CN116863349A

CN116863349A - Remote sensing image change area determining method and device based on triangular network dense matching

Info

Publication number: CN116863349A
Application number: CN202310865015.3A
Authority: CN
Inventors: 修晓龙; 吴瑞姣; 陈光剑; 邓西鹏; 黄裕瑶
Original assignee: Fujian Institute Of Geological Surveying And Mapping
Current assignee: Fujian Institute Of Geological Surveying And Mapping
Priority date: 2023-07-14
Filing date: 2023-07-14
Publication date: 2023-10-10

Abstract

The invention provides a remote sensing image change area determining method and device based on dense matching of a triangular network, comprising the following steps: acquiring a pre-remote sensing image and a post-remote sensing image to be processed; performing feature point matching on the early-stage remote sensing image and the later-stage remote sensing image to obtain an initial matching point set between the early-stage remote sensing image and the later-stage remote sensing image; constructing an irregular triangular network based on the initial matching point set, and performing dense matching on pixel points in each triangle in the irregular triangular network to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image; and determining a change area between the early-stage remote sensing image and the later-stage remote sensing image based on the target dense matching point set. The method is not only suitable for multi-band images, but also suitable for single-band images, has higher stability and robustness, and can automatically extract the change region of the multi-temporal remote sensing image in a batch mode.

Description

Remote sensing image change area determining method and device based on triangular network dense matching

Technical Field

The invention relates to the technical field of remote sensing image change detection, in particular to a method and a device for determining a remote sensing image change area based on dense matching of a triangular network.

Background

At present, the existing remote sensing image change detection methods are mainly divided into two main categories: a method based on multi-scale segmentation and a method based on deep learning neural network. Firstly, the existing method has a severe requirement on data, automatic registration processing is required to be carried out on the images in front and back stages in advance, and most of the images are multi-band; secondly, the processing efficiency of the method based on multi-scale segmentation is low, the classification of the ground object types also needs to be manually participated, the processing precision often depends on the manual intervention degree, the method based on the deep learning neural network needs to train a large number of relevant samples in advance, and the training and parameter adjustment time is more; again, the existing two main methods have poor universality, namely, the parameter and effect are required to be evaluated again when a batch of data is replaced.

Disclosure of Invention

Therefore, the invention aims to provide a method and a device for determining a remote sensing image change area based on dense matching of triangular networks, which are not only applicable to multi-band images, but also applicable to single-band images, have higher stability and robustness, and can automatically extract the change area of multi-time-phase remote sensing images in batches.

In a first aspect, an embodiment of the present invention provides a method for determining a remote sensing image change area based on dense matching of a triangle network, including:

acquiring a pre-remote sensing image and a post-remote sensing image to be processed;

performing feature point matching on the early-stage remote sensing image and the later-stage remote sensing image to obtain an initial matching point set between the early-stage remote sensing image and the later-stage remote sensing image;

constructing an irregular triangular network based on the initial matching point set, and performing dense matching on pixel points in each triangle in the irregular triangular network to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image;

and determining a change area between the early-stage remote sensing image and the later-stage remote sensing image based on the target dense matching point set.

In one embodiment, performing feature point matching on the early-stage remote sensing image and the late-stage remote sensing image to obtain an initial matching point set between the early-stage remote sensing image and the late-stage remote sensing image, including:

constructing a front pyramid image corresponding to the front remote sensing image and a rear pyramid image corresponding to the rear remote sensing image; wherein the early pyramid image and the late pyramid image each comprise a plurality of image layers;

Extracting a plurality of feature points to be matched from the early-stage remote sensing image;

traversing each image layer of the earlier pyramid image and each image layer of the later pyramid image for multiple times through sliding windows with different sizes, so as to perform feature point matching on each feature point to be matched by utilizing a plurality of feature point matching algorithms in each traversing process, and obtaining a matching point set corresponding to each traversing process;

and for each feature point to be matched, if the matching results of the feature points to be matched are consistent in the matching point set corresponding to each traversal process, reserving the matching results of the feature points to be matched and the matching feature points to obtain an initial matching point set.

In one embodiment, performing dense matching on pixel points inside each triangle in the irregular triangle network to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image, where the method includes:

performing dense matching on pixel points in each triangle in the irregular triangle network to obtain an initial dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image;

Determining affine transformation error values of each dense matching point in the initial dense matching point set, and if the affine transformation error values of the dense matching points are larger than a preset error threshold value, eliminating the dense matching points from the initial dense matching point set to obtain an intermediate dense matching point set;

and detecting the missing matching points in the middle dense matching point set, and performing nearest neighbor matching on the missing matching points to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

In one embodiment, detecting a missing matching point in the set of intermediate closely-matched points includes:

dividing the middle dense matching point set into dense matching point subsets corresponding to each triangle in the irregular triangular network;

and circulating each triangle in the irregular triangular network, determining the pixel points which are not successfully matched in the dense matching point subset corresponding to each triangle according to the preset dense matching degree, and determining the pixel points which are not successfully matched as miss-matched points.

In one embodiment, performing nearest neighbor matching on the missed matching point to obtain a target dense matching point set between the early-stage remote sensing image and the late-stage remote sensing image, including:

Determining the distance between the missing matching point and each successfully matched pixel point, and determining a plurality of target pixel points from the successfully matched pixel points according to the distance;

constructing an affine transformation model based on each target pixel point;

and carrying out feature point matching on the missed matching points through the affine transformation model to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

In one embodiment, determining a change region between the early and late remote sensing images based on the set of target dense matching points comprises:

dividing the early-stage remote sensing image into a plurality of image blocks;

determining the number of dense matching points contained in each image partition based on the target dense matching point set;

determining the matching success rate of each image block according to the total pixel number, the invalid pixel number, the dense matching point number and the preset dense matching degree of the early-stage remote sensing image;

and if the matching success rate of the image blocks is smaller than a preset success rate threshold, determining the image blocks as a change area between the front remote sensing image and the rear remote sensing image.

In one embodiment, determining the matching success rate of each image block according to the total number of pixels, the number of invalid pixels, the number of dense matching points and the preset dense matching degree of the early-stage remote sensing image includes:

determining the difference value between the total pixel number and the invalid pixel number of the early-stage remote sensing image;

for each image block, determining the product between the quantity of the dense matching points contained in the image block and the preset dense matching degree;

and determining the ratio of the difference value to the product as the matching success rate of the image block.

In a second aspect, an embodiment of the present invention further provides a remote sensing image change area determining device based on dense matching of a triangle network, including:

the image acquisition module is used for acquiring a pre-remote sensing image and a post-remote sensing image to be processed;

the initial matching module is used for performing characteristic point matching on the early-stage remote sensing image and the later-stage remote sensing image to obtain an initial matching point set between the early-stage remote sensing image and the later-stage remote sensing image;

the dense matching module is used for constructing an irregular triangular network based on the initial matching point set, and carrying out dense matching on pixel points in each triangle in the irregular triangular network to obtain a target dense matching point set between the front-stage remote sensing image and the rear-stage remote sensing image;

And the change area determining module is used for determining a change area between the early-stage remote sensing image and the later-stage remote sensing image based on the target dense matching point set.

In a third aspect, an embodiment of the present invention further provides an electronic device comprising a processor and a memory storing computer-executable instructions executable by the processor to implement the method of any one of the first aspects.

In a fourth aspect, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions which, when invoked and executed by a processor, cause the processor to implement the method of any one of the first aspects.

According to the method and the device for determining the change area of the remote sensing image based on the dense matching of the triangular network, after the front-stage remote sensing image and the rear-stage remote sensing image to be processed are obtained, characteristic point matching is firstly carried out on the front-stage remote sensing image and the rear-stage remote sensing image, and an initial matching point set between the front-stage remote sensing image and the rear-stage remote sensing image is obtained; then an irregular triangular network is constructed based on the initial matching point set, and pixel points in each triangle in the triangular network are subjected to dense matching to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image; and finally, determining a change area between the early-stage remote sensing image and the later-stage remote sensing image based on the target dense matching point set. The method utilizes the characteristics that the areas without change can be successfully matched and the areas with change cannot be successfully matched to provide a remote sensing image change area determination based on triangle network dense matching, and can rapidly and stably judge the change area range of the remote sensing image for refined automatic change detection or artificial judgment detection; in addition, the embodiment of the invention does not need to carry out high-precision registration processing on the front and rear remote sensing images in advance, does not need human intervention and learning and training of a large number of samples, has higher stability and robustness, and can automatically extract the change areas of the multi-temporal remote sensing images in batches.

Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.

In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic flow chart of a remote sensing image change area determining method based on dense matching of a triangular network according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a sliding window according to an embodiment of the present invention;

Fig. 3 is a schematic flow chart of an HOPC algorithm according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of an irregular Delaunay triangulation provided by an embodiment of the present invention;

fig. 5 is a flow chart of another method for determining a change area of a remote sensing image based on dense matching of a triangle network according to an embodiment of the present invention;

fig. 6 is a schematic structural diagram of a remote sensing image change area determining device based on dense matching of a triangular network according to an embodiment of the present invention;

fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.

Detailed Description

For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

At present, the related art provides a remote sensing image change detection method as follows:

1. method and system for detecting remote sensing image change: the invention discloses a method and a system for detecting remote sensing image change. Wherein the method comprises the steps of: dividing the remote sensing image of the next time by taking the vector image spots corresponding to the remote sensing image of the previous time as the reference to obtain the divided image spots of the remote sensing image of the next time; extracting all or part of spectral features and texture features of the remote sensing image area corresponding to the previous time phase of the segmented image spots to form a previous space vector; extracting features of the selected segmentation map spots corresponding to the features of the previous space vector to form a subsequent space vector feature, and constructing a subsequent space vector according to the subsequent space vector feature; and judging whether the selected segmentation pattern spots change or not by comparing the latter space vector with the former space vector. The detection method has the advantages of small salt and pepper noise and high detection precision.

2. A multi-temporal multi-spectral remote sensing image change detection method and system: the invention discloses a multi-temporal multi-spectral remote sensing image change detection method and a system, wherein the method firstly utilizes a non-negative matrix factorization to fuse a change vector amplitude of the multi-temporal remote sensing image and a multi-temporal spectral angle map to acquire a new difference image. Then, FCM algorithm is applied to the difference image to obtain unitary energy item of CRF. And secondly, obtaining a binary energy term of the CRF according to the neighborhood of the image and the difference image. And finally, obtaining a final change detection result by minimizing the energy of the CRF through a cyclic belief propagation algorithm. The invention can better describe the relation between the image neighborhoods, and improves the precision of the change detection; the change detection result is more reliable and more robust.

3. A remote sensing image change detection method based on a twin convolutional neural network comprises the following steps: the invention discloses a remote sensing image change detection method based on a twin convolutional neural network, which relates to the field of remote sensing and mainly solves the problem of poor generalization of the conventional change detection method at present; the method comprises the following steps: obtaining multi-time-phase remote sensing image data, obtaining a mask image, establishing a remote sensing image change detection data set, constructing a twin convolutional neural network model, training the twin convolutional neural network by utilizing the data set, obtaining a training model, detecting changes of a pre-time-phase image and a post-time-phase image to be detected by utilizing the training model, obtaining a preliminary change prediction result, comparing a prediction value of a pixel of the preliminary change prediction result with a preset pixel threshold value, and dividing the preliminary change prediction result into a change region category and a non-change region category, thereby obtaining a change detection result. The method has better generalization performance, simultaneously satisfies end-to-end treatment and is convenient for engineering application.

However, the remote sensing image change detection method has the problems of higher image requirements, lower efficiency, poorer universality and the like, and based on the method and the device, the method and the device for determining the remote sensing image change region based on the dense matching of the triangular network are applied to multiband images, are also applied to single-band images, have higher stability and robustness, and can automatically extract the change region of the multi-temporal remote sensing image in a batch mode.

For the convenience of understanding the present embodiment, first, a method for determining a change area of a remote sensing image based on dense matching of a triangular network disclosed in the present embodiment will be described in detail, referring to a schematic flow chart of a method for determining a change area of a remote sensing image based on dense matching of a triangular network shown in fig. 1, the method mainly includes steps S102 to S108 as follows:

step S102, acquiring a pre-remote sensing image and a post-remote sensing image to be processed.

The front-stage remote sensing image and the rear-stage remote sensing image can be multiband images or single-band images, and the rear-stage remote sensing image can be understood as a remote sensing image of the next moment of the front-stage remote sensing image.

And step S104, performing feature point matching on the early-stage remote sensing image and the later-stage remote sensing image to obtain an initial matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

The initial feature point matching refers to automatically matching a certain number of homonymous points on two-period remote sensing images, wherein the homonymous points are initial matching points (initial matching points for short) for the next step of triangle network construction.

In one embodiment, a plurality of different feature point matching algorithms can be adopted to comprehensively perform initial feature point matching on the early-stage remote sensing image and the later-stage remote sensing image so as to improve the precision of an initial matching point set.

And S106, constructing an irregular triangular network based on the initial matching point set, and performing dense matching on pixel points in each triangle in the irregular triangular network to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

The irregular triangular net comprises a plurality of triangles, and the vertex of each triangle is an initial matching point in the initial matching point set.

In one embodiment, the initial set of matching points may be input to a preset tool, such that an irregular triangle is constructed based on the initial set of matching points using the preset tool; for each triangle in the irregular triangle network, further performing feature point matching on pixel points in the triangle to realize dense matching based on the irregular triangle network, so as to obtain an initial dense matching point set; and detecting the missed matching points of the initial dense matching point set, and performing nearest matching on the missed matching points to obtain a target dense matching point set.

Step S108, determining a change area between the early-stage remote sensing image and the later-stage remote sensing image based on the target dense matching point set.

Wherein the variation region may be a variation rectangular region.

In one embodiment, the characteristics that the area without change can be successfully matched and the area with change cannot be successfully matched are utilized to divide the early-stage remote sensing image into a plurality of image blocks, the matching success rate of each image block is determined based on the target dense matching point set, and if the matching success rate of a certain image block is lower than a preset success rate threshold value, the image block is indicated to be the change area.

The remote sensing image change area determination method based on the triangular network dense matching provided by the embodiment of the invention utilizes the characteristics that the area without change can be successfully matched and the change area can not be successfully matched, and provides the remote sensing image change area determination based on the triangular network dense matching, so that the change area range of the remote sensing image can be rapidly and stably determined for refined automatic change detection or artificial judgment detection; in addition, the embodiment of the invention does not need to carry out high-precision registration processing on the front and rear remote sensing images in advance, does not need human intervention and learning and training of a large number of samples, has higher stability and robustness, and can automatically extract the change areas of the multi-temporal remote sensing images in batches.

In order to facilitate understanding, the embodiment of the invention provides a specific implementation mode of a remote sensing image change area determining method based on dense matching of triangular networks.

For the foregoing step S104, the embodiment of the present invention provides an implementation manner for performing feature point matching on the early-stage remote sensing image and the late-stage remote sensing image to obtain an initial matching point set between the early-stage remote sensing image and the late-stage remote sensing image. The initial feature point matching is to automatically match a certain number of homonymous points on two-period remote sensing images, and the homonymous points are used for the next step of triangle network construction. There are many methods for automatically matching the same name points of images, and there are a correlation coefficient matching method, a SIFT (Scale-invariant feature transform, scale invariant feature transform) matching method, and a phase consistency method (HOPC, histogram of Orientated Phase Congruency) for supporting automatic matching of heterogeneous images. In order to improve the processing efficiency and the matching precision, the three matching algorithms are generally based on the strategy idea of pyramid step-by-step matching. The correlation coefficient matching method and the SIFT matching method can be applicable to automatic matching of the homologous remote sensing images, the efficiency is high, the phase consistency method can be used for automatically matching the homologous remote sensing images and also can be used for automatically matching the heterologous remote sensing images, and the matching success rate is high, and the defects of low efficiency and many error points are overcome.

If registration is not performed on the front-stage remote sensing image and the initial precision is unclear, direct point-by-point dense matching with a larger search radius R is very time-consuming, so that the embodiment of the invention automatically matches a certain number of homonymous points in the front-stage remote sensing image and the rear-stage remote sensing image, the homonymous points can also be called as initial matching points, and the initial matching points can be matched with the initial matching points by using the larger search radius to process the condition of larger initial precision deviation. The precision of the initial matching point has a critical influence on the subsequent treatment, so that the precision of the initial matching point is improved as much as possible, and the initial matching point is ensured to be a correct point.

The embodiment of the invention combines the advantages and disadvantages of a plurality of methods, and the matching initial matching point adopts a pyramid-based correlation coefficient and phase consistency matching method, so that each characteristic point can be finally considered as successful matching only if the two methods are matched at the same time, thereby improving the accuracy of the initial matching point. The matching parameter searching radius R can be set to be 200, the characteristic points are uniformly distributed in the overlapping area of the front and back remote sensing images, and the characteristic point number N can be determined according to the image size. The method can be set in equal proportion according to the fact that 1 ten thousand feature points are extracted from every 1 hundred million pixels, the more the feature points are, the slower the efficiency is, and the less the feature points are, the less dense the triangle net is constructed.

The embodiment of the invention provides a pyramid-based correlation coefficient and phase consistency matching method, and an implementation mode of constructing an initial matching point set is provided, wherein the implementation mode is as follows, see step 1 to step 4:

step 1, constructing a pre-pyramid image corresponding to a pre-remote sensing image and a post-pyramid image corresponding to a post-remote sensing image; wherein, earlier stage pyramid image and later stage pyramid image all include a plurality of image layers.

In specific implementation, the pyramid image is a structure for explaining the image with multiple resolutions, and taking the previous remote sensing image as an example, N image layers with different resolutions are generated by performing multi-scale pixel sampling on the previous remote sensing image. The image layer with the highest level resolution is placed at the bottom and is arranged in a pyramid shape, and a series of image layers with gradually reduced pixels (sizes) are arranged upwards until the image layer with only one pixel point is arranged at the top of the pyramid, so that the pyramid image of the early-stage remote sensing image is formed. Similarly, a later pyramid image corresponding to the later remote sensing image can be constructed, and the embodiment of the invention will not be described herein.

Alternatively, a 2 x 2 early pyramid image of the early remote sensing image may be constructed according to the above method, and a 2 x 2 late pyramid image of the late remote sensing image may be constructed according to the above method.

And 2, extracting a plurality of feature points to be matched from the front-stage remote sensing image.

In one embodiment, the Moravec algorithm may be used to uniformly extract N feature points to be matched on the early-stage remote sensing image.

And 3, traversing each image layer of the front pyramid image and each image layer of the rear pyramid image for multiple times through sliding windows with different sizes, so as to match the characteristic points of each characteristic point to be matched by utilizing a plurality of characteristic point matching algorithms in each traversing process, and obtaining a matching point set corresponding to each traversing process.

Wherein, the sliding window can include a small-size sliding window, a medium-size sliding window and a large-size sliding window, and it should be noted that the large, medium and small are merely relative concepts, and in specific implementation, sliding windows with different sizes can be configured based on actual requirements, which is not limited by the embodiment of the present invention; the feature point matching algorithm can be a correlation coefficient matching method and a phase consistency method.

In consideration of any matching method, if the full search is based on the image, the method is time-consuming and labor-consuming and is easy to generate errors, and the method for generating the pyramid image layer with the image resolution from low to high correspondingly forms a coarse-to-fine image matching scheme, so that the method is the best matching method commonly used at present. Therefore, the embodiment of the invention can circularly search the point with the largest correlation with each feature point to be matched on the corresponding position of the later remote sensing image by using the pyramid matching method.

For example, first, matching feature points to be matched in each image layer by using a small-size sliding window, adopting a correlation coefficient matching method and a phase consistency method by a matching algorithm, and reserving the matching result when the matching results of the two algorithms are consistent, so as to obtain a matching point set traversed at the time; similarly, continuously using the middle-size sliding window to match the feature points to be matched in each image layer to obtain a matching point set traversed at the time; and similarly, continuously using a large-size sliding window to match the feature points to be matched in each image layer to obtain a matching point set traversed at the time.

Further, an embodiment of the present invention provides an implementation manner of traversing by using a sliding window, and refer to a schematic diagram of a sliding window principle shown in fig. 2. Reading a memory block f with the expansion radius of R1 (with the size of (2R1+1) ×2R1+1) by taking a previous image characteristic point PT1 as the center, calculating the pixel coordinate PT2 of the characteristic point in a later image according to the principle that geographic coordinates are approximately the same, and taking the point PT2 as the center to expand the memory block g with the radius of R2 (with the size of (2R2+1) ×2R2+1), wherein R1 is called a template radius, R2 is called a search radius, R2 is required to be larger than R1, so that f can traverse the searched g image through window sliding, and searching for a pixel with the maximum correlation coefficient by utilizing correlation comparison, wherein the larger the searched radius R2 is, the slower the matching speed is; the smaller the template radius, the faster the matching speed, which affects the size of the template and the size of the search window. In addition, the reliability of matching is enhanced with the increase of the search window, but if there is a significant difference between two matched images, a large search window may instead reduce the matching accuracy, even the matching is erroneous. In the case of too small a search window, a mismatch may occur if there is insufficient feature information in the region to be matched. The window size has a large impact on the match success rate. The size of the template can be basically determined according to an empirical value, and the size of the search window is difficult to find a universal suitable size according to different image conditions.

In practical application, in order to achieve the problem of speed and precision, a method of using a dynamic window in a pyramid image is adopted to perform matching operation. Specifically, at least three windows from small to large are used for matching feature points to be matched in each layer of pyramid image, the successful matching is considered only under the condition that the matching results are basically consistent each time, otherwise, the sliding window is enlarged to a certain specified limit value in a stepping mode, and the correct matching size can be achieved.

Further, the embodiment of the invention also provides specific logic of a correlation coefficient matching method and a phase consistency method respectively:

and (one) a correlation coefficient matching method:

the primary measure used for automatic matching is the correlation coefficient. The correlation coefficient matching method uses a correlation coefficient (normalized covariance) as a similarity measure. In statistics, the correlation coefficient is used to represent the correlation between two random variables, and extends into image matching to represent the degree of similarity between two images of the same size.

R (X, Y) is referred to as the correlation coefficient of the two images. Wherein E (X), E (Y) is the gray level average value of the two images, D (X), D (Y) is the variance of the two images, E (XY) is the average value of the two images after the corresponding points are multiplied, and the definition of the E (X, Y) is the same as that of the average value in the general statistical theory.

The correlation coefficient has the following properties:

①R(X,Y)＝R(Y,X)；

②|R(X,Y)|≤1；

(3) the filling condition of R (X, Y) |=1 is that images X and Y are linearly related by 1.

It can be seen that the correlation coefficient R (X, Y) | represents the degree of similarity of the linear relationship between the images X and Y, and the more the correlation coefficient is close to 1 or-1, the more the degree of linear similarity between the images is apparent.

For image positioning application, the coordinate corresponding to the maximum correlation coefficient in the correlation coefficient matrix is the positioning coordinate of the small image or the image point on the large image.

Although the correlation coefficient matching method has high precision, the calculation amount is relatively large, so that the correlation coefficient matching method is not independently applied to real-time image matching, calculation improvement and auxiliary condition constraint are required, the calculation amount is reduced, and the calculation requirement of a real-time image matching system is met. The main measure is to simplify the correlation coefficient formula on the basis of not affecting the correlation coefficient as similarity measurement performance, and to realize approximate calculation to improve the operation speed and reliability.

According to the statistical theory, the calculation formula of the correlation coefficient should be:

the simplified formula is:

wherein f is the image data of the earlier image taking the feature point as the center, and g is the image data of the corresponding search area of the later image. Average value sum of g in primary matching process The constant can be calculated first, and the formula calculation is substituted in the subsequent matching, so that the repetition is avoided.

(II) phase consistency method:

the heterogeneous image matching technology based on phase consistency comprehensively utilizes the intensity and direction information of the phase consistency to construct a descriptor for representing the geometric structural characteristics of the image, and the phase consistency direction histogram (histogram of orientated phase congruency, HOPC) characteristic descriptor can acquire the structural attributes of the image. In addition, a similarity measure called HOPCncc is used for multi-modal registration using orthogonalized correlation coefficients (NCCs) of HOPC descriptors.

The HOPC links together the template window into a final feature description vector by dividing it into blocks and counting the phase consistency direction histogram for each block. Referring to a schematic flow chart of an HOPC algorithm shown in fig. 3, the HOPC extraction process mainly includes the following steps:

(1) and selecting a template window with a certain size from the image.

(2) And calculating the phase consistency intensity value and the direction of each pixel in the template window, and providing characteristic information for the construction of the HOPC.

(3) The template window is divided into a plurality of blocks, wherein each block comprises a plurality of cells, and a basic structure of the HOPC is formed.

(4) Calculating phase consistency direction histograms of the block and the cell, and carrying out normalization processing to eliminate the influence of illumination change.

(5) The gradient direction histogram vectors within all blocks are collected together to form the HOPC feature vector describing the entire template window.

HOPC is a feature descriptor that calculates the internal structure of an image. Since the structural properties are more independent of the image gray scale distribution, this descriptor can be used for matching two images with similar shapes and significant nonlinear radiation differences. Thus, the NCC of HOPC descriptors can be used as a similarity measure (called HOPCncc) for image matching, defined as:

where VA and VB are HOPC descriptors for image area A and image area B.And->Mean values of VA and VB are shown, respectively.

In the process of template matching, a template window moves pixel by pixel in a scene image or a search area, and HOPCncs of each pair of template windows to be matched are calculated. This requires many repeated computations due to the large portion of pixels overlapping in adjacent template windows, which can be resolved using the fast matching mechanism of HOPCncc.

The homonymy point detection using HOPCncc includes two steps: the NCC between the HOPC descriptor and the computation descriptor is extracted. The first step takes a lot of time in the matching process. To extract the HOPC descriptors, the template window is divided into overlapping blocks, and the descriptors of these blocks are computed to form the final dense descriptors. Thus, a block can be used as the basic element of the HOPC descriptor. In order to reduce the computation time of template matching, it is necessary to define the block region of each pixel in the image at the center and extract the HOPC descriptor of each block. Each pixel will have a block HOPC descriptor, which in turn forms a three-dimensional descriptor for the whole image, called block-HOPC image. Block-HOPC descriptors are then collected at intervals of a few pixels to obtain HOPC descriptors for the template window.

In order to ensure the accuracy of the initial matching points, each feature point to be matched needs to be successfully identified as the initial matching point by the correlation coefficient method and the HOPC phase correlation matching method at the same time, and the final matching point format is (x) ₁ ,y ₁ ,x ₂ ,y ₂ )，x ₁ ,y ₁ Line number, x representing the earlier image ₂ ,y ₂ Column and row numbers representing the later images.

And 4, for each feature point to be matched, if the matching results of the feature points to be matched are consistent in the matching point set corresponding to each traversal process, the matching results of the feature points to be matched and the matching feature points are reserved to obtain an initial matching point set.

The matching result of a certain feature point to be matched is assumed to be uniform in a matching point set after traversing sliding windows with three sizes, namely a large size, a medium size and a small size, so that the accuracy of the matching result of the feature point to be matched is higher, and the feature point to be matched and the matching result thereof can be stored as an initial matching point set.

The embodiment of the invention provides a preliminary matching point concept, ensures the precision of the preliminary matching point and reduces the probability of the misplacement on the premise that the existing two matching methods pass through simultaneously. Because the initial precision of the images in the front and rear stages is not necessarily good, the stability of the initial matching point matching is ensured by pyramid step-by-step matching and large searching radius.

For the foregoing step S106, the embodiment of the present invention provides an implementation manner of constructing an irregular triangle network based on the initial matching point set, which may be accomplished by using a CGAL library, and through experiments, the network construction efficiency and the number of the CGAL library are better than those of other libraries, where the irregular triangle network may be an irregular Delaunay triangle network, such as a schematic diagram of an irregular Delaunay triangle network shown in fig. 4.

For the foregoing step S106, the embodiment of the present invention further provides an implementation manner of densely matching the pixel points inside each triangle in the irregular triangle network to obtain the target dense matching point set between the front remote sensing image and the rear remote sensing image, which is referred to in the following steps a to c:

and a, densely matching pixel points in each triangle in the irregular triangle network to obtain an initial densely matched point set between the early-stage remote sensing image and the later-stage remote sensing image.

Each triangular net consists of three primary matching points, and the three points can just calculate affine transformation parameters, and the affine transformation function is linear transformation from one two-dimensional coordinate (the early remote sensing image) to another two-dimensional coordinate (the later remote sensing image), and affine transformation coefficients, namely 6 parameters, are calculated by utilizing primary matching point data, and at least three primary matching points are needed for calculation.

The affine transformation model formula is as follows:

f(x)＝a ₀ x+a ₁ y+a ₂

f(y)＝b ₀ x+b ₁ y+b ₂

the above formula describes affine transformation relation of two-dimensional points x, y to f (x), f (y), six parameters a of affine transformation ₀ 、a ₁ 、a ₂ 、b ₀ 、b ₁ 、b ₂ An object to be solved for the system of equations.

Because the initial matching points already establish the coordinate affine transformation relationship of the early image and the later image, the dense matching method in the triangular network is different from the initial matching points in matching:

(1) The matching method can only use one matching algorithm of the correlation coefficient or the phase consistency, so that the matching time can be saved;

(2) The most time-consuming factor, namely the search radius R, can be adjusted to be 30 or smaller, so that the precision of three primary distribution points of the triangular network is ensured, and the initial precision difference between the early-stage image and the later-stage image is not required to be considered;

(3) The dense matching degree can be provided with a parameter P, the pixel-by-pixel matching is 1, the pixel-by-pixel matching is 2, the more dense the matching is, the more accurate the final change area determination is, but the longest the matching time is, and a value can be integrated according to the requirement and the calculation time length requirement.

Taking pixel-by-pixel dense matching as an example, each pixel (x, y) in the triangular network can calculate initial pixel coordinates (f (x, f (y)) of the corresponding later image of the pixel point through affine transformation parameters calculated by corresponding triangles, reading data f with the pixel (x, y) as a center expansion radius of R1 of the earlier image and data g with the corresponding pixel (f (x, f (y)) as a center expansion radius of R2 of the later image, wherein R1 is required to be smaller than R, so that f can slide on g to calculate a correlation coefficient, and finally, calculating the position coordinates of f in g when the maximum value of the correlation coefficient is calculated, so as to obtain the matching point of the pixel. Affine transformation parameters of the triangle mesh provide a relatively accurate initial matching coordinate range, while correlation coefficient matching solves for accurate coordinates.

According to the embodiment of the invention, the irregular triangular network is built by using the initial matching points, the three points are utilized to determine the surface-to-surface coordinate conversion relation (namely, affine transformation model), the pixels in each triangle are subjected to dense matching, and the affine transformation model built by the initial matching points can be used for accurately positioning the searching position of the later remote sensing image, so that the matching in the triangle can be realized by using a single correlation coefficient method, a pyramid is not needed, the searching radius can be small, and the matching efficiency can be greatly accelerated.

If no initial matching point and triangle net are constructed, all pixels of the front and rear images with large initial precision difference of the two scenes are in large search radius to finish the intensive matching, so that the time consumed by the intensive matching is huge, and the construction of the initial matching point set and the triangle net is the core content of the embodiment of the invention.

And b, determining affine transformation error values of each dense matching point in the initial dense matching point set, and if the affine transformation error values of the dense matching points are larger than a preset error threshold value, eliminating the dense matching points from the initial dense matching point set to obtain an intermediate dense matching point set.

In one embodiment, after the dense matching of the triangulation is completed, a very dense set of points (x ₁ ,y ₁ ,x ₂ ,y ₂ … …) at which time affine parameters (including a) can be formed according to the preliminary points ₀ 、a ₁ 、a ₂ 、b ₀ 、b ₁ 、b ₂ ) To calculate the error value of each dense matching point and delete the wrong matching point according to the error threshold.

In specific implementation, affine transformation error values of all dense points in the triangle can be calculated by taking each triangle as a unit, wherein the affine transformation error values refer to affine transformation error values of the dense matching points which are obtained by bringing the dense matching points into a formula of an affine transformation model so as to obtain corresponding f (x) values and f (y) values, and then, taking the obtained f (x) values and f (y) values as difference values with the actual f (x) values and f (y) values. Further, comparing the affine transformation error value of the dense matching point with a preset error threshold, if the affine transformation error value is larger than the preset error threshold, removing the dense matching point from the initial dense matching point set, otherwise, reserving the dense matching point, and thus obtaining an intermediate dense matching point set.

And c, detecting missed matching points in the middle dense matching point set, and performing nearest neighbor matching on the missed matching points to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

In practical application, dense matching of the triangular mesh constraint obtains matching points of most pixels, but because affine parameters calculated by the triangular mesh are not necessarily capable of completely representing conversion of the whole triangular area, some matching points are deleted in the last step, and some actually changed area is unsuccessful in matching (the area is the area which is finally needed to be changed), therefore, some pixels are unsuccessful in matching, and further leak-repairing matching processing is needed. The purpose of this step is to extract the area that the front and back remote sensing images cannot be successfully matched to the maximum extent.

The nearest matching of the leak matching point is mainly divided into two steps:

(a) Detecting miss-matched points in the set of intermediate dense matching points, including (a 1) to (a 2):

(a1) Dividing the middle dense matching point set into dense matching point subsets corresponding to each triangle in the irregular triangle network;

(a2) And circulating each triangle in the irregular triangle network, determining the pixels which are not successfully matched in the dense matching point subset corresponding to each triangle according to the preset dense matching degree, and determining the pixels which are not successfully matched as miss-matched points.

In one embodiment, each triangle may be cycled, and according to the set degree of dense matching P, which rows and columns of points are not successfully matched is determined, and the row and column numbers of the pixel are recorded.

For example, if the dense matching degree P is set to 1, each pixel needs to be successfully matched, no successful matching needs to be recorded, if the dense matching degree P is set to 2, every other pixel needs to be successfully matched, if the dense matching degree P is set to x, every x-1 pixel needs to be successfully matched, otherwise, the row number of the pixels which are not successfully matched is recorded.

(b) Nearest neighbor matching is carried out on the missing matching points, and a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image is obtained, wherein the method comprises the following steps of (b 1) to (b 3):

(b1) And determining the distance between the missing matching point and each successfully matched pixel point, and determining a plurality of target pixel points from each successfully matched pixel point according to the distance.

In one embodiment, a plurality of pixels closest to the miss-matched point may be selected as the target pixel from the successfully-matched pixels. Wherein the number of target pixel points is at least 3.

(b2) An affine transformation model is constructed based on each target pixel point.

(b3) And carrying out feature point matching on the missed matching points through an affine transformation model to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

In one embodiment, the affine transformation model may be used to automatically match the missed matching points again, so as to obtain a matching result of the missed matching points, and the matching result of the missed matching points is saved to the middle dense matching point set, so as to obtain the target dense matching point set.

The embodiment of the invention can match the pixel points which can be matched with the point as much as possible by executing the steps of detecting the missed matching point and carrying out nearest matching on the missed matching point, thereby being capable of detecting the changed area more accurately.

For the foregoing step S108, the embodiment of the present invention provides an implementation manner for determining a change area between a pre-remote sensing image and a post-remote sensing image based on a target dense matching point set, which can be seen in the following steps one to four:

step one, dividing the early-stage remote sensing image into a plurality of image blocks.

In one embodiment, (1) the pre-image is segmented, and the segmentation size BLOCKSIZE determines the granularity of the change region extraction, and the smaller the value, the finer the granularity, which can be set according to the requirement, and is generally taken as 512.

And secondly, determining the number of dense matching points contained in each image block based on the target dense matching point set.

And thirdly, determining the matching success rate of each image block according to the total pixel number, the invalid pixel number, the dense matching point number and the preset dense matching degree of the early-stage remote sensing image.

In one embodiment, determining a difference between a total number of pixels N and an invalid number of pixels M of the early-stage remote sensing image; for each image block, determining the product between the number C of dense matching points contained in the image block and the preset dense matching degree P; and determining the ratio of the difference value and the product as the matching success rate of the image blocks.

Specifically, the invalid value of each image block is removed, the matching success rate of each image block is counted, for example, the image block is divided according to 512 times 512, the total pixel number is N is 26144, the invalid value occupies M pixels, the matching success point number is C, the dense matching degree is P, and the matching success rate F of each image block is:

wherein F is a floating point number from 0 to 1.

And step four, if the matching success rate of the image blocks is smaller than a preset success rate threshold, determining that the image blocks are the change areas between the early-stage remote sensing images and the later-stage remote sensing images.

In one embodiment, a change threshold Q is set, where F is greater than Q, and if F is less than Q, the image block is represented as a change region, and all the change regions are output as vectors and stored, that is, a final result of the change region.

The embodiment of the invention uses the characteristics that the non-changed areas can be successfully matched and the changed areas cannot be successfully matched to solve the problem of determining the changed areas of the unregistered front and back remote sensing images by using an automatic matching technology, and is also the core content of the embodiment of the invention.

In summary, the remote sensing image change region determining method based on the dense matching of the triangular network provided by the embodiment of the invention has at least the following characteristics:

(1) The characteristics that the areas without change can be successfully matched and the areas with change cannot be successfully matched are utilized, and an automatic matching technology is used for solving the problem of determining the areas with change of the unregistered front and back remote sensing images;

(2) The initial matching point concept is put forward, and on the premise that the existing two matching methods pass through simultaneously, the precision of the initial matching point is ensured, and the probability of the wrong point is reduced. The initial precision of the images at the front and rear stages is not necessarily good, so that the stability of the initial matching point matching is ensured through pyramid step-by-step matching and large searching radius;

(3) The initial matching points are used for establishing an irregular triangle network, a face-to-face coordinate conversion relation (affine transformation model) can be determined by using the three points, and pixels in each triangle are subjected to intensive matching, and as the affine model constructed by the initial matching points is arranged, the affine model can be used for accurately positioning the searching position of the later image, the matching in the triangle can use a single correlation coefficient method, a pyramid is not needed, the searching radius can be small, and the matching efficiency can be greatly accelerated;

(4) In addition, the missing matching point is detected, and nearest matching is performed on the missing matching point so that pixels which can be matched to the point are matched as much as possible, thereby enabling a changed region to be detected more accurately.

The remote sensing image data, especially satellite remote sensing image data, is very large, a scene image is at least tens square kilometers or even hundreds square kilometers, the change detection requirements of land utilization and other applications are vigorous all the time, and if the change area is judged by manpower, the method is accurate, but the method is incapable of facing massive remote sensing data. The method provided by the embodiment of the invention can greatly liberate both hands and eyes of a person, quickly locate the changed rectangular area of the remote sensing image, and can be used for subsequent manual interpretation or be submitted to other algorithms capable of extracting the changed accurate boundary for post-processing.

For easy understanding, the embodiment of the present invention provides another method for determining a change region of a remote sensing image based on dense matching of a triangular network, referring to fig. 5, which is a schematic flowchart of another method for determining a change region of a remote sensing image based on dense matching of a triangular network, and the method mainly includes steps S502 to S510:

Step S502, initial feature point matching;

step S504, constructing a primary distribution point triangle network;

step S506, dense matching of triangular network constraint;

step S508, the nearest matching of the missing matching points;

step S510, a change rectangular area is determined.

In summary, the method for determining the change area of the remote sensing image based on the dense matching of the triangular network provided by the embodiment of the invention utilizes the characteristics that the area without change can be successfully matched and the change area cannot be successfully matched, and adopts the matching strategies such as the constraint of the triangular network and the like to fully automatically and rapidly extract the change rectangular area of the front and rear remote sensing images, and although the change boundary cannot be accurately extracted, the method has great significance as an algorithm for primarily judging the change range, and can save a great amount of labor cost to search the change area on massive front and rear remote sensing images.

For the method for determining the remote sensing image change area based on the dense matching of the triangular network provided in the foregoing embodiment, the embodiment of the present invention provides a remote sensing image change area determining device based on the dense matching of the triangular network, referring to a schematic structural diagram of the remote sensing image change area determining device based on the dense matching of the triangular network shown in fig. 6, the device mainly includes the following parts:

The image acquisition module 602 is configured to acquire a pre-remote sensing image and a post-remote sensing image to be processed;

the initial matching module 604 is configured to perform feature point matching on the early-stage remote sensing image and the late-stage remote sensing image, so as to obtain an initial matching point set between the early-stage remote sensing image and the late-stage remote sensing image;

the dense matching module 606 is configured to construct an irregular triangle network based on the initial matching point set, and perform dense matching on pixel points inside each triangle in the irregular triangle network, so as to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image;

the change region determining module 608 is configured to determine a change region between the early-stage remote sensing image and the late-stage remote sensing image based on the target dense matching point set.

The remote sensing image change area determining device based on the triangular network dense matching provided by the embodiment of the invention utilizes the characteristics that the area without change can be successfully matched and the change area can not be successfully matched, and provides the remote sensing image change area determining device based on the triangular network dense matching, which can rapidly and stably judge the change area range of the remote sensing image for fine automatic change detection or artificial judgment detection, and the embodiment of the invention is not only suitable for multiband images, but also suitable for single-band images; in addition, the embodiment of the invention does not need to carry out high-precision registration processing on the front and rear remote sensing images in advance, does not need human intervention and learning and training of a large number of samples, has higher stability and robustness, and can automatically extract the change areas of the multi-temporal remote sensing images in batches.

In one embodiment, the preliminary matching module 604 is further configured to:

constructing a front-stage pyramid image corresponding to the front-stage remote sensing image and a rear-stage pyramid image corresponding to the rear-stage remote sensing image; wherein, the front pyramid image and the rear pyramid image both comprise a plurality of image layers;

extracting a plurality of feature points to be matched from the front-stage remote sensing image;

traversing each image layer of the early pyramid image and each image layer of the later pyramid image for multiple times through sliding windows with different sizes, so as to match the characteristic points of each characteristic point to be matched by utilizing a plurality of characteristic point matching algorithms in each traversing process, and obtaining a matching point set corresponding to each traversing process;

and for each feature point to be matched, if the matching results of the feature points to be matched are consistent in the matching point set corresponding to each traversal process, the matching results of the feature points to be matched and the matching feature points are reserved to obtain an initial matching point set.

In one embodiment, dense matching module 606 is further configured to:

dividing the middle dense matching point set into dense matching point subsets corresponding to each triangle in the irregular triangle network;

and circulating each triangle in the irregular triangle network, determining the pixels which are not successfully matched in the dense matching point subset corresponding to each triangle according to the preset dense matching degree, and determining the pixels which are not successfully matched as miss-matched points.

In one embodiment, dense matching module 606 is further configured to:

determining the distance between the missing matching point and each successfully matched pixel point, and determining a plurality of target pixel points from each successfully matched pixel point according to the distance;

Constructing an affine transformation model based on each target pixel point;

and carrying out feature point matching on the missed matching points through an affine transformation model to obtain a target dense matching point set between the early-stage remote sensing image and the later-stage remote sensing image.

In one embodiment, the change region determination module 608 is further configured to:

dividing the early-stage remote sensing image into a plurality of image blocks;

if the matching success rate of the image blocks is smaller than a preset success rate threshold value, determining that the image blocks are the change areas between the early-stage remote sensing images and the later-stage remote sensing images.

for each image block, determining the product between the number of dense matching points contained in the image block and the preset dense matching degree;

and determining the ratio of the difference value and the product as the matching success rate of the image blocks.

The device provided by the embodiment of the present invention has the same implementation principle and technical effects as those of the foregoing method embodiment, and for the sake of brevity, reference may be made to the corresponding content in the foregoing method embodiment where the device embodiment is not mentioned.

The embodiment of the invention provides electronic equipment, which comprises a processor and a storage device; the storage means has stored thereon a computer program which, when executed by the processor, performs the method of any of the embodiments described above.

Fig. 7 is a schematic structural diagram of an electronic device according to an embodiment of the present invention, where the electronic device 100 includes: a processor 70, a memory 71, a bus 72 and a communication interface 73, said processor 70, communication interface 73 and memory 71 being connected by bus 72; the processor 70 is arranged to execute executable modules, such as computer programs, stored in the memory 71.

The memory 71 may include a high-speed random access memory (RAM, random Access Memory), and may further include a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. The communication connection between the system network element and the at least one other network element is achieved via at least one communication interface 73 (which may be wired or wireless), which may use the internet, a wide area network, a local network, a metropolitan area network, etc.

Bus 72 may be an ISA bus, a PCI bus, an EISA bus, or the like. The buses may be classified as address buses, data buses, control buses, etc. For ease of illustration, only one bi-directional arrow is shown in FIG. 7, but not only one bus or type of bus.

The memory 71 is configured to store a program, and the processor 70 executes the program after receiving an execution instruction, where the method executed by the apparatus for flow defining disclosed in any of the foregoing embodiments of the present invention may be applied to the processor 70 or implemented by the processor 70.

The processor 70 may be an integrated circuit chip with signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuitry in hardware or instructions in software in the processor 70. The processor 70 may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but may also be a digital signal processor (Digital Signal Processing, DSP for short), application specific integrated circuit (Application Specific Integrated Circuit, ASIC for short), off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA for short), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. The disclosed methods, steps, and logic blocks in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be embodied directly in the execution of a hardware decoding processor, or in the execution of a combination of hardware and software modules in a decoding processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory 71 and the processor 70 reads the information in the memory 71 and in combination with its hardware performs the steps of the method described above.

The computer program product of the readable storage medium provided by the embodiment of the present invention includes a computer readable storage medium storing a program code, where the program code includes instructions for executing the method described in the foregoing method embodiment, and the specific implementation may refer to the foregoing method embodiment and will not be described herein.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, comprising several instructions for causing a computer device (which may be a personal computer, a server, a network device, etc.) to perform all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (RAM, random Access Memory), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

Finally, it should be noted that: the above examples are only specific embodiments of the present invention, and are not intended to limit the scope of the present invention, but it should be understood by those skilled in the art that the present invention is not limited thereto, and that the present invention is described in detail with reference to the foregoing examples: any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or perform equivalent substitution of some of the technical features, while remaining within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention, and are intended to be included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A remote sensing image change area determining method based on triangle network dense matching is characterized by comprising the following steps:

2. The method for determining a change area of a remote sensing image based on dense matching of a triangular network according to claim 1, wherein performing feature point matching on the early remote sensing image and the late remote sensing image to obtain an initial matching point set between the early remote sensing image and the late remote sensing image comprises:

3. The method for determining a change area of a remote sensing image based on dense matching of a triangular network according to claim 1, wherein performing dense matching on pixel points inside each triangle in the irregular triangular network to obtain a set of target dense matching points between the early-stage remote sensing image and the late-stage remote sensing image comprises:

4. The method for determining a change area of a remote sensing image based on dense matching of a triangle network according to claim 3, wherein detecting a missing matching point in the intermediate dense matching point set comprises:

5. The method for determining a change area of a remote sensing image based on dense matching of a triangular network according to claim 3, wherein performing nearest neighbor matching on the missed matching points to obtain a set of target dense matching points between the early remote sensing image and the late remote sensing image comprises:

constructing an affine transformation model based on each target pixel point;

6. The method for determining a change area of a remote sensing image based on dense matching of a triangular network according to claim 1, wherein determining a change area between the early remote sensing image and the late remote sensing image based on the set of target dense matching points comprises:

dividing the early-stage remote sensing image into a plurality of image blocks;

7. The method for determining a change area of a remote sensing image based on dense matching of a triangle network according to claim 6, wherein determining a matching success rate of each image block according to a total number of pixels, a number of invalid pixels, a number of dense matching points and a preset dense matching degree of the previous remote sensing image comprises:

8. The utility model provides a remote sensing image change area determining device based on dense matching of triangle net which characterized in that includes:

9. An electronic device comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor, the processor executing the computer-executable instructions to implement the method of any one of claims 1 to 7.

10. A computer readable storage medium storing computer executable instructions which, when invoked and executed by a processor, cause the processor to implement the method of any one of claims 1 to 7.