CN105741295A - High-resolution remote sensing image registration method based on local invariant feature point - Google Patents

Abstract

The invention relates to a high-resolution remote sensing image registration method based on a local invariant feature point. The high-resolution remote sensing image registration method comprises the following steps: S1: extracting Harris feature points in a benchmark remote sensing image and a remote sensing image to be registered at different time phases in the same area to independently obtain feature point sets P1 and P2; S2: utilizing a SIFT (Scale Invariant Feature Transform) descriptor to independently carry out feature vector description on the feature point sets P1 and P2; S3: searching bidirectional matching point pairs; S4: randomly selecting three groups of matching point pairs PM3 from S3, and obtaining the root-mean-square error RM of the three groups of matching point pairs PM3; S5: judging the threshold value, and returning to S4 or entering S6; S6: calculating an affine transformation relationship matrix Matrix; and S7: utilizing transformation in S6 to obtain a registration image image_R. The high-resolution remote sensing image registration method solves the problem of big registration error of the high-resolution remote sensing image, can realize the high precision and the automation of registration and has a wide application value in the field of the change detection of the remote sensing image.

Description

A kind of high score Remote Sensing Image Matching method based on local invariant feature point

Technical field

The present invention relates to a kind of registration field, specifically a kind of high score Remote Sensing Image Matching method based on local invariant feature point.

Background technology

One good Remote Sensing Image Matching method or scheme should have characteristics that A, robustness are strong, and namely algorithm is not by the impact of the content of input picture and quality；B, wide adaptability, adaptability includes two aspects: be the type of input picture on the one hand, it is possible to be the image of different phase, it is also possible to be the image of different sensors, or even the registration between image and map；It is the type of geometric transformation model on the other hand, it is possible to be similarity transformation, affine transformation or polynomial transformation, even more complicated variation；C, automaticity are high, and algorithm need not operator participate in as far as possible, including the selection of ground control point, parameter and algorithm；D, amount of calculation are less, and this characteristic is particularly important when processing large scale remote sensing images；E, registration accuracy are high, have many applied environment requirement Sub-pixel precisions at present.But, up to the present can be provided simultaneously with above-mentioned five characteristics but without a kind of scheme or method, most method is both for application-specific.

Summary of the invention

The invention provides a kind of high score Remote Sensing Image Matching method based on local invariant feature point, utilize local invariant feature point to carry out extraction and the description of match point, by the calculating of affine transformation relationship matrix, it is possible to achieve the registration accuracy of sub-pixed mapping level.

Target by realizing the present invention be the technical scheme is that method comprises the following steps:

Step 1: extract the Harris characteristic point in the benchmark remote sensing image image1 and remote sensing image image2 subject to registration of areal difference phase, respectively obtain feature point set P₁And P₂；

Step 2: utilize SIFT to describe son respectively to feature point set P₁And P₂Carry out the description of characteristic vector；

Step 3: search characteristics point set P₁To P₂In the characteristic point of all Satisfying Matching Conditions C, obtain matching double points set M₁, meanwhile, search characteristics point set P₂To P₁In the characteristic point of all Satisfying Matching Conditions C, obtain matching double points set M₂, by M₁And M₂Common factor as bi-directional matching point to set Mset；

Step 4: the bi-directional matching point from step 3 randomly selects 3 groups of matching double points PM in set Mset₃, seek its root-mean-square error RE；

Step 5: judge that whether root-mean-square error RE is more than threshold value TH, if it is, return step 4, otherwise enters step 6；

Step 6: utilize 3 groups of matching double points PM in step 4₃Calculate affine transformation relationship matrix Matrix；

Step 7: utilize the affine transformation relationship matrix Matrix in step 6 that remote sensing image image2 subject to registration is converted, obtain registration image image_R.

Matching condition C in described step 3 is:

D_n/D_sn< r

Wherein, D_nFor the arest neighbors Euclidean distance between characteristic point, D_snFor the secondary neighbour's Euclidean distance between characteristic point, r is threshold value, and value is more big, and the coupling accuracy between characteristic point is more high.

Affine transformation relationship matrix Matrix in described step 6 is:

$Matrix=\left[\begin{array}{ccc}{t}_1& t_2& t_3\\ t_4& t_5& t_6\\ 0& 0& 1\end{array}\right]$

And Matrix needs to meet following relation:

$\left[\begin{array}{c}{x}_1\\ y_1\\ 1\end{array}\right]=Matrix\&CenterDot;\left[\begin{array}{c}{x}_2\\ y_2\\ 1\end{array}\right]$

Wherein, (x₁,y₁) and (x₂,y₂) respectively 3 groups of matching double points PM in step 4₃In characteristic point coordinate.

The invention has the beneficial effects as follows: solve the problem that high spatial resolution remote sense image registration error is big, it is possible to realize high accuracy and the automatization of registration, be with a wide range of applications in the change-detection field of remote sensing image.

Accompanying drawing explanation

Fig. 1 is the overall process flow figure of the present invention.

Detailed description of the invention

The specific embodiment of the present invention is described in detail below in conjunction with accompanying drawing.

In step 101, the benchmark remote sensing image image1 and remote sensing image image2 subject to registration of input areal difference phase.

In step 102, extract the Harris characteristic point in image1 and image2, respectively obtain feature point set P₁And P₂。

In step 103, SIFT is utilized to describe son respectively to feature point set P₁And P₂Carry out the description of characteristic vector.

In step 104, search characteristics point set P₁To P₂In the characteristic point of all Satisfying Matching Conditions C, obtain matching double points set M₁, meanwhile, search characteristics point set P₂To P₁In the characteristic point of all Satisfying Matching Conditions C, obtain matching double points set M₂, by M₁And M₂Common factor as bi-directional matching point to set Mset.

In step 105, the bi-directional matching point from step 104 randomly selects 3 groups of matching double points PM in set Mset₃。

3 groups of matching double points PM in step 106, calculation procedure 105₃Root-mean-square error RE.

In step 107, it is judged that whether the root-mean-square error RE in step 106 is more than threshold value TH, if it is, return step 105, otherwise enter step 108.

In step 108, utilize 3 groups of matching double points PM in step 106₃Calculate affine transformation relationship matrix Matrix.

In step 109, utilize the affine transformation relationship matrix Matrix in step 108 that remote sensing image image2 subject to registration is converted, obtain registration image image_R.

In step 110, export registration result.

Claims (3)

1. the high score Remote Sensing Image Matching method based on local invariant feature point, it is characterised in that comprise the following steps:

Step 1: extract the Harris characteristic point in the benchmark remote sensing image image1 and remote sensing image image2 subject to registration of areal difference phase, respectively obtain feature point set P₁And P₂；

Step 2: utilize SIFT to describe son respectively to feature point set P₁And P₂Carry out the description of characteristic vector；

Step 3: search characteristics point set P₁To P₂In the characteristic point of all Satisfying Matching Conditions C, obtain matching double points set M₁, meanwhile, search characteristics point set P₂To P₁In the characteristic point of all Satisfying Matching Conditions C, obtain matching double points set M₂, by M₁And M₂Common factor as bi-directional matching point to set Mset；

Step 4: the bi-directional matching point from step 3 randomly selects 3 groups of matching double points PM in set Mset₃, seek its root-mean-square error RE；

Step 5: judge that whether root-mean-square error RE is more than threshold value TH, if it is, return step 4, otherwise enters step 6；

Step 6: utilize 3 groups of matching double points PM in step 4₃Calculate affine transformation relationship matrix Matrix；

Step 7: utilize the affine transformation relationship matrix Matrix in step 6 that remote sensing image image2 subject to registration is converted, obtain registration image image_R.

2. a kind of high score Remote Sensing Image Matching method based on local invariant feature point according to claim 1, it is characterised in that the matching condition C in step 3 is:

D_n/D_sn< r

Wherein, D_nFor the arest neighbors Euclidean distance between characteristic point, D_snFor the secondary neighbour's Euclidean distance between characteristic point, r is threshold value.

3. a kind of high score Remote Sensing Image Matching method based on local invariant feature point according to claim 1, it is characterised in that the affine transformation relationship matrix Matrix in step 6 is:

$Matrix=\left[\begin{array}{ccc}{t}_1& t_2& t_3\\ t_4& t_5& t_6\\ 0& 0& 1\end{array}\right]$

And Matrix needs to meet following relation:

$\left[\begin{array}{c}{x}_1\\ y_1\\ 1\end{array}\right]=Matrix\&CenterDot;\left[\begin{array}{c}{x}_2\\ y_2\\ 1\end{array}\right]$

Wherein, (x₁,y₁) and (x₂,y₂) respectively 3 groups of matching double points PM in step 4₃In characteristic point coordinate.