a preset time length.

[0130] It should be noted that for other corresponding descriptions of the functional units involved in the image search device provided in the embodiment of the present application, reference can be made to Figures 1 to 4 The corresponding description in the method will not be repeated here.

[0131] The present application also provides a computer device, which may be a personal computer, a server, a network device, etc. Figure 6 As shown, the computer device includes a bus, a processor, a memory and a communication interface, and may also include an input and output interface and a display device. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store location information. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the steps in each method embodiment are implemented.

[0132] Those skilled in the art will understand that Figure 6The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0133] In one embodiment, a computer-readable storage medium is provided. The computer-readable storage medium may be non-volatile or volatile, and stores a computer program thereon. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0134] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0135] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0136] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0137] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0138] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. An image search method, characterized in that: The method comprises: Using sample images collected by vehicles during driving, a distributed database is constructed; Receiving an image to be searched for a target scene, and extracting a first image feature corresponding to the image to be searched through a preset feature extraction model; Extracting the second image feature of each sample image in the distributed database through the feature extraction model; Based on the similarity between the first image feature and each second image feature, a target matching image corresponding to the target scene is obtained in the sample image.

2. The method according to claim 1, characterized in that The method of constructing a distributed database using sample images collected by the vehicle during driving includes: Acquire sample images collected by the vehicle during driving and sample annotation information of the sample images, and determine at least one storage database corresponding to the sample images based on the sample annotation information and database indexes corresponding to each of a plurality of distributed databases; The sample images are stored in the corresponding database.

3. The method according to claim 2, characterized in that The determining, based on the sample annotation information and database indexes corresponding to the plurality of distributed databases, at least one database corresponding to the sample image comprises: When the sample image is a visual image, based on the visual sample annotation information of the visual image and the visual database indexes corresponding to the plurality of visual databases, determining at least one visual database corresponding to the visual image, wherein the distributed database includes the visual database; When the sample image is a point cloud image, at least one point cloud database corresponding to the point cloud image is determined based on the point cloud sample annotation information of the point cloud image and the point cloud database indexes corresponding to each of the multiple point cloud databases, wherein the distributed database includes the point cloud database.

4. The method according to claim 3, characterized in that The visual sample annotation information includes at least one of a camera position, a data collection time, and a data collection scene, and the visual database index includes at least one of a camera position, a data collection time, and a data collection scene; The point cloud sample annotation information includes data collection time and / or data collection scene, and the point cloud database index includes data collection time and / or data collection scene.

5. The method according to claim 2, characterized in that: The extracting the second image features of each sample image in the distributed database by the feature extraction model respectively includes: Determining a target search database from a plurality of distributed databases according to the search image annotation information of the image to be searched; The second image features of each sample image in the target search database are extracted respectively through the feature extraction model.

6. The method according to any one of claims 1 to 5, characterized in that The method of constructing a distributed database using sample images collected by the vehicle during driving includes: Perform dimensionality reduction processing on sample images collected during vehicle driving, and use the sample images after dimensionality reduction processing to build a distributed database; Correspondingly, extracting the first image feature corresponding to the image to be searched by using a preset feature extraction model includes: A dimensionality reduction process is performed on the image to be searched to obtain a reduced-dimensionality image, and a first image feature corresponding to the reduced-dimensionality image is extracted through a feature extraction model.

7. The method according to claim 6, characterized in that The acquiring, in the sample image, a target matching image corresponding to the target scene based on the similarity between the first image feature and each second image feature, comprises: Based on the similarity between the first image feature and each second image feature, obtaining a candidate matching image corresponding to the target scene in the sample image; Extracting a third image feature corresponding to the image to be searched by using the feature extraction model; Performing dimension restoration on the candidate matching image, and extracting a fourth image feature of the candidate matching image after dimension restoration by using the feature extraction model; Based on the similarity between the third image feature and the fourth image feature, a target matching image corresponding to the target scene is obtained from the candidate matching images after dimensionality restoration.

8. The method according to any one of claims 1 to 5, characterized in that Before constructing the distributed database using sample images collected by the vehicle during driving, the method further includes: Obtain the full amount of sample images collected in real time while the vehicle is driving; The sample images are obtained by performing image screening on the full amount of sample images according to a preset sample screening ratio, wherein the preset sample screening ratio is used to constrain the number of images retained within a preset time length.

9. An image search device, characterized in that: The device comprises: A database construction module is used to construct a distributed database using sample images collected by the vehicle during driving; A feature extraction module, configured to receive an image to be searched for a target scene, and extract a first image feature corresponding to the image to be searched through a preset feature extraction model; and extract a second image feature of each sample image in the distributed database through the feature extraction model; The image acquisition module is used to acquire a target matching image corresponding to the target scene in the sample image based on the similarity between the first image feature and each second image feature.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

11. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images