A feature point matching method and device for descending sequence images

Abstract

The application provides a feature point matching method and device for descending sequence images, and the method comprises the following steps: acquiring descending sequence images collected in a descending process; performing feature point extraction and description on the descending sequence images to obtain to-be-matched feature points with scale information; performing same-name feature point matching on the to-be-matched feature points to obtain a feature point tracking sequence; wherein the feature point tracking sequence comprises same-name feature points arranged according to descending time; and performing false matching elimination on the feature point tracking sequence according to the scale information of the to-be-matched feature points to complete feature point matching. The feature point matching method for descending sequence images provided by the scheme effectively improves the matching accuracy of feature points in the descending sequence images.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and spacecraft deep space navigation positioning technology, and particularly relates to a feature point matching method and device for descent sequence images. BACKGROUND

[0002] In order to monitor the landing process of a lander, a large number of sequence images in the landing process are captured by a landing camera installed at the bottom of the lander, which are the descent sequence images. The descent sequence images can be used to retrieve accurate information of the terrain of the landing area, and thus high-precision positioning of the landing point can be achieved. The basis for achieving the above functions is the extraction of feature points and the matching of the same-named points. Due to the scarcity of surface textures of celestial bodies and the serious repetition of ground objects, a large number of false matching points of the same-named features can easily occur between adjacent frames of images, and thus a feature point matching method suitable for the descent sequence images is needed. SUMMARY

[0003] The embodiment of the present application provides a feature point matching method and device for descent sequence images, which can accurately realize feature point matching in the descent sequence images.

[0004] In a first aspect, the embodiment of the present application provides a feature point matching method for descent sequence images, comprising:

[0005] acquiring descent sequence images collected in the descent process;

[0006] extracting and describing feature points of the descent sequence images to obtain to-be-matched feature points with scale information;

[0007] matching the same-named feature points of the to-be-matched feature points to obtain a feature point tracking sequence; wherein the feature point tracking sequence includes the same-named feature points arranged according to the descent time;

[0008] performing false matching elimination on the feature point tracking sequence according to the scale information of the to-be-matched feature points to complete feature point matching.

[0009] Optionally, the extracting and describing feature points of the descent sequence images to obtain to-be-matched feature points with scale information comprises:

[0010] extracting feature points in the descent sequence images by using a first operator; wherein the first operator has a scale invariance property;

[0011] describing the feature points by using a second operator to obtain to-be-matched feature points with scale information.

[0012] Optionally, the matching the same-named feature points of the to-be-matched feature points to obtain a feature point tracking sequence comprises:

[0013] matching homonymic feature points in adjacent sequence images in the descending sequence images, to obtain at least one set of feature points; wherein different feature point sets correspond to different targets, and each set of feature points includes a plurality of feature points corresponding to the target;

[0014] sequentially arranging feature point pairs in each set of feature points according to descending time of the descending sequence images, to obtain a feature point tracking sequence corresponding to the set of feature points.

[0015] Optionally, the feature point matching is completed by removing false matches from the feature point tracking sequence according to the scale information of the feature points to be matched, and includes:

[0016] obtaining scale information corresponding to each homonymic feature point pair in the feature point tracking sequence arranged according to descending time;

[0017] sequentially arranging the scale information corresponding to each homonymic feature point pair according to the descending time, to obtain a scale information sequence;

[0018] determining whether the scale information sequence is monotonically decreasing;

[0019] if yes, the feature point matching is completed;

[0020] if no, false matches are removed from the feature point tracking sequence.

[0021] Optionally, the false matches are removed from the feature point tracking sequence by:

[0022] sequentially arranging the scale information corresponding to each homonymic feature point pair, to obtain a calibration scale information sequence showing a monotonically decreasing trend;

[0023] comparing the scale information sequence with the calibration scale information sequence, and removing homonymic feature points corresponding to scale information arranged differently.

[0024] In a second aspect, the embodiments of the present application further provide a feature point matching device for descending sequence images, including:

[0025] an acquisition module, configured to acquire descending sequence images collected in a descending process;

[0026] a feature point extraction module, configured to extract and describe feature points from the descending sequence images, to obtain feature points to be matched with scale information;

[0027] a matching module, configured to match homonymic feature points in the feature points to be matched, to obtain a feature point tracking sequence; wherein the feature point tracking sequence includes homonymic feature points arranged according to descending time;

[0028] The culling module is configured to perform false match elimination on the feature point tracking sequence according to the scale information of the feature points to be matched, and complete feature point matching.

[0029] Optionally, the matching module is further configured to perform the following operation:

[0030] The same-name feature point matching is performed on adjacent sequence images in the descending sequence images, to obtain at least one set of feature points; wherein different feature point sets correspond to different targets, and each set of feature points includes a plurality of feature points corresponding to the target;

[0031] The feature point pairs in each set of feature points are sorted according to the descending time sequence of the descending sequence images, to obtain a feature point tracking sequence corresponding to the set of feature points.

[0032] Optionally, the culling module is further configured to perform the following operation:

[0033] The scale information corresponding to each same-name feature point pair in the feature point tracking sequence arranged according to the descending time is obtained;

[0034] The scale information corresponding to each same-name feature point pair is sorted according to the descending time, to obtain a scale information sequence;

[0035] It is judged whether the scale information sequence is monotonically decreasing;

[0036] If yes, the feature point matching is completed;

[0037] If no, false match elimination is performed on the feature point tracking sequence.

[0038] In a third aspect, an embodiment of the present application further provides a computing device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the feature point matching method of the descending sequence image according to any one of the above aspects.

[0039] In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium storing a computer program, and the computer program, when executed in a computer, causes the computer to execute the feature point matching method of the descending sequence image according to any one of the above aspects.

[0040] The embodiment of the present application provides a feature point matching method and device of a descending sequence image. The method eliminates false matching points from a feature point tracking sequence according to the scale information of feature points to be matched, and completes feature point matching, according to the property that the scale of the feature points in the descending sequence image monotonically changes with the imaging sequence of the image. Thus, the feature point matching method of the descending sequence image provided by the present application can accurately realize feature point matching in the descending sequence image. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings described below are some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort based on these drawings.

[0042] Figure 1 is a flow chart of a feature point matching method of descending sequence images provided by an embodiment of the present application;

[0043] Figure 2 is a same-axis image scale diagram provided by an embodiment of the present application;

[0044] Figure 3 is a scale monotonous change diagram of descending sequence images provided by an embodiment of the present application;

[0045] Figure 4 is a hardware architecture diagram of a computing device provided by an embodiment of the present application;

[0046] Figure 5 is a structural diagram of a feature point matching device of descending sequence images provided by an embodiment of the present application. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the following will combine the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present application.

[0048] There are mainly two types of existing methods for detecting and removing mis-matching points in existing sequence images:

[0049] (1) The false matching points are removed by solving the projection geometric transformation model that maps the feature points between the images. The mapping model used by the false matching point removal method based on the mapping model includes parameterized models (such as affine transformation, homography and epipolar geometry) and non-parameterized models (such as non-rigid mapping model). The false matching point removal method based on the mapping model is based on the idea of "hypothesis-verification", and searches for a mapping model with the largest consistent inlier set, wherein the relatively classic ones are random sample consensus (RANdom SAmple Consensus, RANSAC) and maximum likelihood estimate sample consensus (Maximum Likelihood Estimate Sample Consensus, MLESAC). This method requires the specific form of the mapping model to be known in advance, however, the descending sequence images have a plane degeneration problem due to the flat terrain of the imaging area, which makes it impossible to obtain a mapping model with a certain form, so this method cannot be directly used for false matching point removal in the descending sequence images.

[0050] (2) In order to solve the above problems, some non-parameterized models have been born, such as identifying correspondence function (Identifying Correspondence Function, ICF) and manifold regularization-based robust point matching (Manifold Regularization-based Robust Point Matching, MR-RPM). ICF searches a pair of correspondence functions that can map the same feature points between two images to each other, and identifies and removes the false matching points according to the deviation values of the matching points with respect to the correspondence functions. MR-RPM assumes that the motion field under the manifold regularization condition is smooth, and then solves the matching problem from the perspective of robust motion field interpolation. These methods have high spatial complexity, so their practicability is also limited.

[0051] More importantly, compared with common multi-view images, the imaging mode of descent sequence images is unique. The descent sequence images belong to sequence images, the imaging baseline is approximately parallel to the camera optical axis, the camera moves forward along the optical axis during the imaging process, and the imaging scene is the preselected landing area of the lander. The landing area is usually selected in a relatively flat area, so the imaging scene is approximately planar, and the plane normal is approximately parallel to the camera optical axis. Although the above mismatch point removal method involves some special scenes, there is no method specifically for the unique imaging mode and scene of the descent sequence images. A large number of false matching points appear in the descent sequence images due to the presence of a large number of similar features. If the fundamental matrix is used as the mapping model for the parameter-based false matching point removal method, an incorrect fundamental matrix may be obtained due to the planar degeneration in the descent sequence images, which may result in the removal of correct matching points. If the plane homography matrix is used as the mapping model, the existing terrain information cannot be effectively used to determine the inlier threshold. In addition, the existing false matching point removal method cannot take advantage of the feature point scale monotonically changing feature of the descent sequence images, so a feature point matching method suitable for the descent sequence images is needed.

[0052] The specific concepts and embodiments of the present application are as follows: Figure 1 The embodiment of the present application provides a feature point matching method of descent sequence images, which comprises the following steps:

[0053] Step 100, acquiring the descent sequence images collected during the descent process;

[0054] Step 102, extracting and describing the feature points of the descent sequence images to obtain the matching feature points with scale information;

[0055] Step 104, matching the matching feature points to obtain the feature point tracking sequence; wherein the feature point tracking sequence comprises the homonym feature points arranged according to the descent time;

[0056] Step 106, removing the false matching points from the feature point tracking sequence according to the scale information of the matching feature points to complete the feature point matching.

[0057] In the embodiment of the present application, the feature points of the acquired descent sequence images are extracted and described to obtain the matching feature points with scale information, the matching feature points are matched to obtain the feature point tracking sequence of the homonym feature points arranged according to the descent time, the scale of the feature points in the descent sequence images has the property of monotonically changing with the imaging sequence, and then the false matching points are removed from the feature point tracking sequence according to the scale information of the matching feature points to complete the feature point matching. In this way, the feature point matching method of the descent sequence images provided by the present application can accurately realize the feature point matching in the descent sequence images.

[0058] It should be noted that during the descent of the lander, the acquisition height of the descent sequence images monotonically decreases with time, so the ground resolution of the descent sequence images monotonically increases with the time sequence of image acquisition, and the scale of the same-named feature points between adjacent frame sequence images also monotonically decreases with the time sequence of image acquisition.

[0059] Specifically, the monotonicity of the scale of the same-named feature points in the descent sequence images is proved as follows:

[0060] For unstructured scenes of extraterrestrial celestial body topography, a feature point extraction and description operator SIFT operator with scale invariance is selected to describe the feature points, and a Gaussian scale transformation of an initial image u can be denoted as:

[0061]

[0062] where G σ u(x) is used to represent the convolution value of the initial image u(x) and a Gaussian kernel with a standard deviation σ. The Gaussian kernel is used to simulate the optical blurring process of the image, and it is more suitable for simulating the blurring process than other function kernels because of its semi-group property, which makes the smoothing of images by Gaussian kernels with different standard deviations σ and γ continuous. For example, the standard deviation of the convolution of Gaussian kernels with standard deviations σ and γ is Therefore, the convolution of G σ u(x) and a Gaussian kernel with a standard deviation γ obtains a new convolution value:

[0063]

[0064] The Gaussian kernel is suitable for simulating the imaging process from far to near. In this process, the scale space can be regarded as a sequence of images obtained by superimposing images with different scaling degrees. The search method of SIFT feature points is to find three-dimensional extreme points in the Difference of Gaussians (DOG) scale space, denoted as v. The calculation formula of DOG is:

[0065]

[0066] where κ is a parameter that controls the sampling density of the scale. The DOG scale space is an approximation of the regularized Laplacian scale space.

[0067] Two images of a scene u0 are obtained by imaging at different positions on the same optical axis, denoted as u λ and u μ , respectively. As shown in Figure 2 , it is assumed that the sampling operators are both S1, and the Gaussian kernel functions are both G c , H λ , and H μ , respectively. u λand u μ , then the two images can be more accurately expressed as:

[0068] u λ = S1G c H λ u0 (4)

[0069] u μ = S1G c H μ u0 (5)

[0070] The scale space of the two images can be calculated from the semigroup property of the Gaussian kernel:

[0071]

[0072] v μ (σ, x) = G σ H μ u0(x) (7)

[0073] where I1 denotes the interpolation operator used. For an image u0 with infinite resolution, its Gaussian scale space is denoted as v0, then according to the commutative relation between similarity scaling matrix and convolution, we can obtain:

[0074] v λ (σ, x) = H λ (G λσ u0(x)) = v0(λσ, λx) (8)

[0075] v μ (σ, x) = v0(μσ, μx) (9)

[0076] In the same way, the DOG scale space can also be calculated from u0 (assuming κ > 1):

[0077] w λ (σ, x) = w0(λσ, λx) (10)

[0078] w μ (σ, x) = w0(μσ, μx) (11)

[0079] Assuming that there is an extreme point (σ0, X0) in the DOG scale space w0, then this extreme point corresponds to (σ1, X1) in w λ and (σ2, X2) in w μ respectively, and satisfies:

[0080] σ0 = λσ1 = μσ2 (12)

[0081] In equation (12), λ and μ respectively represent the similarity scaling matrix of the image u λand images u μ The initial scale is given by σ0, where σ0 is the absolute scale of the feature points, and σ1 and σ2 represent the relative scales of the feature points with respect to the initial scale of the image, i.e., the feature point scales described in the SIFT descriptor. Assuming λ > μ, i.e., u λ Compared to u μ It is a magnified image with higher ground resolution. Therefore, according to the above formula, σ1 < σ2, indicating that the same feature point is in image u. λ The SIFT description scale in the image is smaller than that in the u μ SIFT describes the scale. Since the ground resolution of descent sequence images monotonically increases with the temporal order of image acquisition, the scale of corresponding feature points in descent sequence images monotonically decreases with the temporal order of image acquisition.

[0082] The following description Figure 1 The execution method of each step is shown.

[0083] First, for step 100, acquire the descent sequence images gathered during the descent process. It should be noted that the descent sequence images include multiple image sequences acquired in the order of image imaging.

[0084] For step 102, feature points are extracted and described from the descent sequence image to obtain feature points to be matched with scale information, including:

[0085] Feature points in the descending sequence image are extracted using the first operator; wherein, the first operator has scale-invariant properties;

[0086] The second operator is used to describe the feature points, resulting in feature points to be matched that have scale information.

[0087] Specifically, since the ground resolution of descent sequence images changes continuously with the temporal sequence of image acquisition, a feature point extraction operator with scale-invariant properties (i.e., the first operator) is required, such as SURF, ORB, and SIFT. Because this method utilizes the property that the scale of feature points in descent sequence images monotonically changes with the sequence order of images, a feature point description operator with scale description (i.e., the second operator) is needed, such as SIFT and AKAZE.

[0088] In step 104, the feature points to be matched are matched with corresponding feature points to obtain a feature point tracking sequence, including:

[0089] Matching of corresponding feature points is performed on adjacent image sequences in the descending sequence to obtain at least one set of feature points; wherein different sets of feature points correspond to different targets, and each set of feature points includes multiple feature points corresponding to the target;

[0090] The feature point pairs in each feature point set are sorted according to the descending time sequence of the descending sequence images, so as to obtain a feature point tracking sequence corresponding to the feature point set.

[0091] It should be noted that the selection of the same-name feature point matching algorithm for the descending sequence images should depend on the actual application requirements. If real-time processing is required, a feature point matching algorithm with faster calculation speed, such as the FLANN algorithm, is adopted. If the descending sequence images are transmitted back to the ground for processing at the ground remote operation center, a matching operator with lower calculation time requirement, such as the Brute-Force operator, can be adopted.

[0092] In a preferred embodiment, the Brute-Force operator is adopted for the same-name feature point matching, so as to ensure that as many feature matching points as possible are obtained.

[0093] In the present application, the overlap between any adjacent sequence images is calculated. For adjacent frame images with an overlap greater than 70%, all feature points (i.e., a feature point set) corresponding to a certain ground object or a certain point in the image can be obtained by matching the same-name feature points. Arranging all the feature points corresponding to the ground object according to the descending time sequence can obtain the feature point tracking sequence corresponding to the ground object.

[0094] In step 106, the feature point tracking sequence is subjected to false matching elimination according to the scale information of the feature points to be matched, so as to complete the feature point matching, including:

[0095] Obtaining the scale information corresponding to each same-name feature point in the feature point tracking sequence arranged according to the descending time;

[0096] Sorting the scale information corresponding to each same-name feature point according to the descending time, so as to obtain a scale information sequence;

[0097] Judging whether the scale information sequence is monotonically decreasing;

[0098] If yes, the feature point matching is completed;

[0099] If no, the feature point tracking sequence is subjected to false matching elimination.

[0100] In the present application, based on the fact that the scale of the feature points in the descending sequence images monotonically changes with the imaging sequence, whether the current feature point tracking sequence includes false matching points can be determined intuitively by judging whether the scale information sequence presents a monotonically decreasing trend. When the scale information sequence presents a monotonically decreasing trend, it is determined that the feature point matching is completed, so as to realize accurate matching of the feature points in the descending sequence images.

[0101] In a preferred embodiment, in step 106, the feature point tracking sequence is subjected to false matching elimination, including:

[0102] The scale information corresponding to each same feature point is sorted to obtain a calibration scale information sequence in a monotone decreasing trend;

[0103] The scale information sequence is compared with the calibration scale information sequence, and the same feature points corresponding to the scale information of different orders are removed.

[0104] For example, a feature point tracking sequence continuously exists in six sequence images, the acquisition times of the sequence images are T1, T2, T3, T4, T5 and T6 respectively, and T1 < T2 < T3 < T4 < T5 < T6, the feature point scales in the feature point tracking sequence corresponding to the six sequence images are σ1, σ2, σ3, σ4, σ5 and σ6 in turn, if σ1 > σ2 > σ3 > σ4 > σ5 > σ6 as shown in the figure exists, the feature point matching is completed. Figure 3 If σ1 > σ3 > σ2 > σ4 > σ5 > σ6 exists, it indicates that the feature points in the second sequence image and the third sequence image in the feature point tracking sequence are false same matching points, and should be removed, and the feature point tracking sequence obtained after the removal only has four sequence images, and the four sequence images can be used to inverse the accurate information of the landing area terrain of the lander.

[0105] In the application, when it is judged that the scale information sequence does not present a monotone decreasing trend, it can be directly determined that the current feature point tracking sequence includes false matching points, by comparing with the calibration scale information sequence in a monotone decreasing trend, the same feature points corresponding to the scale information of different orders can be removed, the false matching points are removed, the feature point matching is completed. In this way, by comparing the differences between the scale information sequence and the calibration scale information sequence, the false matching points can be directly removed, the false matching point removal efficiency is effectively improved, and the accuracy of the feature point matching is further ensured.

[0106] As shown in the figure, Figure 4 , Figure 5 The embodiment of the application provides a feature point matching device of a descending sequence image. The device embodiment can be realized by software, or realized by hardware or a combination of software and hardware. From the hardware layer, as shown in the figure, Figure 4 is a hardware architecture diagram of a computing device where a feature point matching device of a descending sequence image provided by the embodiment of the application is located. In addition to the processor, the memory, the network interface and the non-volatile memory shown in the figure, Figure 4 the computing device where the device in the embodiment is usually also can include other hardware, such as a forwarding chip responsible for processing a message and the like. Taking the software implementation as an example, as shown in the figure, Figure 5As shown, as a logical device, it is formed by the CPU of the computing device where it is located to read the corresponding computer program in the non-volatile memory into the memory to run. The embodiment provides a feature point matching device of descending sequence images, including: an acquisition module 500, a feature point extraction module 502, a matching module 504, and a rejection module 506;

[0107] The acquisition module 500 is configured to acquire the descending sequence images collected in the descending process.

[0108] The feature point extraction module 502 is configured to perform feature point extraction and description on the descending sequence images to obtain the to-be-matched feature points with scale information.

[0109] The matching module 504 is configured to perform same-name feature point matching on the to-be-matched feature points to obtain a feature point tracking sequence; wherein the feature point tracking sequence includes same-name feature points arranged according to descending time.

[0110] The rejection module 506 is configured to perform false matching rejection on the feature point tracking sequence according to the scale information of the to-be-matched feature points to complete feature point matching.

[0111] In some specific embodiments, the acquisition module 500 can be configured to perform the above step 100, the feature point extraction module 502 can be configured to perform the above step 102, the matching module 504 can be configured to perform the above step 104, and the rejection module 506 can be configured to perform the above step 106.

[0112] In some specific embodiments, the feature point extraction module 502 is further configured to perform the following operations:

[0113] extracting feature points in the descending sequence images by using a first operator; wherein the first operator has a scale invariance property;

[0114] describing the feature points by using a second operator to obtain the to-be-matched feature points with scale information.

[0115] In some specific embodiments, the matching module 504 is further configured to perform the following operations:

[0116] performing same-name feature point matching on adjacent sequence images in the descending sequence images to obtain at least one set of feature points; wherein different target corresponds to different set of feature points, and each set of feature points includes multiple feature points corresponding to the target;

[0117] sequencing the feature point pairs in each set of feature points according to descending time sequence of the descending sequence images to obtain a feature point tracking sequence corresponding to the set of feature points.

[0118] In some specific embodiments, the rejection module 506 is further configured to perform the following operations:

[0119] obtaining scale information corresponding to each of the same-named feature points arranged in descending time in the feature point tracking sequence;

[0120] sequencing the scale information corresponding to each of the same-named feature points according to descending time to obtain a scale information sequence;

[0121] judging whether the scale information sequence is monotonically decreasing;

[0122] if yes, completing feature point matching;

[0123] if no, sequencing the scale information corresponding to each of the same-named feature points to obtain a calibration scale information sequence in a monotonically decreasing trend; and comparing the scale information sequence with the calibration scale information sequence to eliminate mis-matched same-named feature points corresponding to the scale information of different sequences.

[0124] It can be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the feature point matching device for a descending sequence image. In other embodiments of the present application, the feature point matching device for a descending sequence image can include more or fewer components than the illustration, or combine certain components, or split certain components, or different component arrangement. The illustrated components can be implemented in hardware, software or a combination of software and hardware.

[0125] The information interaction and execution process between the modules in the above device are based on the same concept as the method embodiments of the present application, and the specific content can be referred to the description in the method embodiments of the present application, which will not be repeated here.

[0126] The embodiments of the present application also provide a computing device including a memory and a processor, the memory stores a computer program, and the processor implements the feature point matching method for a descending sequence image in any of the embodiments of the present application when executing the computer program.

[0127] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program makes the processor execute the feature point matching method for a descending sequence image in any of the embodiments of the present application when being executed by the processor.

[0128] Specifically, a system or device equipped with a storage medium can be provided, and the storage medium stores a software program code for realizing the functions of any of the above embodiments, and the computer (or CPU or MPU) of the system or device reads out and executes the program code stored in the storage medium.

[0129] In this case, the program code itself read from the storage medium can implement the functions of any of the above-described embodiments, and thus the program code and the storage medium which stores the program code constitute a part of the present application.

[0130] The storage medium for providing the program code includes a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as a CD-ROM, a CD-R, a CD-RW, a DVD-ROM, a DVD-RAM, a DVD- RW, a DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded to a server computer from an external package via a communication network.

[0131] Further, it is to be understood that not only the program code read out from the computer, but also the operating system or the like operating on the computer based on the instructions of the program code can perform part or all of the actual operations to realize the functions of any of the above-described embodiments.

[0132] Further, it is to be understood that the program code read out from the storage medium is written into the memory provided in the expansion board inserted into the computer or the memory provided in the expansion module connected to the computer, and then part or all of the actual operations are performed by the CPU or the like mounted on the expansion board or the expansion module based on the instructions of the program code to realize the functions of any of the above-described embodiments.

[0133] It is to be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises... a" does not, without more constraints, exclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0134] It is to be understood by those ordinary skilled in the art that all or part of the steps of the above-described method embodiments can be completed by program instruction related hardware, and the aforementioned program can be stored in a computer readable storage medium, and the program, when executed, performs steps including the above-described method embodiments; and the aforementioned storage medium includes ROM, RAM, magnetic disk or optical disk, and various storage media which can store program code.

[0135] It should be pointed out finally that the above embodiments are only used to illustrate the technical solutions of the present application, but not to limit the same; and although the present application has been described in detail with reference to the foregoing embodiments, it should be appreciated by those skilled in the art that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features thereof can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A feature point matching method for descent sequence images, characterized in that, include: Acquire descent sequence images gathered during the descent process; Feature points are extracted and described from the descent sequence image to obtain feature points to be matched with scale information; The feature points to be matched are matched with corresponding feature points to obtain a feature point tracking sequence; wherein, the feature point tracking sequence includes corresponding feature points arranged in descending time. Based on the scale information of the feature points to be matched, the feature point tracking sequence is erroneously eliminated to complete the feature point matching.

2. The method according to claim 1, characterized in that, The step of extracting and describing feature points from the descending sequence image to obtain feature points to be matched with scale information includes: Feature points in the descending sequence image are extracted using a first operator; wherein the first operator has scale-invariant properties. The feature points are described using the second operator to obtain the feature points to be matched with scale information.

3. The method according to claim 1, characterized in that, The step of performing same-name feature point matching on the feature points to be matched to obtain a feature point tracking sequence includes: The adjacent sequence images in the descending sequence image are matched with corresponding feature points to obtain at least one set of feature points; wherein, different feature point sets correspond to different targets, and each set of feature points includes multiple feature points corresponding to the target; According to the descent time order of the descent sequence images, the feature point pairs in each set of feature points are sorted to obtain the feature point tracking sequence corresponding to that set of feature points.

4. The method according to claim 1, characterized in that, The step of removing mismatches from the feature point tracking sequence based on the scale information of the feature points to be matched, thereby completing feature point matching, includes: Obtain the scale information corresponding to each feature point with the same name in the feature point tracking sequence arranged by descent time; The scale information corresponding to each feature point with the same name is sorted according to the descent time to obtain a scale information sequence; Determine whether the scale information sequence is monotonically decreasing; If so, feature point matching is complete; If not, then the feature point tracking sequence is erroneously removed.

5. The method according to claim 4, characterized in that, The process of removing mismatches from the feature point tracking sequence includes: The scale information corresponding to each feature point with the same name is sorted to obtain a calibration scale information sequence that shows a monotonically decreasing trend. The scale information sequence is compared with the calibration scale information sequence, and mismatched feature points corresponding to scale information with different ordering are removed.

6. A feature point matching device for a descending sequence image, characterized in that, include: The acquisition module is used to acquire descent sequence images collected during the descent process; The feature point extraction module is used to extract and describe feature points in the descent sequence image to obtain feature points to be matched with scale information; A matching module is used to perform same-name feature point matching on the feature points to be matched, thereby obtaining a feature point tracking sequence; wherein, the feature point tracking sequence includes same-name feature points arranged in descending time. The elimination module is used to eliminate mismatches in the feature point tracking sequence based on the scale information of the feature points to be matched, thereby completing the feature point matching.

7. The apparatus according to claim 6, characterized in that, The matching module is also used to perform the following operations: The adjacent sequence images in the descending sequence image are matched with corresponding feature points to obtain at least one set of feature points; wherein, different feature point sets correspond to different targets, and each set of feature points includes multiple feature points corresponding to the target; According to the descent time order of the descent sequence images, the feature point pairs in each set of feature points are sorted to obtain the feature point tracking sequence corresponding to that set of feature points.

8. The apparatus according to claim 6 or 7, characterized in that, The rejection module is also used to perform the following operations: Obtain the scale information corresponding to each feature point with the same name in the feature point tracking sequence arranged by descent time; The scale information corresponding to each feature point with the same name is sorted according to the descent time to obtain a scale information sequence; Determine whether the scale information sequence is monotonically decreasing; If so, feature point matching is complete; If not, then the feature point tracking sequence is erroneously removed.

9. A computing device comprising a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program, implements the method as described in any one of claims 1-5.

10. A computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, causes the computer to perform the method of any one of claims 1-5.

Images