A method for evaluating the confidence of SAR scene matching results

Abstract

The present invention belongs to scene matching navigation technology, and specifically relates to a SAR scene matching result confidence evaluation method. The method comprises: determining a landmark point search range according to SAR radar parameters and aircraft flight parameters; retrieving a set of landmark points to be matched from a landmark point database according to the landmark point search range; treating each landmark point T in the set of matching landmark points i , the normalized template matching method is used to calculate the pixel coordinates (x i ,y i ) and the corresponding matching degree P i ; Treat each landmark point pair in the matching landmark point set (T i , T j ), calculate the position correlation error ΔL ij ; According to the matching degree and position correlation error, calculate the landmark point pair (T i , T j )’s matching confidence C ij , where Thd is the position correlation deviation threshold. The matching results are cross-validated using the prior information of the mutual position relationship between multiple landmark points, which effectively reduces the probability of landmark point mismatching and improves the reliability of the matching results.

Description

Technical Field

[0001] The present invention belongs to the field of scene matching navigation technology, and particularly relates to a method for evaluating the confidence level of SAR scene matching results. Background Art

[0002] Scene matching navigation uses an airborne image sensor to sense ground terrain information to form a real-time image, performs registration processing with a pre-prepared reference map, and then calculates the accurate position of the carrier aircraft in the earth coordinate system through inverse positioning calculation. It is an important airborne navigation method. Scene matching navigation has the characteristics of strong autonomy, strong anti-interference ability, high navigation accuracy, and non-accumulation of errors over time. It is applicable to carrier aircraft platforms such as high-altitude long-endurance strategic bombers, various types of reconnaissance drones, fighter jets, and helicopters, and has broad application value and application prospects.

[0003] The scene matching navigation system is divided into a visible light scene matching system and an SAR radar scene matching system. The imaging equipment of the visible light matching system is mature, and both the navigation accuracy and the data update rate are relatively high. However, it is greatly affected by weather and light, does not have the ability to work all day and all weather, and has relatively high requirements for the preparation of the reference map. The pre-use preparation work is cumbersome and the maintenance cost is high. It is mainly used in missile-borne guidance systems and low-altitude unmanned aircraft auxiliary navigation systems.

[0004] The SAR radar scene matching navigation system uses an SAR radar as an imaging sensor. The SAR radar is a high-resolution microwave imaging radar that uses the synthetic aperture principle, pulse compression technology, and signal processing methods to obtain two-way high-resolution remote sensing imaging in the range direction and the azimuth direction with a real small-aperture antenna. Compared with general optoelectronic sensors in the visible light band, the SAR radar has the ability to work all day and all weather, and at the same time has the characteristics of a long imaging distance and a wide swath width. It can be applied to various types of high-altitude and high-speed carrier aircraft platforms, as well as various types of fighter jets and unmanned aircraft, and has very strong environmental adaptability. It can greatly improve the performance of the navigation system of the carrier aircraft platform in a satellite denial environment. However, due to the limitations of the SAR radar imaging principle, its images often have a relatively high noise level, resulting in insufficient credibility of the matching results of individual landmark points and unable to be directly provided for the aircraft integrated navigation system. Summary of the Invention

[0005] Object of the Invention: To provide a method for evaluating the confidence level of SAR scene matching results, and solve the problem of insufficient credibility of the matching results of individual landmark points caused by high noise in SAR radar images.

[0006] Technical Solution:

[0007] In a first aspect, a method for evaluating the confidence level of SAR scene matching results is provided, including:

[0008] Determine the landmark point search range according to the SAR radar parameters and the aircraft flight parameters;

[0009] Retrieve the set of landmark points to be matched from the landmark point database according to the landmark point search range;

[0010] For each landmark point T in the set of landmark points to be matched i , calculate the pixel coordinates (x i , y i , y i ) and the corresponding matching degree P i of the matching point of the landmark point T in the airborne SAR radar image, where the set of landmark points to be matched includes n landmark points, and the value of i ranges from 1 to n;

[0011] For each pair of landmark points (T i , T j ) in the set of landmark points to be matched, calculate the position correlation error ΔL ij ;

[0012] Calculate the matching confidence C i , T j ) of the pair of landmark points (T ij , where the value of j ranges from 1 to n and j is not equal to i.

[0013] Furthermore, determining the set of landmark points to be matched specifically includes: determining the landmark point search range according to the SAR radar parameters and the aircraft flight parameters; retrieving the set of landmark points to be matched from the landmark point database according to the search range.

[0014] Furthermore, for each pair of landmark points (T i , T j ) in the set, calculate the position correlation error ΔL ij , specifically including:

[0015] Calculate the coordinates (x i , y j , z ie , y ie , z ie ,) and (x je , y je , z je ,) of the landmark points T

[0016] in the Earth rectangular coordinate system respectively; i Calculate the first relative position vector E j of the landmark points T ij , E ij = (x ie - y je , yie -y je )

[0017] Calculate the landmark point T i and T j The second relative position vector M between the corresponding matching points in the SAR radar image ij where M ij =(x i -y j , y i -y j )·RES, where RES is the pixel resolution of the airborne SAR image, unit: meter / pixel;

[0018] According to the first relative position vector E ij and the second relative position vector M ij Calculate the position correlation error ΔL of the landmark point pair (T i , T j ), ΔL ij where ΔL ij =|E ij -M ij |.

[0019] Furthermore, for each landmark point pair (T i , T j ) in the set, calculate the matching confidence C ij , specifically including:

[0020] According to the matching degrees P i and P j of the landmark points T i and T j , calculate the probability P ij that the matching result is correct under the condition that the relative position relationship of the two landmark points is correct, P i =1-(1-P j )(1-P ij );

[0021] Construct a probability function to calculate the probability that the relative position relationship of the two landmark points is correct when the position correlation error is ΔL where Thd is the position correlation deviation threshold;

[0022] According to the conditional probability formula, calculate the matching confidence C ij as:

[0023] Furthermore, calculate the position correlation error ΔL ij， , specifically including:

[0024] Calculate the landmark points T i and T jCoordinates in the Earth's polar coordinate system where R is the radius of the Earth;

[0025] Calculate the landmark point T i 、T j The first relative position vector E in the horizontal direction of the Earth ij ,

[0026] Calculate the landmark point T i 、T j The second relative position vector M between the corresponding matching points in the SAR radar image ij ,M ij =(x i -y j ,y i -y j )·RES, where RES is the pixel resolution of the airborne SAR image, unit: m / pixel;

[0027] According to the first relative position vector E ij and the second relative position vector M ij Calculate the position correlation error ΔL of the landmark point pair (T i ,T j ), ΔL ij ,ΔL ij =|E ij -M ij |.

[0028] Furthermore, the matching confidence C of the landmark point pair (T i ,T j ) adopts a binary calculation method, specifically ij For further details

[0029] Furthermore, the position correlation deviation threshold Thd is determined according to the specific SAR radar performance and imaging quality, specifically Thd = m·RES + Δh / sin(h / R min ) + Δθ·R min where m is the matching pixel error, Δh is the relative height error, h is the relative height, R min is the minimum working distance, and Δθ is the track angle error.

[0030] Furthermore, the method further includes: for multiple landmark point subsets (T1, 2,...T n ) in the set, calculate the matching confidence C 1-n , specifically:[[]]

[0031] According to the matching degrees (P1, P2,...P of the landmark point subsets (T1, T2,...T n )n ) Calculate the probability that the matching result is correct under the condition that the relative position relationship is correct

[0032] Construct a probability function to calculate the position correlation error ΔL between landmark point 1 and landmark point i (i = 2, 3, … n) 1i The probability that the relative position relationship between the two landmark points is correct when ΔL

[0033] According to the conditional probability formula, calculate the matching confidence C 1-n It is:

[0034] Beneficial effects:

[0035] Aiming at the problem of insufficient confidence in the matching result of a single landmark point in the airborne SAR radar matching navigation system, this method uses the prior information of the mutual position relationship between multiple landmark points to cross-verify the matching result, effectively reducing the probability of incorrect matching of landmark points and improving the reliability of the matching result. At the same time, a method for evaluating the confidence of the matching result is proposed, which can effectively improve the navigation ability of the aircraft platform in the satellite denial environment. Specific implementation manners

[0036] To solve the above problems, the present invention uses the prior information of the mutual position relationship between multiple landmark points to cross-verify the matching result of the SAR radar scene matching system, effectively reducing the probability of incorrect matching of landmark points and improving the reliability of the matching result. Based on the principles of statistical probability, the matching confidence is calculated according to the matching degree of a single landmark point and the position correlation error of the landmark point pair, and the result can be provided to the integrated navigation computer as an evaluation index for the availability of the navigation source.

[0037] Example 1:

[0038] Step 1: Determine the landmark point search range according to the SAR radar parameters and the aircraft flight parameters. Define the landmark point search range for database retrieval.

[0039] Step 2: According to the search range, retrieve the set of landmark points to be matched in the landmark point database. Reduce the amount of computation for matching calculation and reduce the calculation time.

[0040] Step 3: For each landmark point Ti in the set, use the normalized template matching method to calculate the pixel coordinates (x i of the matching point of landmark point T in the airborne SAR radar image i , y i ) and the corresponding matching degree P iPrepare landmark point matching degree and matching coordinate data for position correlation cross-validation. Among them, the set of landmark points to be matched includes n landmark points, and the value of i ranges from 1 to n as an integer.

[0041] Step 4: For each pair of landmark points (T i , T j ) in the set of landmark points to be matched, calculate the position correlation error ΔL ij . Obtaining the position correlation error can intuitively reflect the matching correctness of the landmark point pair and prepare for calculating the matching confidence of the punctuation pair.

[0042] Step 401: Calculate the coordinates (x i , y j , z ie ) and (x ie , y ie , z je ) of the landmark points T je , T je in the Earth rectangular coordinate system respectively.

[0043] Step 402: Calculate the relative position vector Eij of the landmark points Ti, Tj in the Earth horizontal plane direction.

[0044] E ij = (x ie - y je , y ie - y je )

[0045] Step 403: Calculate the relative position vector Mij between the corresponding matching points of the landmark points Ti, Tj in the SAR radar image.

[0046] M ij = (x i - y j , y i - y j ) · RES

[0047] where RES is the pixel resolution of the airborne SAR image, unit: meter / pixel.

[0048] Step 404: Calculate the position correlation error ΔL i , T j ) of the landmark point pair (T ij .

[0049] ΔL ij = |E ij - M ij |

[0050] Step 5: According to the matching degree and the position correlation error, calculate the landmark point pair (Ti , T j ) matching confidence C ij . The value of j ranges from 1 to n and j is not equal to i.

[0051]

[0052] Where Thd is the position correlation deviation threshold, which is determined by the specific SAR radar performance and imaging quality. Specifically,

[0053] Thd = m·RES + Δh / sin(h / R min ) + Δθ·R min

[0054] Where m is the matching pixel error, Δh is the relative height error, h is the relative height, R min is the minimum working distance, and Δθ is the track angle error.

[0055] The matching confidence of the landmark point pair is obtained, which can be provided to the integrated navigation computer as an evaluation index of the availability of the navigation source. Compared with the single landmark point, the reliability of the landmark point pair passing the test is significantly improved.

[0056] Specifically:

[0057] Step 501: According to the matching degrees P i , T j of T i , P j , calculate the probability P ij = 1 - (1 - P i )(1 - P j ) when the relative position relationship between the two landmark points is correct and the matching result is correct;

[0058] Step 502: Construct a probability function to calculate the probability ij that the relative position relationship between the two landmark points is correct when the position correlation error is ΔL

[0059] Step 503: According to the conditional probability formula, calculate the matching confidence C ij as:

[0060] Furthermore, the matching confidence C i of the landmark point pair (T j , T ij can also adopt a binarization calculation method. Specifically,

[0061] Example 2:

[0062] Step 1: Determine the landmark point search range based on SAR radar parameters and aircraft flight parameters. Define the landmark point search range for database retrieval.

[0063] Step 2: Retrieve the set of landmark points to be matched from the landmark point database according to the search range. Reduce the computational workload and calculation time of the matching calculation.

[0064] Step 3: For each landmark point Ti in the set, use the normalized template matching method to calculate the pixel coordinates (x i , y i ) of the matching point in the airborne SAR radar image and the corresponding matching degree P i . Prepare the landmark point matching degree and matching coordinate data for position correlation cross-verification.

[0065] Step 4: For each pair of landmark points (T i , T j ) in the set, calculate the position correlation error ΔL ij . Obtain the position correlation error, which can intuitively reflect the correctness of the landmark point pair matching and prepare for calculating the matching confidence of the landmark point subset.

[0066] Step 401: Calculate the coordinates of landmark points T i , T j in the earth's polar coordinate system where R is the radius of the earth.

[0067] Step 402: Calculate the relative position vector Eij of landmark points Ti and Tj in the earth's horizontal plane direction.

[0068]

[0069] Step 403: Calculate the relative position vector M ij between the corresponding matching points of landmark points Ti and Tj in the SAR radar image.

[0070] M ij = (x i - y j , y i - y j ) · RES

[0071] where RES is the pixel resolution of the airborne SAR image, unit: m / pixel.

[0072] Step 404: Calculate the position correlation error ΔL i , T j ) of the landmark point pair (T ij .

[0073] ΔLij = |E ij - M ij |

[0074] Step 5: Calculate the matching confidence C of the subset of landmark points (T1, T2,... T n ) according to the matching degree and the position correlation error 1-n .

[0075]

[0076] where Thd is the position correlation deviation threshold, which is determined by the specific SAR radar performance and imaging quality. Specifically

[0077] Thd = m·RES + Δh / sin(h / R min ) + Δθ·R min , where m is the matching pixel error, Δh is the relative height error, h is the relative height, R min is the minimum working distance, and Δθ is the track angle error. Obtaining the matching confidence of the subset of landmark points can be provided to the integrated navigation computer as an evaluation index for the availability of the navigation source. Compared with a single landmark point, the reliability of the subset of landmark points passing the test is significantly improved.

[0078] Specifically

[0079] Step 501: Calculate the probability that the matching result is correct under the condition that the relative position relationship is correct according to the matching degrees (P1, P2,... P n ) of the subset of landmark points (T1, T2,... T n ).

[0080] Step 502: Construct a probability function to calculate the probability that the relative position relationship between landmark point 1 and landmark point i (i = 2, 3,... n) is correct when the position correlation error is ΔL 1i .

[0081] Step 503: Calculate the matching confidence C among the n points in the set according to the conditional probability formula 1-n as follows

Claims

1. A method for evaluating the confidence of SAR scene matching results, characterized in that, Including: Determine the landmark point search range according to the SAR radar parameters and the aircraft flight parameters; Retrieve the set of landmark points to be matched from the landmark point database according to the landmark point search range; For each landmark point T in the set of landmark points to be matched i , use the normalized template matching method to calculate the pixel coordinates (x i , y i , i ) and the corresponding matching degree P i of the matching point of the landmark point T in the airborne SAR radar image; among them, the set of landmark points to be matched includes n landmark points, and the value of i ranges from 1 to n;​ For each pair of landmark points (T i , T j ) in the set of matched landmark points, calculate the position correlation error ΔL ij ; Calculate the matching confidence C of the landmark point pair (T i , T j ) according to the matching degree and the position correlation error, where the value of j ranges from 1 to n and j is not equal to i. ij ​ 2. The method according to claim 1, characterized in that, Determining the set to be matched specifically includes: determining the landmark point search range according to the SAR radar parameters and the aircraft flight parameters; retrieving the set of landmark points to be matched from the landmark point database according to the search range.

3. The method according to claim 1, characterized in that, For each pair of landmark points (T i , T j ) in the set, calculate the position correlation error ΔL ij , specifically including: Calculate the landmark points T i and T j coordinates (x ie , y ie , z ie ) in the Earth rectangular coordinate system, (x je , y je , z je ); Calculate the landmark point T i 、T j The first relative position vector E in the direction of the earth's horizontal plane ij ,E ij =(x ie -y je ,y ie -y je ) Calculate the landmark point T i , T j The second relative position vector Mij, M between the corresponding matching points of ij ij =(x i -y j , y i -y j )·RES, where the pixel coordinates (x i ) of the matching point of the landmark point T i , y i ) in the airborne SAR radar image, the pixel coordinates (x j ) of the matching point of the landmark point T j , y j ) in the airborne SAR radar image, and RES is the pixel resolution of the airborne SAR image, unit: m / pixel; According to the first relative position vector E ij and the second relative position vector M ij calculate the position correlation error ΔL i , ΔL j ) of the landmark point pair (T ij , ΔL ij = |E ij - M ij |.

4. The method according to claim 3, characterized in that, For each pair of landmark points (T i , T j ) in the set, the calculated matching confidence C ij is specifically as follows: Based on punctuation mark T i and T j matching degree P i and P j calculate the probability P that the matching result is correct under the condition that the relative position relationship between two landmark points is correct ij = 1 - (1 - P i )(1 - P j ); Construct a probability function to calculate the position correlation error as ΔL ij The probability that the relative position relationship between two landmark points is correct where Thd is the position correlation deviation threshold; Calculate the matching confidence C according to the conditional probability formula ij as follows:

5. The method according to claim 4, characterized in that, Calculate the position correlation error ΔL ij , which specifically includes: Calculate the landmark points T separately i , T j coordinates in the Earth's polar coordinate system where R is the radius of the Earth; Calculate the landmark point T i 、T j The first relative position vector E in the direction of the earth's horizontal plane ij , Calculate the landmark point T i and the second relative position vector M j between the corresponding matching points of T ij in the SAR radar image, where M ij =(x i -y j , y i -y j )·RES, where RES is the pixel resolution of the airborne SAR image, unit: m / pixel; According to the first relative position vector E ij and the second relative position vector M ij calculate the position correlation error ΔL i , ΔL j ) of the landmark point pair (T ij , ΔL ij = |E ij - M ij |.

6. The method according to claim 5, characterized in that, Calculate the matching confidence C of landmark point pairs (T i , T j ), where ij A binarization calculation method is adopted, specifically Thd is the position correlation deviation threshold.

7. The method according to claim 5, characterized in that, The position correlation deviation threshold Thd is determined according to the specific SAR radar performance and imaging quality, specifically Thd = m·RES + Δh / sin(h / R min ) + Δθ·R min , where m is the matching pixel error, Δh is the relative height error, h is the relative height, R min is the minimum operating distance, and Δθ is the track angle error.

8. The method according to claim 7, characterized in that, The method further includes: for multiple landmark point subsets (T1, T2,... T n ) in the set, calculating the matching confidence C 1-n between n points, specifically: Based on the matching degrees (P1, P2,... P n ) of the landmark point subsets (T1, T2,... T n ), calculate the probability that the matching result is correct under the condition that the relative position relationship is correct Construct a probability function to calculate the position correlation error between landmark point 1 and landmark point i (i = 2, 3, … n) as ΔL 1i The probability that the relative position relationship between the two landmark points is correct Calculate the matching confidence C according to the conditional probability formula 1-n as follows:

Images