Underwater vision closed-loop detection method, device, equipment and medium

Abstract

The invention relates to the technical field of computers, and discloses an underwater vision closed-loop detection method, device and equipment and a medium, global feature information and local feature information of an obtained key frame image are extracted at the same time through a convolutional neural network, and the local feature information of the key frame image is extracted on the basis of the global feature information of the current key frame. Quantifying global similarity and time relevance between the current key frame and other key frames to construct a closed-loop database; meanwhile, on the basis of global feature information and inter-frame adjacency characteristics of the current key frame, a closed-loop candidate frame is searched from a closed-loop database, the closed-loop candidate frame is verified according to local feature information of the current key frame and local feature information of the closed-loop candidate frame, and when verification is passed, the current key frame is verified; taking the closed-loop candidate frame as a target closed-loop frame of the current key frame; the closed-loop detection process is constrained through image similarity, time relevance and inter-frame adjacency, and the underwater vision closed-loop precision and recall rate are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an underwater visual closed-loop detection method, device, equipment and medium. Background Art

[0002] With the continuous development of positioning technology, simultaneous localization and mapping (SLAM) technology has been widely used in underwater research.

[0003] Closed-loop detection is an important part of SLAM technology, which can effectively reduce positioning errors caused by trajectory drift. However, the single structure and insufficient texture of the underwater environment pose great challenges to closed-loop detection, resulting in low recall rate and poor positioning accuracy of underwater closed-loop detection.

[0004] It can be seen that how to solve the problem of low positioning accuracy and recall rate of underwater visual closed-loop detection has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] The present invention provides an underwater visual closed-loop detection method, device, equipment and medium to solve the problem of how to improve the positioning accuracy and recall rate of underwater visual closed-loop detection.

[0006] In order to solve the above technical problems, the first aspect of the present invention provides an underwater visual closed-loop detection method, comprising:

[0007] Acquire multiple key frames and their corresponding key frame images in sequence, extract global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially put the key frames into a queue of frames waiting to be inserted;

[0008] Acquire the current key frame in the queue of frames waiting to be inserted, and calculate the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted based on the global feature information of the current key frame, so as to construct a closed-loop database;

[0009] Based on the global feature information and inter-frame adjacency of the current key frame, searching for closed-loop candidate frames from the closed-loop database;

[0010] The closed-loop candidate frame is verified according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, the closed-loop candidate frame is used as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection.

[0011] As one of the preferred solutions, the step of sequentially acquiring multiple key frames and their corresponding key frame images, and extracting global feature information and local feature information of the key frame images through a convolutional neural network model includes:

[0012] Acquire multiple key frames and their corresponding key frame images in sequence from the front end of the SLAM system;

[0013] Extracting local features of the key frame image through the convolution layer of the convolution neural network model, and respectively extracting high-distinctive information and low-dimensional representation information of the local features through the attention layer of the convolution neural network model and the autoencoder of the convolution neural network model, to obtain local feature information of the key frame image;

[0014] The global features of the key frame image are extracted through the convolution layer of the convolutional neural network model, and the contribution of the global features is weighted through the generalized average pooling layer of the convolutional neural network model to obtain the global feature information of the key frame image.

[0015] As one of the preferred solutions, the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted are calculated based on the global feature information of the current key frame to construct a closed-loop database, including:

[0016] Calculate the global similarity and time correlation between the current key frame and the head key frame in the queue of frames waiting to be inserted according to the global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted;

[0017] In the determination phase, if the global similarity and the time correlation meet the first requirement determination or the second requirement determination, the head key frame is inserted into the closed-loop database and removed from the queue of frames waiting to be inserted, so as to update the head key frame in the queue of frames waiting to be inserted; the first requirement determination is that the global similarity meets the similarity requirement, and the second requirement determination is that the global similarity does not meet the similarity requirement and the time correlation meets the correlation requirement;

[0018] Based on the global feature information of the current key frame, the determination step is repeated until the number of frames waiting to be inserted into the frame queue reaches a preset number, and the closed-loop database is output.

[0019] As one of the preferred solutions, after calculating the global similarity and time correlation between the global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted, the method includes:

[0020] When the global similarity and the time correlation meet the third requirement, the current key frame is updated, and the global similarity and time correlation between the two are recalculated according to the updated global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted; the third requirement is that the global similarity does not meet the similarity requirement and the time correlation does not meet the correlation requirement;

[0021] The demand is determined based on the recalculated global similarity and time correlation, and the head key frame or the updated current key frame is updated based on the determination result. The determination step is repeated based on the updated result until all current key frames in the queue of frames waiting to be inserted are traversed and the closed-loop database is output.

[0022] As one of the preferred solutions, searching for closed-loop candidate frames from the closed-loop database based on the global feature information of the current key frame and inter-frame adjacency includes:

[0023] Respectively calculating the similarity between the global feature information of each closed-loop frame in the closed-loop database and the global feature information of the current key frame, and sorting them according to the similarity to construct a closed-loop frame queue;

[0024] Constructing a closed-loop frame sequence according to the current closed-loop frame, the first closed-loop frame and the second closed-loop frame in the closed-loop frame queue, and using the closed-loop frame sequence as the nearest neighbor of the current closed-loop frame; the first closed-loop frame and the second closed-loop frame are respectively the previous closed-loop frame and the next closed-loop frame of the current closed-loop frame;

[0025] Each closed-loop frame except the head closed-loop frame in the closed-loop frame queue is taken as a closed-loop frame to be verified and verified in sequence, and each closed-loop frame to be verified that meets the first verification condition or the second verification condition is taken as a preferred closed-loop frame, and the head closed-loop frame in the closed-loop frame sequence is taken as the closed-loop candidate frame of the current key frame; the first verification condition is that when the first closed-loop frame of the closed-loop frame to be verified belongs to the nearest neighbor of the head closed-loop frame, the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame; the second verification condition is that when the first closed-loop frame of the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame, the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame and the closed-loop frame to be verified does not belong to the nearest neighbor of the first closed-loop frame of the closed-loop frame to be verified.

[0026] As one of the preferred solutions, the closed-loop candidate frame is verified according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, the closed-loop candidate frame is used as the target closed-loop frame of the current key frame to realize underwater visual closed-loop detection, including:

[0027] Based on the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, the closed-loop candidate frame is geometrically verified by a random sampling consistency algorithm to obtain the number of inner points of each closed-loop candidate frame;

[0028] The closed-loop candidate frame with the largest number of inliers is used as a preliminary closed-loop candidate frame for timing verification, and the preliminary closed-loop candidate frame that passes the timing verification is used as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection.

[0029] As one of the preferred solutions, the closed-loop candidate frame with the largest number of inliers is used as a preliminary closed-loop candidate frame for timing verification, including:

[0030] Taking the closed-loop candidate frame with the largest number of inliers as a first preliminary closed-loop candidate frame, and constructing the nearest neighbor of the first preliminary closed-loop candidate frame;

[0031] The current key frame is updated, and the closed-loop candidate frame search and geometric verification steps of the updated current key frame are repeated to obtain the updated preliminary closed-loop candidate frame of the current key frame as the second preliminary closed-loop candidate frame;

[0032] When the second preliminary closed-loop candidate frame is the nearest neighbor of the first preliminary closed-loop candidate frame, the timing verification of the first preliminary closed-loop candidate frame passes; otherwise, the timing verification of the first preliminary closed-loop candidate frame fails, and the timing verification step is performed on the second preliminary closed-loop candidate frame until all key frames in the waiting frame queue are iteratively verified.

[0033] A second aspect of the present invention provides an underwater visual closed-loop detection device, comprising:

[0034] A feature calculation module is used to sequentially obtain multiple key frames and their corresponding key frame images, extract global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially put the key frames into a queue of frames waiting to be inserted;

[0035] A database construction module, used for obtaining the current key frame in the queue of frames waiting to be inserted, and calculating the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted based on the global feature information of the current key frame, so as to construct a closed-loop database;

[0036] A candidate frame search module, configured to search for closed-loop candidate frames from the closed-loop database based on global feature information and inter-frame adjacency of the current key frame;

[0037] The candidate frame verification module is used to verify the closed-loop candidate frame according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, the closed-loop candidate frame is used as the target closed-loop frame of the current key frame to realize underwater visual closed-loop detection.

[0038] A third aspect of the present invention provides an electronic device, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the underwater visual closed-loop detection method as described above when executing the computer program.

[0039] A fourth aspect of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the underwater visual closed-loop detection method as described above is implemented.

[0040] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0041] (1) By extracting global and local feature information from key frame images, it can comprehensively and accurately capture the important features of the image, providing a solid foundation for subsequent analysis and matching;

[0042] (2) Based on the global feature information of the current keyframe, the global similarity and temporal correlation between it and other keyframes are calculated, so that a closed-loop database with strong correlation can be constructed, which not only contains the visual features of the keyframes, but also incorporates temporal information, which helps to identify closed loops more accurately;

[0043] (3) Considering the difficulties in building a closed-loop database and screening closed-loop frames, it is proposed to use image similarity, temporal correlation and inter-frame adjacency to constrain the closed-loop detection process, which enhances the effectiveness of global and local features of the image and improves the accuracy and recall of underwater visual closed-loop. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the implementation mode will be briefly introduced below. Obviously, the drawings described below are only some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0045] Figure 1 is a flow chart of an underwater visual closed-loop detection method provided by a certain embodiment of the present invention;

[0046] Figure 2is a flowchart of image feature extraction provided by a certain embodiment of the present invention;

[0047] Figure 3 is a closed-loop database construction flow chart provided by a certain embodiment of the present invention;

[0048] Figure 4 is a structural diagram of an underwater visual closed-loop detection device provided by a certain embodiment of the present invention;

[0049] Figure 5 It is a structural diagram of an electronic device provided by a certain embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following is a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings and embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0052] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0053] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0054] At present, visual simultaneous localization and mapping technology uses cameras to perceive the surrounding environment and can complete self-localization without the need for a priori maps. Due to its low cost and good visibility, it is gradually being widely used in underwater research. The closed-loop detection module is an important part of SLAM. It is used to identify the location that the robot has been to before, and then perform closed-loop correction, thereby effectively reducing the positioning error caused by trajectory drift. However, the single structure and insufficient texture of the underwater environment bring great challenges to closed-loop detection.

[0055] To address this issue, for example, the patent with an authorization date of January 25, 2022, an authorization announcement number of CN113112565B, and an invention name of "A map construction method, storage medium and device based on robot vision closed-loop detection algorithm", uses Gaussian HOG descriptors to extract global features and performs secondary detection on the position matching of strong feature areas to achieve efficient closed-loop detection, but this scheme does not consider the impact of the construction of a closed-loop database on the closed-loop recall rate in the closed-loop frame search; for another example, the patent with an authorization date of December 23, 2022, an authorization announcement number of CN113033555B, and an invention name of "Visual SLAM closed-loop detection method based on metric learning", uses a triplet loss function to train the metric learning in the SLAM system, and performs similarity judgment on the key frame feature vectors obtained by the learning model, but this scheme does not involve the verification process of the closed-loop candidate frames, resulting in a certain loss of closed-loop accuracy. At the same time, it also does not consider the construction of a closed-loop database, and the closed-loop recall rate is not high.

[0056] Based on this, the convolutional neural network features extracted from specific layers of the present invention can show the advantage of higher recognition ability compared with hand-crafted image features with the help of convolutional neural networks. Convolutional neural networks are often used in closed-loop detection. The convolutional neural network model is used to extract global features of frame images to quickly search for similar images, and local features of frame images are simultaneously extracted for image geometric verification. Considering the influence of closed-loop database construction and HNSW graph connection frames on closed-loop detection, closed-loop detection is constrained by image similarity, temporal correlation and inter-frame adjacency to improve closed-loop recall and accuracy.

[0057] In one embodiment, if Figure 1 As shown, the first aspect of the present invention provides an underwater visual closed-loop detection method, comprising:

[0058] S1. Obtain multiple key frames and their corresponding key frame images in sequence, extract global feature information and local feature information of the key frame images through a convolutional neural network model, and put the key frames into a queue of frames waiting to be inserted in sequence;

[0059] In one embodiment, sequentially acquiring a plurality of key frames and their corresponding key frame images, and extracting global feature information and local feature information of the key frame images through a convolutional neural network model, includes:

[0060] Acquire multiple key frames and their corresponding key frame images in sequence from the front end of the SLAM system;

[0061] Extracting local features of the key frame image through the convolution layer of the convolution neural network model, and respectively extracting high-distinctive information and low-dimensional representation information of the local features through the attention layer of the convolution neural network model and the autoencoder of the convolution neural network model, to obtain local feature information of the key frame image;

[0062] The global features of the key frame image are extracted through the convolution layer of the convolutional neural network model, and the contribution of the global features is weighted through the generalized average pooling layer of the convolutional neural network model to obtain the global feature information of the key frame image.

[0063] Specifically, considering the structural uniformity of the underwater environment, the present invention adopts the ResNet-50 network as the backbone convolutional neural network to extract features of the key frame image: after the key frame image is obtained from the key frame queue at the front end of the SLAM system, the local features are extracted through the convolutional layer Conv2 of the ResNet-50 network, and the attention layer of the ResNet-50 network is used to extract the feature information with high distinctiveness in the local features, and then the convolutional autoencoder of the ResNet-50 network is used to learn the low-dimensional representation of the local features, and the local feature information of the key frame image can be obtained by combining the two; at the same time, the convolutional layer Conv2 of the ResNet-50 network is used to extract the global features of the key frame image, and the generalized average pooling layer of the ResNet-50 network is used to weight the contribution of each feature to obtain the global feature description information of the key frame image. The feature extraction process of the key frame image is as follows: Figure 2 shown.

[0064] In the initial stage of the SLAM system, closed loops will not occur, so the keyframes can be directly inserted into the closed-loop database after feature extraction. When the number of frames in the closed-loop database reaches a certain threshold, the new keyframes are sequentially placed in the queue of frames waiting to be inserted. Considering the motion characteristics of the underwater carrier and the construction elements of the closed-loop database, this example sets the keyframe number threshold to 50 frames. The present invention takes into account the impact of low-quality underwater images on visual closed-loop detection, and uses different layers of convolutional neural networks to simultaneously learn the global and local features of keyframe images to improve the search efficiency of closed-loop candidate frames and the topological information associated with the data.

[0065] S2, obtaining the current key frame in the queue of frames waiting to be inserted, and based on the global feature information of the current key frame, calculating the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to construct a closed-loop database;

[0066] In one embodiment, the global feature information of the current key frame is used to calculate the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to construct a closed-loop database, including:

[0067] Calculate the global similarity and time correlation between the current key frame and the head key frame in the queue of frames waiting to be inserted according to the global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted;

[0068] In the determination phase, if the global similarity and the time correlation meet the first requirement determination or the second requirement determination, the head key frame is inserted into the closed-loop database and removed from the queue of frames waiting to be inserted, so as to update the head key frame in the queue of frames waiting to be inserted; the first requirement determination is that the global similarity meets the similarity requirement, and the second requirement determination is that the global similarity does not meet the similarity requirement and the time correlation meets the correlation requirement;

[0069] Based on the global feature information of the current key frame, the determination step is repeated until the number of frames waiting to be inserted into the frame queue reaches a preset number, and the closed-loop database is output.

[0070] Specifically, after determining the current key frame in the queue of frames waiting to be inserted, the global similarity and temporal correlation between the two are calculated based on its global feature information and the global feature information of the first key frame in the queue of frames waiting to be inserted; wherein the global similarity, also known as image similarity, is used to characterize the global similarity between the current key frame and the first key frame in the queue of frames waiting to be inserted, and is expressed by the following formula:

[0071]

[0072] In the formula, gf i is the global feature information of the current key frame; gf q It is the global feature information of the first key frame of the team; is the L2 norm; α is the global feature similarity threshold, which is obtained through the similarity comparison experiment of underwater images;

[0073] The temporal correlation is used to characterize the temporal similarity between the current key frame and the first key frame in the queue of frames waiting to be inserted, and is expressed by the following formula:

[0074] |t i -t q |·f≥β

[0075] Where, t i is the timestamp corresponding to the global feature information of the current key frame; t q is the timestamp corresponding to the global feature information of the first key frame; f is the camera frequency; β is the time correlation threshold.

[0076] The present invention utilizes image similarity and time correlation to determine whether the head key frame in the queue of frames waiting to be inserted can be inserted into a closed-loop database. Specifically, in the judgment link, if the global similarity is not less than the global feature similarity threshold, that is, the global similarity meets the similarity requirement, and the global similarity calculation formula is established; or if the global similarity is less than the global feature similarity threshold, and the time correlation is not less than the time correlation threshold, that is, the time correlation meets the correlation requirement, and the time correlation calculation formula is established, then the head key frame in the queue of inserted frames can be inserted into the closed-loop database as a closed-loop frame, and can be removed from the queue of frames waiting to be inserted to obtain a new head key frame; then the judgment link is repeated based on the global feature information of the new head key frame and the global feature information of the current key frame until the number of frames in the queue of frames waiting to be inserted reaches a preset number, and the preset number is preferably zero, to obtain a closed-loop database.

[0077] The present invention takes into account the influence of the closed-loop database construction method of the SLAM system on the closed-loop detection effect, and proposes to use the image similarity and time correlation between the current key frame and the head key frame in the queue of frames waiting to be inserted to reasonably judge whether the head key frame can be inserted into the closed-loop database as a closed-loop frame, thereby effectively avoiding unnecessary repeated data addition, reducing database redundancy, and improving system resource utilization; the dual judgment conditions (first demand judgment and second demand judgment) adopted ensure that only those key frames that are similar in features and reasonable in time (that is, not too far or too close to avoid interference from outdated information) can be selected to be added to the closed-loop database, thereby improving the accuracy and efficiency of closed-loop detection.

[0078] In one embodiment, after calculating the global similarity and time correlation between the global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted, the method includes:

[0079] When the global similarity and the time correlation meet the third requirement, the current key frame is updated, and the global similarity and time correlation between the two are recalculated according to the updated global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted; the third requirement is that the global similarity does not meet the similarity requirement and the time correlation does not meet the correlation requirement;

[0080] The demand is determined based on the recalculated global similarity and time correlation, and the head key frame or the updated current key frame is updated based on the determination result. The determination step is repeated based on the updated result until all current key frames in the queue of frames waiting to be inserted are traversed and the closed-loop database is output.

[0081] Specifically, when the global similarity between the current key frame and the first key frame in the team is less than the global feature similarity threshold, and the time correlation between the two is less than the time correlation threshold, that is, the global similarity does not meet the similarity requirement and the time correlation does not meet the correlation requirement, the current key frame is updated, and the key frame at the next moment can be used as the current key frame, or other updating methods can be used. The specific method used is not limited here; then the global similarity and time correlation between the new current key frame and the first key frame in the team are calculated, and the calculated results are used to make a demand judgment, and the first key frame in the team is updated according to the judgment result and the closed-loop database is enriched or the current key frame is updated, and the judgment link is repeated with the updated data until all the key frames in the queue of frames waiting to be inserted are traversed, that is, only key frames that do not meet the similarity requirements and correlation requirements are left in the queue of frames waiting to be inserted, and the final closed-loop database is output for subsequent search.

[0082] The closed-loop database construction process is as follows Figure 3As shown, the present invention calculates the global similarity and time correlation between the current key frame and the head key frame in the queue of frames waiting to be inserted, and makes demand judgments based on these calculation results to decide when and how to update the key frame, which helps to ensure that only those key frames that truly have potential closed-loop value are selected, thereby improving the accuracy and efficiency of closed-loop detection; when the global similarity and time correlation do not meet the third demand judgment, the dynamic update and re-evaluation mechanism adopted enables the system to respond to complex and changeable scenarios more flexibly, thereby improving the adaptability and robustness of the system; by constructing a closed-loop database through gradual iterative update and recalculation, unnecessary repeated calculations and data processing are avoided, thereby saving system resources and improving overall performance.

[0083] S3, searching for closed-loop candidate frames from the closed-loop database based on the global feature information of the current key frame and inter-frame adjacency;

[0084] In one embodiment, step S3 includes:

[0085] Respectively calculating the similarity between the global feature information of each closed-loop frame in the closed-loop database and the global feature information of the current key frame, and sorting them according to the similarity to construct a closed-loop frame queue;

[0086] Constructing a closed-loop frame sequence according to the current closed-loop frame, the first closed-loop frame and the second closed-loop frame in the closed-loop frame queue, and using the closed-loop frame sequence as the nearest neighbor of the current closed-loop frame; the first closed-loop frame and the second closed-loop frame are respectively the previous closed-loop frame and the next closed-loop frame of the current closed-loop frame;

[0087] Each closed-loop frame except the head closed-loop frame in the closed-loop frame queue is taken as a closed-loop frame to be verified and verified in sequence, and each closed-loop frame to be verified that meets the first verification condition or the second verification condition is taken as a preferred closed-loop frame, and the head closed-loop frame in the closed-loop frame sequence is taken as the closed-loop candidate frame of the current key frame; the first verification condition is that when the first closed-loop frame of the closed-loop frame to be verified belongs to the nearest neighbor of the head closed-loop frame, the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame; the second verification condition is that when the first closed-loop frame of the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame, the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame and the closed-loop frame to be verified does not belong to the nearest neighbor of the first closed-loop frame of the closed-loop frame to be verified.

[0088] Specifically, using the current key frame Kf iThe global feature information of the frame is used to search for the top k closed-loop candidate frames similar to it in the closed-loop database using similarity algorithms such as the color histogram method, the texture feature method, and the GIST feature method, and sort them by similarity to form a closed-loop frame queue; a closed-loop frame is selected from the closed-loop frame queue and combined with its previous closed-loop frame and its next closed-loop frame to obtain the closed-loop frame sequence of the closed-loop frame, that is, the nearest neighbor of the closed-loop frame; then, each closed-loop frame in the closed-loop frame queue except the head closed-loop frame is used as a closed-loop frame to be verified and is verified in sequence. Indicates the first closed-loop frame to be verified in the closed-loop frame sequence, that is, the closed-loop frame most similar to the current key frame. If satisfied That is, if the first verification condition is met, it can be used as a preferred closed-loop frame. The closed-loop frame to be verified The nearest neighbor of , if not satisfied, then the closed-loop frame to be verified is an invalid closed-loop frame; and for the closed-loop frame to be verified If it satisfies the following formula:

[0089]

[0090] The closed-loop frame to be verified To select the optimal closed-loop frame, the aforementioned method is iterated and analogized in sequence, and all the closed-loop frames to be verified in the closed-loop frame queue are verified in order, and the preferred closed-loop frame that passes the verification and the closed-loop frame at the head of the queue are used as closed-loop candidate frames for the current key frame. It should be noted that during the verification, since the closed-loop frame at the head of the queue has the highest similarity with the current key frame, it does not need to participate in the verification process and can be used as a closed-loop candidate frame for the current key frame.

[0091] The present invention efficiently constructs a closed-loop frame queue based on the similarity between the global feature information of the closed-loop frame and the current key frame, so as to effectively reduce the interference of irrelevant frames and improve the efficiency and accuracy of closed-loop detection; by constructing the nearest neighbor of the current closed-loop frame, it is helpful to maintain the temporal or spatial coherence between closed-loop frames, and to more accurately identify and verify closed-loop events in subsequent steps; by verifying the closed-loop frame to be verified and screening out the preferred closed-loop frame according to the verification result, the influence of mismatch and noise can be further reduced to ensure that the correlation between the finally selected closed-loop candidate frame and the current key frame is more reliable, and at the same time, the range of candidate frames is gradually narrowed, thereby reducing unnecessary calculations, so as to avoid obtaining multiple closed-loop candidate key frames as adjacent frames, and the closed-loop candidate frames are screened by the nearest neighbor, so as to suppress the selection of closed-loop candidate frames of adjacent frames, thereby improving the closed-loop recall rate.

[0092] S4, verifying the closed-loop candidate frame according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, using the closed-loop candidate frame as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection;

[0093] In one embodiment, step S4 includes:

[0094] Based on the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, the closed-loop candidate frame is geometrically verified by a random sampling consistency algorithm to obtain the number of inner points of each closed-loop candidate frame;

[0095] The closed-loop candidate frame with the largest number of inliers is used as a preliminary closed-loop candidate frame for timing verification, and the preliminary closed-loop candidate frame that passes the timing verification is used as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection.

[0096] Specifically, the present invention first uses a feature matching algorithm (such as brute force matching, FLANN, etc.) to match the local feature information of the current key frame with the local feature information of each closed-loop candidate frame; then initializes the RANSAC algorithm to determine the parameters required by the RANSAC algorithm, including the maximum number of iterations, the inlier determination threshold, the minimum number of samples required for model estimation (such as the four point pairs required for point cloud alignment), etc., and randomly extracts a group of samples from the matching features of the current key frame and the closed-loop candidate frame to estimate the geometric transformation model (such as rigid body transformation, affine transformation, projection transformation, etc.); substitutes the remaining feature matching pairs into the estimated model, and determines the inlier determination threshold (such as distance error, angle error, etc.) according to the inlier determination threshold. Determine which matching pairs are inliers and count the number of inliers that conform to the model; repeat the above process of random sampling, model estimation and inlier determination until the maximum number of iterations is reached or other stopping conditions are met (such as the number of inliers has reached a preset number), and after each iteration, select the model with the largest number of inliers as the current optimal model, and for each closed-loop candidate frame, record the number of inliers under the optimal model; finally output the number of inliers for each closed-loop candidate frame under the optimal model as an important indicator for evaluating the consistency of the geometric relationship between the closed-loop candidate frame and the current key frame, and use the closed-loop candidate frame with the largest number of inliers as the preliminary closed-loop candidate frame for timing verification, and the preliminary closed-loop candidate frame that passes the verification can be used as the target closed-loop frame of the current key frame.

[0097] By matching the local feature information of the current key frame and the closed-loop candidate frame, the present invention can more accurately capture the geometric relationship between the two to reduce the possible mismatch caused by the global feature information; by performing geometric verification through the RANSAC algorithm, the closed-loop candidate frames that are consistent with the geometric relationship of the current key frame can be effectively screened out, thereby improving the accuracy of closed-loop detection, and accurately identifying closed-loop candidate frames in complex or dynamic environments, thereby enhancing the robustness of the system.

[0098] In one embodiment, the step of using the closed-loop candidate frame with the largest number of inliers as a preliminary closed-loop candidate frame for timing verification includes:

[0099] Taking the closed-loop candidate frame with the largest number of inliers as a first preliminary closed-loop candidate frame, and constructing the nearest neighbor of the first preliminary closed-loop candidate frame;

[0100] The current key frame is updated, and the closed-loop candidate frame search and geometric verification steps of the updated current key frame are repeated to obtain the updated preliminary closed-loop candidate frame of the current key frame as the second preliminary closed-loop candidate frame;

[0101] When the second preliminary closed-loop candidate frame is the nearest neighbor of the first preliminary closed-loop candidate frame, the timing verification of the first preliminary closed-loop candidate frame passes; otherwise, the timing verification of the first preliminary closed-loop candidate frame fails, and the timing verification step is performed on the second preliminary closed-loop candidate frame until all key frames in the waiting frame queue are iteratively verified.

[0102] Specifically, the present invention takes the closed-loop candidate frame with the largest number of internal points as the first preliminary closed-loop candidate frame, and determines the nearest neighbor of the first preliminary closed-loop candidate frame with reference to the closed-loop frame sequence construction method to perform timing verification, which can also be called adjacency verification: the current key frame is updated and the updated preliminary closed-loop candidate frame of the current key frame is obtained as the second preliminary closed-loop candidate frame. If the second preliminary closed-loop candidate frame belongs to the nearest neighbor of the first preliminary closed-loop candidate frame, it means that the first preliminary closed-loop candidate frame of the current key frame has passed the timing verification, and it is used as the target closed-loop frame of the current key frame; if it does not meet the requirements, it means that the timing verification has failed, and the second preliminary closed-loop candidate frame is subjected to timing verification until all key frames in the waiting frame queue are iteratively verified.

[0103] The present invention uses the closed-loop candidate frame with the largest number of inner points as the preliminary closed-loop candidate frame, which can ensure that the geometric relationship between the selected candidate frame and the current key frame is the most consistent, thereby improving the accuracy of closed-loop detection; by constructing the nearest neighbor of the preliminary closed-loop candidate frame, it is helpful to quickly screen out the candidate frame most similar to the current key frame in the subsequent verification process, further improving the efficiency of closed-loop detection; and the addition of the timing verification step can ensure that the selected closed-loop frame is consistent with the current key frame in timing, avoiding closed-loop detection errors caused by mismatching or abnormal data; by constructing the nearest neighbor and iteratively updating, the huge amount of calculation brought by the global search is avoided, and the consumption of computing resources is reduced; the closed-loop candidate frame and the search frame are double-time-verified by the nearest neighbor sequence, thereby ensuring the accuracy of closed-loop detection.

[0104] In the embodiment of the present application, based on the problem of how to improve the positioning accuracy and recall rate of underwater visual closed-loop detection, an underwater visual closed-loop detection method is designed, which realizes sequentially acquiring multiple key frames and their corresponding key frame images, extracting the global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially placing the key frames into a queue of frames waiting to be inserted; acquiring the current key frame in the queue of frames waiting to be inserted, and based on the global feature information of the current key frame, calculating the global similarity and temporal correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to obtain the global feature information of the current key frame and the global similarity and temporal correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to obtain the global feature information of the current key frame and the global similarity and temporal correlation between the current key frame and the ... Construct a closed-loop database; search for closed-loop candidate frames from the closed-loop database based on the global feature information of the current key frame and the inter-frame adjacency; verify the closed-loop candidate frames according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frames, and when the verification passes, use the closed-loop candidate frames as the target closed-loop frames of the current key frame to implement a technical solution for underwater visual closed-loop detection; constrain closed-loop detection by image similarity, temporal correlation and inter-frame adjacency, thereby enhancing the effectiveness of global and local features of the image and improving the underwater visual closed-loop accuracy and recall rate.

[0105] It should be noted that although the steps in the above flowchart are shown in sequence according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders.

[0106] In another embodiment, if Figure 4 As shown, the second aspect of the present invention provides an underwater visual closed-loop detection device, comprising:

[0107] The feature calculation module 10 is used to sequentially obtain multiple key frames and their corresponding key frame images, extract global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially put the key frames into a queue of frames waiting to be inserted;

[0108] A database construction module 20 is used to obtain the current key frame in the queue of frames waiting to be inserted, and based on the global feature information of the current key frame, calculate the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to construct a closed-loop database;

[0109] A candidate frame search module 30, configured to search for closed-loop candidate frames from the closed-loop database based on global feature information and inter-frame adjacency of the current key frame;

[0110] The candidate frame verification module 40 is used to verify the closed-loop candidate frame according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, the closed-loop candidate frame is used as the target closed-loop frame of the current key frame to realize underwater visual closed-loop detection.

[0111] It should be noted that each module in the above-mentioned underwater visual closed-loop detection device can be fully or partially implemented by software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules. For the specific definition of an underwater visual closed-loop detection device, please refer to the definition of an underwater visual closed-loop detection method above. The two have the same functions and effects, which will not be repeated here.

[0112] A third aspect of the present invention provides an electronic device, the electronic device comprising:

[0113] processor, memory, and bus;

[0114] The bus is used to connect the processor and the memory;

[0115] The memory is used to store operation instructions;

[0116] The processor is used to call the operation instruction, and the executable instruction enables the processor to perform operations corresponding to an underwater visual closed-loop detection method as shown in the first aspect of the present application.

[0117] In an alternative embodiment, an electronic device is provided, such as Figure 5 As shown, Figure 5The electronic device 5000 shown includes: a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, such as through a bus 5002. Optionally, the electronic device 5000 may also include a transceiver 5004. It should be noted that in actual applications, the transceiver 5004 is not limited to one, and the structure of the electronic device 5000 does not constitute a limitation on the embodiments of the present application.

[0118] Processor 5001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 5001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0119] The bus 5002 may include a path to transmit information between the above components. The bus 5002 may be a PCI bus or an EISA bus, etc. The bus 5002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 5 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0120] The memory 5003 may be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an EEPROM, a CD-ROM or other optical disk storage, an optical disk storage (including a compressed optical disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0121] The memory 5003 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 5001. The processor 5001 is used to execute the application code stored in the memory 5003 to implement the content shown in any of the above method embodiments.

[0122] Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc.

[0123] The fourth aspect of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, an underwater visual closed-loop detection method shown in the first aspect of the present application is implemented.

[0124] Another embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding content in the aforementioned method embodiment.

[0125] In addition, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, the steps of the above method are implemented.

[0126] In summary, the present invention relates to the field of computer technology, and discloses an underwater visual closed-loop detection method, device, equipment and medium, wherein the method comprises sequentially acquiring a plurality of key frames and their corresponding key frame images, extracting global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially placing the key frames into a queue of frames waiting to be inserted; acquiring the current key frame in the queue of frames waiting to be inserted, and based on the global feature information of the current key frame, calculating the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to construct a closed-loop database; based on the current key frame, The global feature information of the key frame and the inter-frame adjacency are used to search for closed-loop candidate frames from the closed-loop database; the closed-loop candidate frames are verified according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frames, and when the verification is passed, the closed-loop candidate frames are used as the target closed-loop frames of the current key frame to achieve underwater visual closed-loop detection; the present invention takes into account the difficulties in closed-loop database construction and closed-loop frame screening, constrains closed-loop detection through image similarity, time correlation and inter-frame adjacency, enhances the effectiveness of global and local features of the image, and improves the underwater visual closed-loop accuracy and recall rate.

[0127] Each embodiment in this specification is described in a progressive manner, and the same or similar parts of each embodiment can be directly referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. It should be noted that the technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above-mentioned embodiments are not 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.

[0128] The above-mentioned embodiments only express several preferred implementation modes of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in the technical field, several improvements and substitutions can be made without departing from the technical principles of the present invention, and these improvements and substitutions should also be regarded as the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be based on the protection scope of the claims.

Claims

1. An underwater visual closed-loop detection method, characterized in that: include: Acquire multiple key frames and their corresponding key frame images in sequence, extract global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially put the key frames into a queue of frames waiting to be inserted; Acquire the current key frame in the queue of frames waiting to be inserted, and calculate the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted based on the global feature information of the current key frame, so as to construct a closed-loop database; Based on the global feature information and inter-frame adjacency of the current key frame, searching for closed-loop candidate frames from the closed-loop database; The closed-loop candidate frame is verified according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, the closed-loop candidate frame is used as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection.

2. The underwater visual closed-loop detection method according to claim 1, characterized in that: The step of sequentially acquiring a plurality of key frames and their corresponding key frame images, and extracting global feature information and local feature information of the key frame images through a convolutional neural network model includes: Acquire multiple key frames and their corresponding key frame images in sequence from the front end of the SLAM system; Extracting local features of the key frame image through the convolution layer of the convolution neural network model, and respectively extracting high-distinctive information and low-dimensional representation information of the local features through the attention layer of the convolution neural network model and the autoencoder of the convolution neural network model, to obtain local feature information of the key frame image; The global features of the key frame image are extracted through the convolution layer of the convolutional neural network model, and the contribution of the global features is weighted through the generalized average pooling layer of the convolutional neural network model to obtain the global feature information of the key frame image.

3. The underwater visual closed-loop detection method according to claim 1, characterized in that: The global feature information of the current key frame is used to calculate the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted, so as to construct a closed-loop database, including: Calculate the global similarity and time correlation between the current key frame and the head key frame in the queue of frames waiting to be inserted according to the global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted; In the determination phase, if the global similarity and the time correlation meet the first requirement determination or the second requirement determination, the head key frame is inserted into the closed-loop database and removed from the queue of frames waiting to be inserted, so as to update the head key frame in the queue of frames waiting to be inserted; the first requirement determination is that the global similarity meets the similarity requirement, and the second requirement determination is that the global similarity does not meet the similarity requirement and the time correlation meets the correlation requirement; Based on the global feature information of the current key frame, the determination step is repeated until the number of frames waiting to be inserted into the frame queue reaches a preset number, and the closed-loop database is output.

4. The underwater visual closed-loop detection method according to claim 3, characterized in that: After calculating the global similarity and time correlation between the global feature information of the current key frame and the global feature information of the first key frame in the queue of frames waiting to be inserted, the method includes: When the global similarity and the time correlation meet the third requirement, the current key frame is updated, and the global similarity and time correlation between the two are recalculated according to the updated global feature information of the current key frame and the global feature information of the head key frame in the queue of frames waiting to be inserted; the third requirement is that the global similarity does not meet the similarity requirement and the time correlation does not meet the correlation requirement; The demand is determined based on the recalculated global similarity and time correlation, and the head key frame or the updated current key frame is updated based on the determination result. The determination step is repeated based on the updated result until all current key frames in the queue of frames waiting to be inserted are traversed and the closed-loop database is output.

5. The underwater visual closed-loop detection method according to claim 1, characterized in that: The step of searching the closed-loop candidate frame from the closed-loop database based on the global feature information of the current key frame and the inter-frame adjacency comprises: Respectively calculating the similarity between the global feature information of each closed-loop frame in the closed-loop database and the global feature information of the current key frame, and sorting them according to the similarity to construct a closed-loop frame queue; Constructing a closed-loop frame sequence according to the current closed-loop frame, the first closed-loop frame and the second closed-loop frame in the closed-loop frame queue, and using the closed-loop frame sequence as the nearest neighbor of the current closed-loop frame; the first closed-loop frame and the second closed-loop frame are respectively the previous closed-loop frame and the next closed-loop frame of the current closed-loop frame; Each closed-loop frame except the head closed-loop frame in the closed-loop frame queue is taken as a closed-loop frame to be verified and verified in sequence, and each closed-loop frame to be verified that meets the first verification condition or the second verification condition is taken as a preferred closed-loop frame, and the head closed-loop frame in the closed-loop frame sequence is taken as the closed-loop candidate frame of the current key frame; the first verification condition is that when the first closed-loop frame of the closed-loop frame to be verified belongs to the nearest neighbor of the head closed-loop frame, the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame; the second verification condition is that when the first closed-loop frame of the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame, the closed-loop frame to be verified does not belong to the nearest neighbor of the head closed-loop frame and the closed-loop frame to be verified does not belong to the nearest neighbor of the first closed-loop frame of the closed-loop frame to be verified.

6. The underwater visual closed-loop detection method according to claim 5, characterized in that: The verifying the closed-loop candidate frame according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, using the closed-loop candidate frame as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection, includes: Based on the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, the closed-loop candidate frame is geometrically verified by a random sampling consistency algorithm to obtain the number of inner points of each closed-loop candidate frame; The closed-loop candidate frame with the largest number of inliers is used as a preliminary closed-loop candidate frame for timing verification, and the preliminary closed-loop candidate frame that passes the timing verification is used as the target closed-loop frame of the current key frame to achieve underwater visual closed-loop detection.

7. The underwater visual closed-loop detection method according to claim 6, characterized in that: The step of taking the closed-loop candidate frame with the largest number of inliers as a preliminary closed-loop candidate frame for timing verification includes: Taking the closed-loop candidate frame with the largest number of inliers as a first preliminary closed-loop candidate frame, and constructing the nearest neighbor of the first preliminary closed-loop candidate frame; The current key frame is updated, and the closed-loop candidate frame search and geometric verification steps of the updated current key frame are repeated to obtain the updated preliminary closed-loop candidate frame of the current key frame as the second preliminary closed-loop candidate frame; When the second preliminary closed-loop candidate frame is the nearest neighbor of the first preliminary closed-loop candidate frame, the timing verification of the first preliminary closed-loop candidate frame passes; otherwise, the timing verification of the first preliminary closed-loop candidate frame fails, and the timing verification step is performed on the second preliminary closed-loop candidate frame until all key frames in the waiting frame queue are iteratively verified.

8. An underwater visual closed-loop detection device, characterized in that: include: A feature calculation module is used to sequentially obtain multiple key frames and their corresponding key frame images, extract global feature information and local feature information of the key frame images through a convolutional neural network model, and sequentially put the key frames into a queue of frames waiting to be inserted; A database construction module, used for obtaining the current key frame in the queue of frames waiting to be inserted, and calculating the global similarity and time correlation between the current key frame and other key frames in the queue of frames waiting to be inserted based on the global feature information of the current key frame, so as to construct a closed-loop database; A candidate frame search module, configured to search for closed-loop candidate frames from the closed-loop database based on global feature information and inter-frame adjacency of the current key frame; The candidate frame verification module is used to verify the closed-loop candidate frame according to the local feature information of the current key frame and the local feature information of the closed-loop candidate frame, and when the verification passes, the closed-loop candidate frame is used as the target closed-loop frame of the current key frame to realize underwater visual closed-loop detection.

9. An electronic device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the underwater visual closed-loop detection method as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the device where the computer-readable storage medium is located executes the computer program, the underwater visual closed-loop detection method as described in any one of claims 1 to 7 is implemented.

Images