Multi-temporal image matching method based on angle and distance hash mapping

Abstract

The invention discloses a multi-temporal image matching method based on angle and distance Hash mapping, which belongs to the technical field of multi-temporal image matching, and comprises the following steps: S1, obtaining a feature point set of a reference image and a to-be-matched image; s2, obtaining feature descriptors of the reference image and the to-be-matched image; s3, rough matching between the reference image and the image to be matched; and S4, fine matching between the reference image and the to-be-matched image. According to the method, the feature descriptors of SIFT are improved, and Hash mapping constraints of angles and distances are introduced in a descriptor matching stage, so that a multi-temporal image matching result is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of multi-temporal image matching, and in particular to a multi-temporal image matching method based on angle and distance hash mapping. Background Art

[0002] Multi-temporal image matching is widely used in remote sensing, image fusion, scene matching and other fields. Since multi-temporal images have significant differences in brightness, contrast and texture details, such image matching problems have always been a difficult problem for academic and industry research. Multi-temporal image matching methods are mainly divided into grayscale-based matching and feature-based matching. Grayscale-based matching methods directly or indirectly use grayscale information for matching, such as mutual information-based and self-similarity-based matching methods. Due to different imaging times, there are large differences in the grayscale information of images of the same scene at different times. Therefore, the performance of such methods is significantly reduced when applied to multi-temporal image matching. Feature-based matching methods generate feature descriptors by statistically analyzing the feature information of local areas, and use distance or correlation functions as matching measures for feature matching. They can better achieve stable matching between multi-temporal images. How to find and extract stable and reliable common features between multi-temporal images is the key to achieving accurate matching.

[0003] At present, the method based on SIFT feature, i.e. scale-invariant feature transformation, is a very representative method in feature matching. However, if the imaging time span of multi-temporal images is large and the grayscale difference between images is large, the directional information of the feature descriptors of the same-name points extracted by the SIFT method is prone to inversion, which leads to a certain deviation in the matching results. Therefore, a multi-temporal image matching method based on angle and distance hash mapping is proposed. Summary of the invention

[0004] The technical problem to be solved by the present invention is: how to improve the matching accuracy of multi-temporal images. A multi-temporal image matching method based on angle and distance hash mapping is provided. By improving the SIFT feature descriptor and introducing angle and distance hash mapping constraints in the descriptor matching stage, the multi-temporal image matching results are made more accurate.

[0005] The present invention solves the above technical problems through the following technical solutions, and the present invention comprises the following steps:

[0006] S1: Obtain the feature point set of the reference image and the image to be matched

[0007] The feature point extraction method in the SIFT method is used to obtain the feature point set of the reference image and the image to be matched;

[0008] S2: Obtain feature descriptors of the reference image and the image to be matched

[0009] After obtaining the feature points of the reference image and the image to be matched, the feature descriptor of each feature point is calculated using the HOG (Histogram of Gradients) method;

[0010] S3: Rough matching between the reference image and the image to be matched

[0011] Matching the feature descriptors of the reference image and the image to be matched based on the minimum Euclidean distance to obtain a rough matching result between the feature points of the reference image and the image to be matched;

[0012] S4: Precise matching between the reference image and the image to be matched

[0013] According to the constraints of angle and distance hash mapping, the coarse matching result between the feature points of the reference image and the image to be matched is optimized to obtain the precise matching result between the feature points of the reference image and the image to be matched.

[0014] Furthermore, in step S1, the process of obtaining the feature point set of the reference image and the image to be matched is specifically as follows: B The SIFT method is used to obtain the corresponding feature point set P = {p 1, p 2,…, p M}, treat the matching image I in the same way T The SIFT method is used to obtain the corresponding feature point set Q = {q 1, q 2,…, q N}, where M is the total number of feature points in the reference image, and N is the total number of feature points in the image to be registered.

[0015] Furthermore, in step S2, the process of obtaining the feature descriptors of the reference image and the image to be matched is specifically as follows: traversing the reference image I B Each feature point p in the feature point set P i , i∈[1,M]: p i As the center, r is the radius, select image block I Bi , and the HOG method is used to calculate the image block I Bi The HOG information is used as the feature point p i The feature descriptor V i , V i is an F-dimensional vector, that is, V i =[v i1 ,v i2 ,…,v iF ]; Similarly, traverse the image to be matched I T Each feature point q in the feature point set Q j , j∈[1,N]: with q jAs the center, r is the radius, select image block I Tj , and the HOG method is used to calculate the image block I Tj The HOG information is used as the feature point q j The feature descriptor W j , W j is an F-dimensional vector, that is, W j =[w j1 ,w j2 ,…,w jF ].

[0016] Furthermore, in step S3, the specific processing process is as follows:

[0017] S31: Traverse the reference image I B Each feature point p i , calculate its matching with the image I according to the following formula T Each feature point q j The distance between d(i,j):

[0018]

[0019] Among them, sqrt(.) represents the square root operation;

[0020] S32: After obtaining the distances d(i,1), d(i,2), ..., d(i,N), calculate the feature point subscript j* corresponding to the minimum distance, that is:

[0021] j*=argmin[d(i,j)];

[0022] S33: Feature point p i The corresponding matching feature point in the image to be matched is q j* , which is denoted as s i .

[0023] Furthermore, in step S4, the specific processing process is as follows:

[0024] S41: Based on the reference image I B Each feature point p i The corresponding matching feature point s in the image to be matched i , calculate p i With s i The slope k of the straight line between i , according to the slope k of the straight line i Calculate the corresponding angle θ i , using a non-negative integer β as the angle interval, divide the angle interval [-90°, 90°] to obtain a series of angle subintervals U n ;

[0025] S42: To establish an angle hash map H(U n ), first H(U n ) is initialized to 0, and then traverses each angle θ i , for H(U n ) is updated, and H(U1), H(U2), …, H(U z ), calculate the angle sub-interval subscript n* corresponding to the maximum value, and obtain the reference image I B The angle corresponding to the feature point is in the interval U n* The point set C in

[0026] S43: Base image I B Feature point c in point set C l The matching feature point in the image to be matched is denoted as a l , where l∈[1,L], a1, a2, …, a L The point set composed is denoted as A, and c is calculated. l with a l The distance between l , calculate the distance lower limit μ1 and the distance upper limit μ2 according to the formula, divide the distance interval [μ1,μ2] into a series of distance subintervals Φ using a non-negative integer δ as the distance interval m ;

[0027] S44: To establish a distance hash map Ψ(Φ m ), first replace Ψ(Φ m ) is initialized to 0, and then traverses each distance γ l , for Ψ(Φ m ) is updated, and Ψ(Φ1), Ψ(Φ2), …, Ψ(Φ ξ ), calculate the angle sub-interval subscript m* corresponding to the maximum value, and obtain the reference image I B The distance corresponding to the feature point is in the interval Φ m* The point set Λ within;

[0028] S45: Record with reference image I B The feature point set Λ in the image to be matched corresponds to Θ, and the obtained feature point set Λ and feature point set Θ are the precise matching results between the feature points of the reference image and the image to be matched.

[0029] Furthermore, in step S41, the straight line slope k i The calculation formula is as follows:

[0030] k i =(p i (y)-s i (y)) / (p i(x)-s i (x)

[0031] Among them, p i (x), p i (y) represent the feature points p i Column coordinates, row coordinates, s i (x), s i (y) represents the feature point s i Column coordinates and row coordinates of

[0032] Angle θ i The calculation formula is as follows:

[0033] θ i =arctan(k i );

[0034] Angle subinterval U n It is expressed as follows:

[0035] U n =[-90°+(β+1)*(n-1),-90°+(β+1)*(n-1)+β]∩[-90°,90°]

[0036] Where n is the angle subinterval number, and n∈[1,z], z is the largest positive integer not exceeding 181 / (β+1).

[0037] Furthermore, in step S42, H(U n ) is updated as follows:

[0038] H(U n )=H(U n )+1,ifθ i ∈U n ;

[0039] The calculation formula for the angle subinterval subscript n* is as follows:

[0040] n*=argmax[H(U n )];

[0041] The point set C is represented as follows:

[0042] C=[p o |, o∈[1,M] and p o ∈P,θ o ∈U n* ]

[0043] And the point set C is abbreviated as:

[0044] C=[c1,c2,…,c L ]

[0045] Where L is the total number of elements contained in the point set C.

[0046] Furthermore, in step S43, the distance γ l The calculation formula is as follows:

[0047] γ l =sqrt([c l (y)-a l (y)] 2 +[c l (x)-a l (x)] 2 )

[0048] Among them, c l (x), c l (y) represent the feature points c l The column and row coordinates of a l (x), a l (y) represent feature points a l Column coordinates and row coordinates of

[0049] The calculation formulas for the lower distance limit μ1 and the upper distance limit μ2 are as follows:

[0050] μ1=floor(min(γ1,γ2,…,γ L ))

[0051] μ2=ceil(max(γ1,γ2,…,γ L ))

[0052] Among them, floor() and ceil() represent the rounding down operation and rounding up operation respectively;

[0053] Distance subinterval Φ m It is expressed as follows:

[0054] Φ m =[μ1+(δ+1)*(m-1),μ1+(δ+1)*(m-1)+δ]∩[μ1,μ2]

[0055] Wherein, m is the distance subinterval number, and m∈[1,ξ], ξ is the maximum positive integer not exceeding (μ2-μ1+1) / (δ+1).

[0056] Furthermore, in step S44, Ψ(Φ m ) is updated as follows:

[0057] Ψ(Φ m )=Ψ(Φ m )+1, if γ l ∈Φ m;

[0058] The calculation formula for the angle subinterval subscript m* is as follows:

[0059] m*=argmax[Ψ(Φ m )];

[0060] The point set Λ is expressed as follows:

[0061] Λ=[c h |, h∈[1,L] and c h ∈C,Φ h ∈Φ m* ]

[0062] The point set Λ is simply written as:

[0063] Λ=[λ1,λ2,…,λ E ]

[0064] Where E is the total number of elements contained in the point set Λ.

[0065] Compared with the prior art, the present invention has the following advantages: the multi-temporal image matching method based on angle and distance hash mapping combines the stability of SIFT feature points and the robustness of HOG feature descriptors, so that the extracted feature points and feature descriptors are suitable for matching multi-temporal images; in the precise matching stage, the angle and distance constraints between feature descriptors are introduced to improve the matching accuracy, and the hash mapping method is adopted to make the calculation complexity of the matching only O(N). BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 It is a flowchart of a multi-temporal image matching method based on angle and distance hash mapping in an embodiment of the present invention;

[0067] Figure 2 3 is a comparison diagram of multi-phase optical image matching effects in an embodiment of the present invention, wherein (a) is the traditional SIFT-based matching effect, and (b) is the matching effect based on angle and distance hash mapping of the present invention. DETAILED DESCRIPTION

[0068] The following is a detailed description of an embodiment of the present invention. This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiment.

[0069] This embodiment provides a technical solution: a multi-temporal image matching method based on angle and distance hash mapping, comprising the following steps:

[0070] (1) Obtaining the feature point set of the reference image and the image to be matched

[0071] The feature point extraction method in the SIFT method is used to obtain the feature point set of the reference image and the image to be matched;

[0072] (2) Obtaining feature descriptors of the reference image and the image to be matched

[0073] After obtaining the feature points of the reference image and the image to be matched, the feature descriptor of each feature point is calculated using the HOG (Histogram of Gradients) method;

[0074] (3) Rough matching between the reference image and the image to be matched

[0075] Matching the feature descriptors of the reference image and the image to be matched based on the minimum Euclidean distance to obtain a rough matching result between the feature points of the reference image and the image to be matched;

[0076] (4) Precise matching between the reference image and the image to be matched

[0077] According to the constraints of angle and distance hash mapping, the coarse matching result between the feature points of the reference image and the image to be matched is optimized to obtain the precise matching result between the feature points of the reference image and the image to be matched.

[0078] In this embodiment, in step (1), the process of obtaining the feature point set of the reference image and the image to be matched is as follows: B The SIFT method is used (firstly, an image difference pyramid is established, then extreme points are calculated in the pyramid, and finally, the edge response and extreme point filtering are eliminated) to obtain the corresponding feature point set P = {p 1, p 2,…, p M}, treat the matching image I in the same way T The SIFT method is used to obtain the corresponding feature point set Q = {q 1, q 2,…, q N}, where M is the total number of feature points in the reference image, and N is the total number of feature points in the image to be registered.

[0079] In this embodiment, in step (2), the process of obtaining the feature descriptors of the reference image and the image to be matched is as follows: traverse the reference image I B Each feature point p in the feature point set P i , i∈[1,M]: p i As the center, r is the radius, select image block I Bi , and the HOG method is used (calculating the image block gradient, segmenting the cell unit, constructing the gradient histogram, normalizing, and finally generating the feature descriptor) to calculate the image block I Bi The HOG information is used as the feature point pi The feature descriptor V i , V i is an F-dimensional vector, that is, V i =[v i1 ,v i2 ,…,v iF ]; Similarly, traverse the image to be matched I T Each feature point q in the feature point set Q j , j∈[1,N]: with q j As the center, r is the radius, select image block I Tj , and the HOG method is used to calculate the image block I Tj The HOG information is used as the feature point q j The feature descriptor W j , W j is an F-dimensional vector, that is, W j =[w j1 ,w j2 ,…,w jF ].

[0080] In step (3), the rough matching process between the reference image and the image to be matched is specifically as follows:

[0081] Traverse the benchmark image I B Each feature point p i , calculate its matching with the image I according to the following formula T Each feature point q j The distance between d(i,j):

[0082]

[0083] Among them, sqrt(.) represents the square root operation;

[0084] After obtaining the distances d(i,1), d(i,2), ..., d(i,N), calculate the feature point subscript j* corresponding to the minimum distance, that is:

[0085] j*=argmin[d(i,j)],j∈[1,N]

[0086] At this time, the feature point p i The corresponding matching feature point in the image to be matched is q j* , which is denoted as s i .

[0087] In this embodiment, in step (4), the precise matching process between the reference image and the image to be matched is specifically as follows:

[0088] In the rough matching stage, the reference image I is obtained. B Each feature point pi The corresponding matching feature point s in the image to be matched i , calculate p as follows i With s i The slope k of the straight line between i :

[0089] k i =(p i (y)-s i (y)) / (p i (x)-s i (x)

[0090] Among them, p i (x), p i (y) represent the feature points p i Column coordinates, row coordinates, s i (x), s i (y) represents the feature point s i Column coordinates and row coordinates of

[0091] The slope k of the straight line i Perform inverse tangent calculation to obtain the corresponding angle θ i :

[0092] θ i =arctan(k i ), i∈[1,M]

[0093] Using a non-negative integer β as the angle interval, the angle interval [-90°, 90°] is divided to obtain the following series of angle subintervals U n :

[0094] U n =[-90°+(β+1)*(n-1),-90°+(β+1)*(n-1)+β]∩[-90°,90°]

[0095] Where n is the angle subinterval number, and n∈[1,z], z is the largest positive integer not exceeding 181 / (β+1);

[0096] To create an angle hash map H(U n ), first H(U n ) is initialized to 0, and then traverses each angle θ i , according to the following method H(U n ) to update:

[0097] H(U n )=H(U n )+1,ifθ i ∈U n

[0098] After obtaining H(U1), H(U2), ..., H(U z ), calculate the angle sub-interval subscript n* corresponding to the maximum value, that is:

[0099] n*=argmax[H(U n )], n∈[1,z]

[0100] At this time, the reference image I is obtained by statistics as follows: B The angle corresponding to the feature point is in the interval U n* The point set C inside:

[0101] C=[p o |, o∈[1,M] and p o ∈P,θ o ∈U n* ]

[0102] The point set C is simply written as:

[0103] C=[c1,c2,…,c L ]

[0104] Where L is the total number of elements contained in the point set C;

[0105] The reference image I B Feature point c in point set C l (l∈[1,L]) corresponds to the matching feature point in the image to be matched, denoted as a l ;

[0106] Let a1, a2, ..., a L The point set composed of is denoted as A, and c is calculated as follows l with a l The distance between l :

[0107] γ l =sqrt([c l (y)-a l (y)] 2 +[c l (x)-a l (x)] 2 )

[0108] Among them, c l (x), c l (y) represent the feature points c l The column and row coordinates of a l (x), a l (y) represent feature points a lThe column and row coordinates of ; the lower limit μ1 and the upper limit μ2 of the distance are calculated according to the following formula:

[0109] μ1=floor(min(γ1,γ2,…,γ L ))

[0110] μ2=ceil(max(γ1,γ2,…,γ L ))

[0111] Among them, floor() and ceil() represent the rounding down operation and rounding up operation respectively;

[0112] Taking a non-negative integer δ as the distance interval, the distance interval [μ1,μ2] is divided to obtain the following series of distance subintervals Φ m :

[0113] Φ m =[μ1+(δ+1)*(m-1),μ1+(δ+1)*(m-1)+δ]∩[μ1,μ2]

[0114] Where m is the distance subinterval number, and m∈[1,ξ], ξ is the largest positive integer not exceeding (μ2-μ1+1) / (δ+1);

[0115] To establish the distance hash map Ψ(Φ m ), firstly, Ψ(Φ m ) is initialized to 0, and then traverses each distance γ l , according to the following method to Ψ(Φ m ) to update:

[0116] Ψ(Φ m )=Ψ(Φ m )+1, if γ l ∈Φ m

[0117] In order to find Ψ(Φ1), Ψ(Φ2), ..., Ψ(Φ ξ ), calculate the angle sub-interval subscript m* corresponding to the maximum value, that is:

[0118] m*=argmax[Ψ(Φ m )],m∈[1,ξ]

[0119] At this time, the reference image I is obtained by statistics as follows: B The distance corresponding to the feature point is in the interval Φ m* Point set Λ within:

[0120] Λ=[c h |, h∈[1,L] and c h ∈C,Φ h∈Φ m* ]

[0121] The point set Λ is simply written as:

[0122] Λ=[λ1,λ2,…,λ E ]

[0123] Where E is the total number of elements contained in the point set Λ;

[0124] Let I be the reference image B The feature point set Λ in the image to be matched corresponds to Θ, and the obtained feature point set Λ and feature point set Θ are the precise matching results between the feature points of the reference image and the image to be matched.

[0125] like Figure 2 The figure shows the comparison of multi-phase optical image matching effects. The imaging time of the reference image and the image to be matched is different, and the grayscale information obtained is quite different. The matching effect of the traditional SFIT method (see Figure 2 The deviation of (a) in FIG. 1 is large, while the method of the present invention (see FIG. 1 ) using the improved SIFT and adding angle and distance constraints Figure 2 (b) in the figure can achieve a more accurate matching effect.

[0126] In summary, the multi-temporal image matching method based on angle and distance hash mapping in the above embodiment improves the SIFT feature descriptor and introduces angle and distance hash mapping constraints in the descriptor matching stage, so that the multi-temporal image matching results are more accurate.

[0127] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A multi-temporal image matching method based on angle and distance hash mapping, characterized in that: The following steps are involved: S1: Obtain the feature point set of the reference image and the image to be matched The feature point extraction method in the SIFT method is used to obtain the feature point set of the reference image and the image to be matched; S2: Obtain feature descriptors of the reference image and the image to be matched After obtaining the feature points of the reference image and the image to be matched, the feature descriptor of each feature point is calculated using the HOG (Histogram of Gradients) method; S3: Rough matching between the reference image and the image to be matched Matching the feature descriptors of the reference image and the image to be matched based on the minimum Euclidean distance to obtain a rough matching result between the feature points of the reference image and the image to be matched; S4: Precise matching between the reference image and the image to be matched According to the constraints of angle and distance hash mapping, the coarse matching result between the feature points of the reference image and the image to be matched is optimized to obtain the precise matching result between the feature points of the reference image and the image to be matched.

2. The multi-temporal image matching method based on angle and distance hash mapping according to claim 1 is characterized in that: In step S1, the process of obtaining the feature point set of the reference image and the image to be matched is as follows: B The SIFT method is used to obtain the corresponding feature point set P = {p 1, p 2,…, p M }, treat the matching image I in the same way T The SIFT method is used to obtain the corresponding feature point set Q = {q 1, q 2,…, q N }, where M is the total number of feature points in the reference image, and N is the total number of feature points in the image to be registered.

3. The multi-temporal image matching method based on angle and distance hash mapping according to claim 2 is characterized in that: In step S2, the process of obtaining the feature descriptors of the reference image and the image to be matched is as follows: traverse the reference image I B Each feature point p in the feature point set P i , i∈[1,M]: p i As the center, r is the radius, select image block I Bi , and the HOG method is used to calculate the image block I Bi The HOG information is used as the feature point p i The feature descriptor V i , V i is an F-dimensional vector, that is, V i =[v i1 ,v i2 ,…,v iF ]; Similarly, traverse the image to be matched I T Each feature point q in the feature point set Q j , j∈[1,N]: with q j As the center, r is the radius, select image block I Tj , and the HOG method is used to calculate the image block I Tj The HOG information is used as the feature point q j The feature descriptor W j , W j is an F-dimensional vector, that is, W j =[w j1 ,w j2 ,…,w jF ].

4. The multi-temporal image matching method based on angle and distance hash mapping according to claim 3 is characterized in that: In step S3, the specific processing process is as follows: S31: Traverse the reference image I B Each feature point p i , calculate its matching with the image I according to the following formula T Each feature point q j The distance between d(i,j): Among them, sqrt(.) represents the square root operation; S32: After obtaining the distances d(i,1), d(i,2), ..., d(i,N), calculate the feature point subscript j* corresponding to the minimum distance, that is: j*=argmin[d(i,j)]; S33: Feature point p i The corresponding matching feature point in the image to be matched is q j* , which is denoted as s i .

5. The multi-temporal image matching method based on angle and distance hash mapping according to claim 4 is characterized in that: In step S4, the specific processing process is as follows: S41: Based on the reference image I B Each feature point p i The corresponding matching feature point s in the image to be matched i , calculate p i With s i The slope k of the straight line between i , according to the slope k of the straight line i Calculate the corresponding angle θ i , using a non-negative integer β as the angle interval, divide the angle interval [-90°, 90°] to obtain a series of angle subintervals U n ; S42: To establish the angle hash map H(U n ), first H(U n ) is initialized to 0, and then traverses each angle θ i , for H(U n ) is updated, and H(U1), H(U2), …, H(U z ), calculate the angle sub-interval subscript n* corresponding to the maximum value, and obtain the reference image I B The angle corresponding to the feature point is in the interval U n* The point set C in S43: Base image I B Feature point c in point set C l The matching feature point in the image to be matched is denoted as a l , where l∈[1,L], a1, a2, …, a L The point set composed is denoted as A, and c is calculated. l with a l The distance between l , calculate the distance lower limit μ1 and the distance upper limit μ2 according to the formula, divide the distance interval [μ1,μ2] into a series of distance subintervals Φ using a non-negative integer δ as the distance interval m ; S44: To establish a distance hash map Ψ(Φ m ), first replace Ψ(Φ m ) is initialized to 0, and then traverses each distance γ l , for Ψ(Φ m ) is updated, and Ψ(Φ1), Ψ(Φ2), …, Ψ(Φ ξ ), calculate the angle sub-interval subscript m* corresponding to the maximum value, and obtain the reference image I B The distance corresponding to the feature point is in the interval Φ m* The point set Λ within; S45: Record with reference image I B The feature point set Λ in the image to be matched corresponds to Θ, and the obtained feature point set Λ and feature point set Θ are the precise matching results between the feature points of the reference image and the image to be matched.

6. The multi-temporal image matching method based on angle and distance hash mapping according to claim 5, characterized in that: In step S41, the straight line slope k i The calculation formula is as follows: k i =(p i (y)-s i (y)) / (p i (x)-s i (x)) Among them, p i (x), p i (y) represent the feature points p i Column coordinates, row coordinates, s i (x), s i (y) represents the feature point s i Column coordinates and row coordinates of Angle θ i The calculation formula is as follows: θ i =arctan(k i ); Angle subinterval U n It is expressed as follows: U n =[-90°+(β+1)*(n-1),-90°+(β+1)*(n-1)+β]∩[-90°,90°] Where n is the angle subinterval number, and n∈[1,z], z is the largest positive integer not exceeding 181 / (β+1).

7. The multi-temporal image matching method based on angle and distance hash mapping according to claim 5, characterized in that: In step S42, H(U n ) is updated as follows: H(U n )=H(U n )+1,ifθ i ∈U n ; The calculation formula for the angle subinterval subscript n* is as follows: n*=argmax[H(U n )]; The point set C is represented as follows: C=[p o |, o∈[1,M] and p o ∈P,θ o ∈U n* ] And the point set C is abbreviated as: C=[c1,c2,…,c L ] Among them, L is the total number of elements contained in the point set C.

8. The multi-temporal image matching method based on angle and distance hash mapping according to claim 5, characterized in that: In step S43, the distance γ l The calculation formula is as follows: γ l =sqrt([c l (y)-a l (the)] 2 +[c l (x)-a l (x)] 2 ) Among them, c l (x), c l (y) represent the feature points c l The column and row coordinates of a l (x), a l (y) represent feature points a l Column coordinates and row coordinates of The calculation formulas for the lower distance limit μ1 and the upper distance limit μ2 are as follows: μ1=floor(min(γ1,γ2,…,γ L )) μ2=ceil(max(γ1,γ2,…,γ L )) Among them, floor() and ceil() represent the rounding down operation and rounding up operation respectively; Distance subinterval Φ m It is expressed as follows: F m =[μ1+(δ+1)*(m-1),μ1+(δ+1)*(m-1)+δ]∩[μ1,μ2] Wherein, m is the distance subinterval number, and m∈[1,ξ], ξ is the maximum positive integer not exceeding (μ2-μ1+1) / (δ+1).

9. The multi-temporal image matching method based on angle and distance hash mapping according to claim 5, characterized in that: In step S44, Ψ(Φ m ) is updated as follows: P(F m )=Ψ(Φ m )+1,ifγ l ∈Φ m ; The calculation formula for the angle subinterval subscript m* is as follows: m*=argmax[Ψ(Φ m )]; The point set Λ is expressed as follows: Λ = [c h |, h ∈ [1, L] and c h ∈ C, Φ h ∈ Φ m* ​ The point set Λ is simply written as: Λ=[λ1,λ2,…,λ E ] Where E is the total number of elements contained in the point set Λ.

Images