Indirect image matching method and device, electronic equipment and storage medium

Abstract

The present application relates to an indirect image matching method, device, electronic equipment and storage medium, the first feature of the target image is extracted, the second feature and the third feature of the reference image are extracted, the first association of the target image and the reference image is established based on the first feature and the second feature; the fourth feature of the image to be matched is extracted, the second association of the reference image and the image to be matched is established based on the third feature and the fourth feature; the target image and the image to be matched are matched based on the first association and the second association; the first feature of the target image, the second feature and the third feature of the reference image are different from the fourth feature of the image to be matched extracted. The speed of image feature extraction and the accuracy of image matching are considered, and the demand of real-time processing of video stream can be met.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of augmented reality, and particularly to indirect image matching methods, apparatuses, electronic devices, and storage media. Background Technology

[0002] Augmented Reality (AR) is an emerging technology that combines the digital and real worlds, aiming to integrate virtual environments into the real world to enhance "reality." Typically, an AR system involves the following steps: video image acquisition, virtual image generation, and video synthesis and output. Before video synthesis, it's usually necessary to identify the specific content captured by the camera in a partial image. This involves quickly matching the partial image with the complete target image to pinpoint the location of the captured content within the target image. Feature-based image matching algorithms provide the technological foundation for rapid video recognition in AR systems.

[0003] Currently, commonly used algorithms for planar image matching include Scale-invariant Feature Transform (SIFT), oriented FAST, and rotation-invariant BRIEF (Oriented FAST and Rotated BRIEF, ORB) feature matching. These algorithms detect image features and calculate the transformation matrix between images based on these features. SIFT is a very stable feature, and while it can achieve relatively accurate feature matching, its feature extraction and description are quite complex. Furthermore, the matching of feature vectors uses Euclidean distance calculations between floating-point numbers, making the algorithm time-consuming and difficult to meet the real-time processing requirements of video images in AR systems.

[0004] To improve processing speed, the ORB algorithm is widely used. The ORB algorithm can quickly remove features in advance. Although the algorithm is time-saving, the matching effect is unstable. It is usually only suitable for matching between consecutive images in a video. For matching images with large differences, such as matching between local images and the whole image, or matching images with large changes in camera pose, the ORB algorithm has poor matching effect. Its image matching effect is difficult to meet the requirements of stable processing of real-time video streams.

[0005] To balance image matching accuracy and video processing speed, existing technologies employ image matching methods that fuse SIFT and ORB algorithms. However, these methods all use both SIFT and ORB algorithms to process the same image to improve feature extraction quality and speed. In other words, SIFT and ORB algorithms process the same object, complementing each other. These methods require using multiple different algorithms for feature extraction on the same image and then fusing the features extracted by different algorithms, increasing the system's image processing time. Therefore, there is an urgent need in the existing technology for an image matching method, device, electronic device, and storage medium that can stably, accurately, and quickly achieve video image matching. Summary of the Invention

[0006] The purpose of this invention is to solve the technical problem that image matching in the prior art cannot simultaneously take into account matching speed and matching accuracy, and cannot meet the requirements of accurate matching in real-time video stream processing. It provides an indirect image matching method, device, electronic device and storage medium that can meet the real-time requirements of video image matching, and is fast and accurate.

[0007] To address the aforementioned technical problems, embodiments of the present invention provide an indirect image matching method, comprising:

[0008] Extract a first feature from the target image, extract a second and a third feature from the reference image, and establish a first association between the target image and the reference image based on the first and second features;

[0009] Extract the fourth feature of the image to be matched, and establish a second association between the reference image and the image to be matched based on the third feature and the fourth feature;

[0010] Based on the first association and the second association, the target image and the image to be matched are matched;

[0011] The extraction of the first feature of the target image, the second feature of the reference image, and the third feature is out of sync with the extraction of the fourth feature of the image to be matched.

[0012] Preferably, the first feature operator for extracting the first feature of the target image and the second feature of the reference image, and the second feature operator for extracting the third feature of the reference image and the fourth feature of the image to be matched, are different feature extraction operators. The feature extraction speed of the second feature operator is greater than that of the first feature operator, and the feature extraction accuracy of the first feature operator is higher than that of the second feature operator.

[0013] Preferably, the first feature operator is SIFT and the second feature operator is ORB.

[0014] Preferably, the first association relationship and the second association relationship are transformation matrices, and matching the target image and the image to be matched based on the first association relationship and the second association relationship specifically includes:

[0015] The transformation matrix from the image to be matched to the target image is obtained based on the first transformation matrix and the second transformation matrix, and the position of the image to be matched on the target image is located.

[0016] Preferably, the image to be matched is a local image captured in real time by a camera, the target image is a global image, and the transformation matrix from the image to be matched to the target image is H = H1 × H2, where H1 is the transformation matrix between the target image and the reference image, H2 is the transformation matrix between the reference image and the image to be matched, and the first transformation matrix and the second transformation matrix are homography matrices.

[0017] Preferably, the reference image is acquired according to a first cycle, and the second feature and the third feature of the reference image are calculated using a first feature operator and a second feature operator, and a first transformation matrix is ​​calculated;

[0018] The image to be matched is acquired according to the second cycle, the fourth feature of the image to be matched is calculated using the second feature operator, and the second transformation matrix is ​​calculated based on the third feature and the fourth feature.

[0019] In the second period, the image to be matched and the target image are matched according to the first transformation matrix and the second transformation matrix, and the time length of the first period is n times that of the second period.

[0020] The present invention also provides an indirect image matching device, comprising:

[0021] A backup extraction module is used to extract a first feature of the target image, extract a second and a third feature of the reference image, and establish a first association between the target image and the reference image based on the first and second features.

[0022] A real-time extraction module is used to extract a fourth feature of the image to be matched, and to establish a second association between the reference image and the image to be matched based on the third feature and the fourth feature;

[0023] The extraction of the first feature of the target image, the second feature of the reference image, and the third feature is asynchronous with the extraction of the fourth feature of the image to be matched.

[0024] A matching module is used to match the target image with the image to be matched based on the first association and the second association.

[0025] Preferably, the backup extraction module is configured to acquire the reference image according to a first cycle, calculate the second feature and the third feature of the reference image using a first feature operator and a second feature operator, and calculate a first transformation matrix; the real-time extraction module is configured to acquire the image to be matched according to a second cycle, calculate the fourth feature of the image to be matched using a second feature operator, calculate a second transformation matrix based on the third feature and the fourth feature, and the matching module is further configured to match the image to be matched and the target image based on the first transformation matrix and the second transformation matrix, wherein the time length of the first cycle is n times that of the second cycle.

[0026] This invention provides an electronic device, comprising:

[0027] The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the aforementioned indirect image matching method.

[0028] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the aforementioned indirect image matching method.

[0029] Compared with the prior art, the indirect image matching method, apparatus, electronic device, and storage medium provided by this invention utilize a reference image as an intermediate medium and adopt a dual-thread independent working mode. One thread performs feature extraction of the reference image and the target image in advance, saving online real-time processing time while ensuring the accuracy of feature extraction. The other thread performs online real-time processing, extracting features and performing matching calculations on the image to be matched. The entire image matching time is the calculation time of the second thread, shortening the time for feature extraction and matching calculations, thereby improving the speed of image matching. While optimizing the image matching time, it can also ensure the accuracy of image matching, balancing speed and accuracy, and achieving real-time processing of video streams.

[0030] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0031] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.

[0032] Figure 1 A flowchart of the indirect image matching method provided in an embodiment of the present invention;

[0033] Figure 2 A flowchart of an indirect image matching method provided in another embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of the structure of the indirect image matching device provided in an embodiment of the present invention. Detailed Implementation

[0035] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the various embodiments of the present invention will be described in detail below with reference to the accompanying drawings. However, those skilled in the art will understand that many technical details are presented in the various embodiments of the present invention to facilitate a better understanding of this application. However, the technical solutions claimed in this application can be implemented even without these technical details and various changes and modifications based on the following embodiments. The division of the various embodiments below is for ease of description and should not constitute any limitation on the specific implementation of the present invention. The various embodiments can be combined with and referenced by each other without contradiction.

[0036] The first embodiment of the present invention relates to an indirect image matching method, such as... Figure 1 As shown, the indirect image matching method specifically includes: extracting a first feature of a target image, extracting a second and a third feature of a reference image, and establishing a first association between the target image and the reference image based on the first and second features; extracting a fourth feature of the image to be matched, and establishing a second association between the reference image and the image to be matched based on the third and fourth features; matching the target image and the image to be matched based on the first and second associations; the extraction of the first feature of the target image, the second feature of the reference image, and the third feature is asynchronous with the extraction of the fourth feature of the image to be matched.

[0037] Compared to methods that use SIFT and ORB feature operators to jointly extract features from the image to be matched and then match it with the target image, the indirect image matching method provided by this invention uses a reference image as an intermediate medium and employs a dual-threaded approach. One thread extracts features from the target image and the intermediate medium image and calculates a first association, i.e., a first transformation matrix, which is kept in reserve. The other thread extracts features from the image to be matched and calculates a second association, i.e., a second transformation matrix, based on the features of the latest acquired reference image kept in reserve. Combining this second association with the already calculated first transformation matrix determines the matching relationship between the image to be matched and the target image. Throughout the matching process, the reference image serves as an intermediate transformation medium. The first association and the features of the reference image are pre-calculated and kept in reserve. These are performed simultaneously with or before the second thread. Therefore, the entire image matching time is the computation time of the second thread, i.e., the feature extraction time of the image to be matched, thus shortening the feature extraction and matching calculation time and improving the speed of image matching. Furthermore, the first thread is independent and pre-computed, and the first association has sufficient computation time to achieve accurate calculation of the target image. While optimizing the image matching time, it can also ensure the accuracy of image matching, balancing speed and accuracy, and can achieve real-time processing of video streams.

[0038] Specifically, the first feature of the target image is obtained using a first feature operator, the second and third features of the latest reference image are obtained using the first and second feature operators respectively, and the fourth feature of the image to be matched is obtained using the second feature operator. The first feature operator used to extract the first feature of the target image and the second feature of the reference image, and the second feature operator used to extract the third feature of the reference image and the fourth feature of the image to be matched are different feature extraction operators. The feature extraction speed of the second feature operator is greater than that of the first feature operator, and the feature extraction accuracy of the first feature operator is higher than that of the second feature operator. As an optional embodiment, the first feature operator is SIFT and the second feature operator is ORB. This invention utilizes two feature operators with different extraction speeds and accuracies for feature extraction. In two independent threads, the first thread, which can be prepared in advance, uses the high-precision SIFT for high-precision feature extraction, ensuring accurate image matching and pre-extracting features from a reference image for later use. The second thread, due to higher real-time requirements, employs the faster ORB for real-time feature extraction of the image to be matched, further improving the image matching speed compared to the first thread. By using two feature operators and two independent threads, the image matching speed is accelerated while achieving better matching results than using the second feature operator alone.

[0039] The first association and the second association are transformation matrices, respectively. Matching the target image and the image to be matched based on the first and second associations specifically includes: obtaining the transformation matrix from the image to be matched to the target image according to the first and second transformation matrices, and locating the position of the image to be matched on the target image. Wherein, the image to be matched is a local image captured in real-time by a camera, and the target image is a pre-prepared global image. The transformation matrix from the image to be matched to the target image is H = H1 × H2, where H1 is the transformation matrix between the target image and the latest reference image, and H2 is the transformation matrix between the latest reference image and the image to be matched. The first and second transformation matrices are homography matrices.

[0040] The image matching method provided by this invention utilizes an intermediate reference image as a medium to quickly achieve matching between local and global images. It rapidly and accurately locates the position of the local image to be matched within the global image, completing video recognition and thus improving the processing speed of video processing and AR systems. Compared to existing technologies that use SIFT matching (which is effective but slow) or ORB features (which are fast but ineffective), this invention provides an indirect image matching method that improves matching speed to match ORB feature matching while enhancing matching effectiveness, making it superior to ORB feature matching.

[0041] Further, the reference image is acquired according to a first cycle, and the second and third features of the reference image are calculated using the first and second feature operators to calculate a first transformation matrix. The image to be matched is acquired according to a second cycle, and the fourth feature of the image to be matched is calculated using the second feature operator. A second transformation matrix is ​​calculated based on the third and fourth features. Within the second cycle, the image to be matched and the target image are matched based on the first and second transformation matrices. The duration of the first cycle is n times the duration of the second cycle, where n is a positive number. As an optional embodiment, at each first cycle, the latest reference image is acquired from the video stream, the second and third features are calculated, and the first transformation matrix is ​​calculated, thus reserving the first transformation matrix and the third feature for later use. At each second cycle, the fourth feature is calculated, and the latest third feature is read to calculate the second transformation matrix. The second transformation matrix is ​​multiplied by the latest first transformation matrix to achieve image matching. The reference image and the image to be matched differ by an integer number of frames.

[0042] The image matching method provided by this invention has two threads, such as... Figure 2 As shown, after calculating the SIFT features of the target image, thread 1 and thread 2 read images from the video stream as the latest reference image and the image to be matched, respectively. Thread 1 needs to calculate the SIFT and ORB features of the reference image, and use the latest SIFT features of the reference image to match the target image, calculating the latest homography matrix from the reference image to the target image. Then, the result is saved as the pre-prepared reference image features and the homography matrix of the image transformation. Thread 2 calculates the ORB features of the current frame, and uses the ORB features to match the pre-prepared reference image ORB features to obtain the homography matrix from the image to be matched to the reference image. Then, the homography matrix from the image to be matched to the reference image is multiplied by the prepared homography matrix from the reference image to the target image to indirectly obtain the homography matrix from the image to be matched to the target image.

[0043] Specifically, thread 1 performs the following tasks: on the one hand, it detects SIFT features of the target image to obtain SIFT keypoints and corresponding descriptors; on the other hand, it uses the input reference image to perform SIFT feature detection and ORB feature detection to obtain SIFT keypoints and corresponding descriptors of the reference image and ORB keypoints and corresponding descriptors of the reference image. Furthermore, thread 1 matches SIFT keypoints based on the target image SIFT keypoints and corresponding descriptors and the reference image SIFT keypoints and corresponding descriptors, filters outliers, and calculates the homography matrix from the reference image to the target image. Then, it updates the reference image information using the reference image ORB keypoints and corresponding descriptors and the homography matrix from the reference image to the target image.

[0044] Feature point matching results may contain outliers, meaning some feature points may deviate from the matched feature points due to environmental influences. The presence of outliers can introduce errors into subsequent calculations. Filtering outliers makes feature matching more accurate.

[0045] Thread 2 performs the following tasks: It uses the input image to be matched to perform ORB feature detection, obtaining ORB keypoints and corresponding descriptors of the image to be matched; it uses the prepared reference image ORB features and the ORB keypoints and corresponding descriptors of the image to be matched to match ORB keypoints, and after filtering out outliers, it calculates the homography matrix from the image to the reference image. Then, based on the homography matrix from the image to the reference image and the prepared homography matrix from the reference image to the target image, it calculates the homography matrix from the image to the target image, completing the image matching.

[0046] Thread 1 and Thread 2 loop independently. Thread 2 requires Thread 1 to complete at least one loop before it begins. Thread 1 uses SIFT for feature extraction calculations, therefore its processing time is several times that of Thread 2, let's say N times (N < 10). Before the feature calculation of the new reference image is completed, the image to be matched is matched against the previous calculation result of the reference image. The maximum difference between the image to be matched and the reference image used for ORB feature matching is 2N-1 frames. Under normal camera conditions without severe shaking, the image to be matched and the reference image have a large overlap area, allowing ORB feature matching to quickly obtain relatively accurate results. In practical applications, the processing time per frame is the same as the loop time of Thread 2, comparable to directly using the ORB matching algorithm, but with better performance.

[0047] Embodiment 2 of the present invention provides an indirect image matching device, such as Figure 3 As shown, the indirect image matching device includes:

[0048] A backup extraction module is used to extract a first feature of the target image, extract a second and a third feature of the reference image, and establish a first association between the target image and the reference image based on the first and second features.

[0049] A real-time extraction module is used to extract a fourth feature of the image to be matched, and to establish a second association between the reference image and the image to be matched based on the third feature and the fourth feature;

[0050] The extraction of the first feature of the target image, the second feature of the reference image, and the third feature is asynchronous with the extraction of the fourth feature of the image to be matched.

[0051] A matching module is used to match the target image with the image to be matched based on the first association and the second association.

[0052] Specifically, a first feature of the target image is obtained using a first feature operator. The second and third features of the latest reference image are obtained using both the first and second feature operators. A fourth feature of the image to be matched is obtained using the second feature operator. The first feature operator used to extract the first feature of the target image and the second feature of the reference image, and the second feature operator used to extract the third feature of the reference image and the fourth feature of the image to be matched, are different feature extraction operators. The feature extraction speed of the second feature operator is greater than that of the first feature operator, and the feature extraction accuracy of the first feature operator is higher than that of the second feature operator. As an optional embodiment, the first feature operator is SIFT, and the second feature operator is ORB.

[0053] Furthermore, the backup extraction module is configured to acquire the reference image according to a first cycle, calculate the second and third features of the reference image using the first and second feature operators, and calculate the first transformation matrix; the real-time extraction module is configured to acquire the image to be matched according to a second cycle, calculate the fourth feature of the image to be matched using the second feature operator, and calculate the second transformation matrix based on the third and fourth features; the matching module is further configured to match the image to be matched and the target image based on the first and second transformation matrices, wherein the duration of the first cycle is n times that of the second cycle.

[0054] It is not difficult to see that this embodiment is a device embodiment corresponding to the first embodiment, and this embodiment can be implemented in conjunction with the first embodiment. The relevant technical details mentioned in the first embodiment are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the first embodiment.

[0055] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this invention, this embodiment does not introduce units that are not closely related to solving the technical problem proposed by this invention; however, this does not mean that other units are absent from this embodiment.

[0056] A third embodiment of the present invention also provides an indirect image matching electronic device, comprising: a memory, a processor, and a computer program for an indirect image matching method stored in the memory and executable on the processor, wherein:

[0057] The processor invokes a computer program stored in memory to perform the following steps: extracting a first feature of a target image, extracting a second and a third feature of a reference image, and establishing a first association between the target image and the reference image based on the first and second features; extracting a fourth feature of the image to be matched, and establishing a second association between the reference image and the image to be matched based on the third and fourth features; matching the target image and the image to be matched based on the first and second associations; the extraction of the first feature of the target image, the second feature of the reference image, and the third feature is asynchronous with the extraction of the fourth feature of the image to be matched.

[0058] Specifically, a first feature of the target image is obtained using a first feature operator. A second and third feature of the latest reference image are obtained using both the first and second feature operators. A fourth feature of the image to be matched is obtained using the second feature operator. The first feature operator used to extract the first feature of the target image and the second feature of the reference image, and the second feature operator used to extract the third feature of the reference image and the fourth feature of the image to be matched, are different feature extraction operators. The feature extraction speed of the second feature operator is greater than that of the first feature operator, and the feature extraction accuracy of the first feature operator is higher than that of the second feature operator. As an optional embodiment, the first feature operator is SIFT, and the second feature operator is ORB.

[0059] The first and second association relationships are transformation matrices, respectively. Matching the target image and the image to be matched based on the first and second association relationships specifically includes: obtaining the transformation matrix from the image to be matched to the target image according to the first and second transformation matrices, and locating the position of the image to be matched on the target image. Wherein, the image to be matched is a local image captured in real-time by a camera, and the target image is a pre-prepared global image. The transformation matrix from the image to be matched to the target image is H = H1 × H2, where H1 is the transformation matrix between the target image and the latest reference image, and H2 is the transformation matrix between the latest reference image and the image to be matched. The first and second transformation matrices are homography matrices.

[0060] Furthermore, the reference image is acquired according to the first cycle, and the second and third features of the reference image are calculated using the first and second feature operators to calculate the first transformation matrix; the image to be matched is acquired according to the second cycle, and the fourth feature of the image to be matched is calculated using the second feature operator; the second transformation matrix is ​​calculated based on the third and fourth features; and the image to be matched and the target image are matched according to the first and second transformation matrices within the second cycle, wherein the time length of the first cycle is n times that of the second cycle.

[0061] The electronic device provided in this embodiment of the invention can be applied to devices such as smartphones, computers, and servers capable of indirect image matching. The electronic device provided in this embodiment of the invention can implement each process of the indirect image matching method described above, and can achieve the same beneficial effects. To avoid repetition, further details are omitted here.

[0062] The fourth embodiment of the present invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the indirect image matching method or the application-side indirect image matching method provided in the embodiments of the present invention, and can achieve the same technical effect. To avoid repetition, it will not be described again here.

[0063] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0064] Those skilled in the art will understand that the above embodiments are specific examples of implementing the present invention, and in practical applications, various changes in form and detail may be made without departing from the spirit and scope of the present invention.

Claims

1. An indirect image matching method, characterized in that, The indirect image matching method includes: Extract a first feature from the target image, extract a second and a third feature from the reference image, and establish a first association between the target image and the reference image based on the first and second features; Extract the fourth feature of the image to be matched, and establish a second association between the reference image and the image to be matched based on the third feature and the fourth feature; Based on the first association and the second association, the target image and the image to be matched are matched; The extraction of the first feature of the target image, the second feature and the third feature of the reference image are asynchronous with the extraction of the fourth feature of the image to be matched. Specifically, the reference image is acquired according to a first cycle, and the second and third features of the reference image are calculated using a first feature operator and a second feature operator to calculate a first transformation matrix. The image to be matched is acquired according to a second cycle, and the fourth feature of the image to be matched is calculated using the second feature operator. A second transformation matrix is ​​calculated based on the third and fourth features, and within the second cycle, the image to be matched and the target image are matched based on the first and second transformation matrices. The duration of the first cycle is n times the duration of the second cycle, where n is a positive number.

2. The indirect image matching method according to claim 1, characterized in that: The first feature operator for extracting the first feature of the target image and the second feature of the reference image, and the second feature operator for extracting the third feature of the reference image and the fourth feature of the image to be matched, are different feature extraction operators. The feature extraction speed of the second feature operator is greater than that of the first feature operator, and the feature extraction accuracy of the first feature operator is higher than that of the second feature operator.

3. The indirect image matching method according to claim 2, characterized in that: The first feature operator is SIFT, and the second feature operator is ORB.

4. The indirect image matching method according to claim 1, characterized in that, The first association relationship and the second association relationship are transformation matrices, respectively. Matching the target image and the image to be matched based on the first association relationship and the second association relationship specifically includes: The transformation matrix from the image to be matched to the target image is obtained based on the first transformation matrix and the second transformation matrix, and the position of the image to be matched on the target image is located.

5. The indirect image matching method according to claim 4, characterized in that, The image to be matched is a local image captured in real time by the camera, and the target image is a global image. The transformation matrix from the image to be matched to the target image is H = H1 × H2, where H1 is the transformation matrix between the target image and the reference image, and H2 is the transformation matrix between the reference image and the image to be matched. The first transformation matrix and the second transformation matrix are homography matrices.

6. An indirect image matching device, characterized in that, The indirect image matching device includes: A backup extraction module is used to extract a first feature of the target image, extract a second and a third feature of the reference image, and establish a first association between the target image and the reference image based on the first and second features. A real-time extraction module is used to extract a fourth feature of the image to be matched, and to establish a second association between the reference image and the image to be matched based on the third feature and the fourth feature; The extraction of the first feature of the target image, the second feature of the reference image, and the third feature is asynchronous with the extraction of the fourth feature of the image to be matched. A matching module is used to match the target image and the image to be matched based on the first association and the second association; wherein, the reference image is acquired according to a first period, the second feature and the third feature of the reference image are calculated using a first feature operator and a second feature operator, and a first transformation matrix is ​​calculated; the image to be matched is acquired according to a second period, the fourth feature of the image to be matched is calculated using the second feature operator, a second transformation matrix is ​​calculated based on the third feature and the fourth feature, and the image to be matched and the target image are matched according to the first transformation matrix and the second transformation matrix within the second period, wherein the time length of the first period is n times the second period, and n is a positive number.

7. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the indirect image matching method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the indirect image matching method as described in any one of claims 1 to 5.

Images