CN102968780B - A kind of remote sensing images joining method based on human-eye visual characteristic - Google Patents

Abstract

The invention discloses a remote sensing image splicing method based on human visual characteristics, which comprises the following steps: 1) extracting feature points on a reference image and an image to be spliced, and establishing matching feature point pairs; 2) eliminating wrong matching feature point pairs , to obtain the initial transformation matrix H between the reference image and the image to be stitched; 3) Divide the reference image into an area of interest and a common area, and use the ratio of the average gradient values of the two areas to obtain the weight factor α; 4) Set The incremental matrix P is used as the optimized increment of the initial transformation matrix H, and the incremental matrix P is iteratively solved by using the coordinate information of the matching feature point pairs in step 2) and the weight factor α in step 3), and using each The incremental matrix P obtained at one time can be calculated to obtain the corresponding transformation matrix H, until the iterative process ends after convergence; 5) Apply the final transformation matrix H to perform projective transformation on the image to be stitched, and stitch and fuse it with the reference image to complete the stitching of the image.

Description

A remote sensing image mosaic method based on human visual characteristics

technical field

The invention relates to the technical field of computer image processing, in particular to a remote sensing image splicing method based on human visual characteristics.

Background technique

Remote sensing images are images obtained through high-altitude flying platforms such as satellites, airplanes, or hot air balloons. They play a very important role in disaster warning, resource detection, military reconnaissance, map surveying, and other fields. important way.

In the process of acquiring remote sensing images, it is always hoped to obtain high-resolution images with sufficient field of view, which can facilitate people to quickly extract useful information. However, under the existing technical conditions, the field of view and resolution are two contradictions that are difficult to reconcile.

The existing solution is to continue to increase the resolution of the sensor on the one hand, and on the other hand to obtain multiple overlapping images with high resolution by imaging a large area multiple times, and restore them into a large-coverage high-resolution image through subsequent image processing technology. rate remote sensing images.

Remote sensing image stitching can be divided into two parts: image registration and image fusion.

Image registration is the core link of image mosaic technology, and it is a means to obtain the transformation relationship of two overlapping images.

In order to obtain a more accurate relationship between images, it is usually optimized in image registration, and there are many optimization methods.

However, the previous optimization algorithms were uniform optimization, that is, they did not distinguish the importance of image content. The actual result is that it is possible to achieve relatively high stitching accuracy in some areas that are not useful for actual production and life, but in On the contrary, some very important content got very poor results, which is not conducive to subsequent applications.

Contents of the invention

The purpose of the present invention is to provide an optimization method for different application environments for differences in the importance of remote sensing image scenes, so that the optimized results conform to the visual judgment characteristics of human eyes.

In order to achieve the above object, the present invention provides a remote sensing image splicing method based on human visual characteristics, comprising the following steps:

1) Extract the feature points on the reference image and the image to be stitched, set up matching feature point pairs, and obtain the initial matching relationship between the feature points of the two images;

2) Eliminate the wrong matching relationship, obtain the matching feature point pair with the correct matching relationship, and obtain the initial transformation matrix H between the reference image and the image to be stitched;

3) divide the reference image into an area of interest and a common area, calculate the average gradient values S _interest (G) and S _normal (G) of the two areas respectively, and use the ratio of the average gradient values to obtain the weight factor α;

4) Set the increment matrix P as the optimization increment of the initial transformation matrix H, use the coordinate information of the matching feature point pairs in step 2) and the weight factor α in step 3) to iteratively solve the increment matrix P, and use the iterative The incremental matrix P obtained each time in the process can be calculated to obtain the corresponding transformation matrix H, until the convergence ends the iterative process;

5) Apply the final transformation matrix H to carry out projective transformation on the image to be spliced, and splice and fuse it with the reference image to complete the splicing of the image.

The feature points of the reference image and the image to be stitched are selected using the SIFT algorithm, and the initial matching relationship between the feature points is obtained by using the Euclidean geometric distance principle; and the RANSAC method is used to eliminate the wrong matching relationship, that is, the matching feature points with the correct matching relationship are obtained Yes, the transformation matrix obtained through the correct matching relationship can make the two spliced images overlap better, while the wrong matching will result in false splicing.

Let the homogeneous coordinates of the feature points of the reference image in step 2) be X ₁ =(x ₁ , y ₁ , 1), and the homogeneous coordinates of the feature points of the image to be stitched be X ₂ =(x ₂ , y ₂ , 1 ), and the initial transformation matrix H is:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

In the formula, h ₁ ~ h ₈ are the elements in the transformation matrix H, which represent the projective transformation relationship between the reference image and the image to be stitched, and the coordinate transformation relationship between the reference image and the image to be stitched corresponds to the following formula:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

The gradient value G(x, y) of a pixel in an image is represented by the following formula:

$\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}$

${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; ,</span>$ ${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}$

In the formula, (x, y) is the coordinate of the pixel in the image, I _x represents the gradient of the image in the x direction, I _y represents the gradient of the image in the y direction, and I(x, y) represents the gray value of the image.

The weight factor α is expressed as:

$\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; interest</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; normal</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; )</span>}$

$\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}$

In the formula: S _interest (G) is the average value of the gradient values of all pixels in the region of interest, and S _normal (G) is the average value of the gradient values of all pixels in the normal region.

The incremental matrix P is expressed as:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

In the formula, the elements p ₁ to p ₈ in the increment matrix P correspond to the tiny increments of the corresponding elements in the transformation matrix H, respectively.

Setting the intermediate coordinate variable X′ ₂ =(x′ ₂ , y′ ₂ , 1) is convenient for mathematical expression calculation:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; \&prime;</span>}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; \&prime;</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

The increment matrix P represents the small increment of the projection transformation matrix H in each iteration process, and its function is to continuously optimize and approximate H, and its expression is:

(I+P)H→H

In the formula, I is the unit matrix, using the intermediate coordinate variable X′ ₂ =(x′ ₂ , y′ ₂ , 1) and the formula:

(I+P)HX ₂ =(I+P)X′ ₂ =X3=(x ₃ , y ₃ , 1)

The calculation can be simplified, and X ₃ =(x ₃ , y ₃ , 1) represents the coordinates of the feature points of the image to be stitched after projective transformation.

Introduce the weight factor α to divide the image into two parts:

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>$

In the formula, Respectively represent the coordinates of the nth feature point of the reference image and the coordinates of the nth feature point coordinates of the image to be stitched after projective transformation, Represent the coordinates of the mth feature point of the reference image and the coordinates of the mth feature point of the image to be stitched after projective transformation, α is the weight factor, S ₁ and S ₂ represent the area of interest and the common area respectively.

The incremental matrix P obtained in each iteration process, according to (I+P)H→H, the transformation matrix H in the next iteration process is obtained, and the termination condition of the iteration is convergence or exceeding the limit of the number of iterations, and the final The transformation matrix H of .

Substituting the final transformation matrix H into the coordinate transformation relationship between the reference image and the image to be spliced:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

Complete the projective transformation of the image to be stitched, and obtain the stitched image.

The present invention has the following advantages:

The present invention introduces the concept of weight optimization in the process of remote sensing image registration optimization. According to the importance of remote sensing image content, different weight values are given, so that the results of remote sensing image registration optimization are more in line with human visual judgment. In important remote sensing areas, the error of image registration can be as small as possible, and the registration accuracy can reach a relatively high level, so that the final stitched image can be applied to a wider range of actual needs.

Description of drawings

Fig. 1 is the operation flowchart of the remote sensing image splicing method based on human visual characteristics in the present invention;

Fig. 2a, Fig. 2b, Fig. 2c and Fig. 2d are the process of establishing image feature point matching relationship in the present invention;

Figure 2a is a reference image;

Figure 2b is the image to be spliced;

Figure 2c is the initial matching relationship of the feature points of the two images;

Figure 2d shows the correct matching relationship obtained after the RANSAC method;

Figure 3a is the mosaic result optimized by the common Levenberg-Marquardt algorithm;

Figure 3b and Figure 3c are the corresponding marked areas in Figure 3a;

Figure 4a is the mosaic result after the mosaic method optimization of the present invention;

Figure 4b and Figure 4c are the corresponding marked areas in Figure 4a;

Figure 5a is a reference image for simulation;

Figure 5b is a simulated image to be stitched;

Figure 6a is the splicing result optimized by the common LM algorithm;

Figure 6b, Figure 6c and Figure 6d are the corresponding marked areas in Figure 6a;

Figure 7a is the mosaic result after the mosaic method optimization of the present invention;

Figure 7b, Figure 7c and Figure 7d are the corresponding marked areas in Figure 7a;

Figure 8a is a reference image of a remote sensing image;

Fig. 8b is the image to be spliced of the remote sensing image;

Fig. 8c is the mosaic result after the splicing method of the present invention is optimized.

Detailed ways

As shown in FIG. 1 , the present invention is a remote sensing image mosaic method based on human visual characteristics, and the specific implementation manners are illustrated by the following two specific examples.

Example 1: Figure 2a, Figure 2b, Figure 2c and Figure 2d are used to illustrate the process of the entire splicing method.

(1) Figure 2a and Figure 2b are the reference image and the image to be stitched respectively. The feature points of the reference image and the image to be stitched are extracted using SIFT technology, and Figure 2c shows the initial matching relationship of the feature points of the two images obtained.

(2) Use the projection transformation model to represent the transformation relationship between the two images, that is, the transformation matrix contains 8 parameter variables, and then use the RANSAC technology to eliminate the wrong matching points to obtain the feature point pairs with the correct matching relationship, and at the same time get the initial Transformation matrix H. Let the homogeneous coordinates of the feature points of the reference image be X ₁ = (x ₁ , y ₁ , 1), the homogeneous coordinates of the feature points of the image to be spliced are X ₂ = (x ₂ , y ₂ , 1), and the matrix H is 3x3 The projection transformation matrix of , where the transformation matrix H is:

$\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

In the formula, h ₁ ~ h ₈ are the elements in the projection transformation matrix H, which represent the projection transformation relationship between the reference image and the image to be spliced, and the coordinate transformation relationship corresponds to the following formula:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

(3) Adopt a weight optimization algorithm that considers the importance of content. The quantification of the importance of remote sensing image content can be set into two methods: subjective method and objective method. The subjective algorithm is the artificially added weight value of various contents of the remote sensing image by the user. For example, airport runways and military targets can be assigned relatively high weights, and the range of weight values is the interval [0, 1]. Aiming at the requirements of automatic optimization algorithm, the gradient information of the image is designed to represent the importance of information in a certain area. The gradient of the image pixel is determined by the following formula:

$\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}$

${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; ,</span>$ ${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&PartialD;</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}$

In the formula, (x, y) is the coordinate of the pixel in the image, I _x represents the gradient of the image in the x direction, I _y represents the gradient of the image in the y direction, and I(x, y) represents the gray value of the image.

The weight factor α is expressed as:

$\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; interest</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; normal</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; )</span>}$

$\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&beta;</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}$

In the formula: S _interest (G) is the average value of the gradient values of all pixels in the region of interest, and S _normal (G) is the average value of the gradient values of all pixels in the normal region.

(4) Set the incremental matrix P as the optimized increment of the initial transformation matrix H, and use the weight factor α and the coordinate positions of the matching feature point pairs to iteratively solve the incremental matrix P. In the iterative process, the incremental matrix P obtained each time and the initial transformation matrix H constitute the initial transformation matrix H of the next iteration, and the iterative process ends until convergence.

The incremental matrix P is expressed as:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

Setting the intermediate coordinate variable X′ ₂ =(x′ ₂ , y′ ₂ , 1) is convenient for mathematical expression calculation:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; \&prime;</span>}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}^{<span\; class="notranslate">\; \&prime;</span>}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

The increment matrix P represents the small increment of the projection transformation matrix H in each iteration process, and its function is to continuously optimize and approximate H, and its expression is:

(I+P)H→H

I is the unit matrix, and the calculation can be simplified by using the intermediate coordinate variable X′ ₂ and (I+P)HX ₂ =(I+P)X′ ₂ =X ₃ =(x ₃ , y ₃ , 1).

In order to get the incremental matrix P, we need to solve the following formula:

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>$

In the formula, represent the coordinates of the nth feature point of the reference image and the coordinates of the nth feature point coordinates of the image to be stitched after projective transformation, respectively.

Introduce the weight factor α to divide the image into two parts:

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>$

In the formula, Respectively represent the coordinates of the nth feature point of the reference image and the coordinates of the nth feature point coordinates of the image to be stitched after projective transformation, Represent the coordinates of the mth feature point of the reference image and the coordinates of the mth feature point of the image to be stitched after projective transformation, α is the weight factor, S ₁ and S ₂ represent the area of interest and the common area respectively.

The incremental matrix P is solved by using the Levenberg-Marquardt algorithm. The incremental matrix P obtained in each iteration process, the transformation matrix H in the next iteration process is obtained according to (I+P)H→H, and the termination condition of the iteration is convergence or exceeding the limit of the number of iterations. The criterion for convergence is that ε is less than 1 pixel:

$\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}\sqrt{{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; no</span>}$

In the formula, Indicates the coordinates of the feature points of the reference image in the region of interest, Indicates the transformed coordinates of the feature points of the image to be stitched. Preferably, the maximum number of iterations is set to 100. In some cases, the situation of the two acquired images is more complicated, and the average error of the region of interest cannot be reduced to within one pixel, so the number of iterations is used as the termination condition of the iteration.

Using the incremental matrix P and the transformation matrix H, (I+P)H→H can be iteratively solved to obtain the final transformation matrix H, and then according to the projection transformation relation between the reference image and the image to be spliced:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$

The image to be stitched is subjected to projection transformation, and the final optimized stitched image is obtained.

Comparing the mosaic result of the optimization method of the present invention with the mosaic result after optimization of the common LM algorithm, it can be obtained: Fig. 3a is the mosaic result after optimization of the common Levenberg-Marquardt algorithm, and Fig. 4a is the mosaic result of the optimization method of the present invention. Figure 3b and Figure 3c are the enlarged images of the corresponding areas in Figure 3a, respectively, where Figure 3b is from the normal area, and Figure 3c is from the area of interest, we can find that the optimization results of Figure 3b are relatively good, but the optimization of Figure 3c The result is relatively poor. But in Figure 4b and Figure 4c, we can find that the image optimization effect of the region of interest in Figure 4c is better, but the optimization effect of Figure 4b is relatively poor.

Example 2: The characteristics of the method of the present invention in terms of human vision are illustrated by comparing the optimized splicing results of the simulated image and the remote sensing image.

(1) Figure 5a and Figure 5b are the input reference image and the image to be stitched.

(2) Figure 6a shows the optimized stitching structure without weight values. Figure 6b and Figure 6c come from the region of interest we defined, and the result of optimized splicing has a relatively large error. Figure 6d is from the common region we defined.

(3) Figure 7a is the optimized splicing result of the present invention. Compared with Fig. 6b and Fig. 6c, the optimization error of Fig. 7b and Fig. 7c is very small, basically eliminating the difference on the visual level, while Fig. 7d has basically no big difference visually compared with Fig. 6d.

(4) Figure 8a and Figure 8b are reference images and images to be stitched of remote sensing images, and Figure 8c is the stitching result using the present invention, wherein the optimization error of feature points in the region of interest has reached the sub-pixel level.

Claims (8)

1. a remote sensing image mosaic method based on human visual characteristics, is characterized in that, comprises the following steps: 1) Extract the feature points on the reference image and the image to be stitched, set up matching feature point pairs, and obtain the initial matching relationship between the feature points of the two images; 2) Eliminate the wrong matching relationship, obtain the matching feature point pair with the correct matching relationship, and obtain the initial transformation matrix H between the reference image and the image to be stitched; 3) divide the reference image into an area of interest and a common area, calculate the average gradient values S _interest (G) and S _normal (G) of the two areas respectively, and use the ratio of the average gradient values to obtain the weight factor α; The weight factor α is expressed as: $\mathrm{\&alpha;}=\frac{1}{\mathrm{\&beta;}+1},\mathrm{\&beta;}=\frac{S_{i\mathrm{no}tere\mathrm{the\; s}t}\left(G\right)}{S_{\mathrm{no}ormal}\left(G\right)};$ 4) Set the increment matrix P as the optimization increment of the initial transformation matrix H, use the coordinate information of the matching feature point pairs in step 2) and the weight factor α in step 3) to iteratively solve the increment matrix P, and use the iterative The incremental matrix P obtained each time in the process can be calculated to obtain the corresponding transformation matrix H, until the convergence ends the iterative process; The criterion for convergence is that ε is less than 1 pixel: $\mathrm{<span\; class="notranslate">\; \&epsiv;</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\sqrt{{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; no</span>}$ In the formula, Indicates the coordinates of the feature points of the reference image in the region of interest, Indicates the transformed coordinates of the feature points of the image to be spliced; Introducing the formula: $\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; \&Element;</span>{\mathrm{<span\; class="notranslate">\; S</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}}{<span\; class="notranslate">\; \&Sigma;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&alpha;</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}^{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; |</span><span\; class="notranslate">\; |</span>$ In the formula, Respectively represent the coordinates of the nth feature point of the reference image and the coordinates of the nth feature point coordinates of the image to be stitched after projective transformation, Represent the coordinates of the mth feature point of the reference image and the coordinates of the mth feature point coordinates of the image to be stitched after projective transformation, α is the weight factor, S ₁ , S ₂ represent the area of interest and the common area respectively; and Use the Levenberg-Marquardt algorithm to iteratively solve the incremental matrix P; 5) Apply the final transformation matrix H to carry out projective transformation on the image to be stitched, and stitch and fuse it with the reference image to complete the stitching of the image. 2. The remote sensing image mosaic method based on human visual characteristics as claimed in claim 1, wherein the homogeneous coordinates of the reference image feature points in said step 2) are X ₁ =(x ₁ , y ₁ , 1), the homogeneous coordinates of the image feature points to be spliced are X ₂ =(x ₂ ,y ₂ ,1), and the initial transformation matrix H is: $\mathrm{<span\; class="notranslate">\; h</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$ In the formula, h ₁ ~ h ₈ are the elements in the transformation matrix H, which represent the projective transformation relationship between the reference image and the image to be stitched. 3. the remote sensing image mosaic method based on human visual characteristics as claimed in claim 2, is characterized in that, the gradient value G (x, y) of pixel in the image is represented by following formula: $\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\sqrt{{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}$ ${\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; I</span>}}_{\mathrm{<span\; class="notranslate">\; the\; y</span>}}<span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>}{<span\; class="notranslate">\; \&part;</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}}$ In the formula, (x, y) is the coordinate of the pixel in the image, I _x represents the gradient of the image in the x direction, I _y represents the gradient of the image in the y direction, and I(x, y) represents the gray value of the image. 4. the remote sensing image splicing method based on human visual characteristics as claimed in claim 3, is characterized in that, described incremental matrix P is expressed as: $\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; p</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$ In the formula, the elements p ₁ to p ₈ in the increment matrix P correspond to the tiny increments of the corresponding elements in the transformation matrix H, respectively. 5. the remote sensing image splicing method based on human visual characteristics as claimed in claim 4, is characterized in that, the relational expression of increment matrix P and transformation matrix H is: (I+P)H→H In the formula, I is the identity matrix. 6. the remote sensing image splicing method based on human visual characteristics as claimed in claim 5, is characterized in that, the coordinates of the feature point coordinates of the image to be spliced after projection transformation are: X ₃ =(x _3, y ₃ ,1)=(I+P)HX ₂ =(I+P)X' ₂ In the formula, I is the unit matrix, X' ₂ is the intermediate coordinate variable, and X' ₂ =(x' ₂ ,y' ₂ ,1). 7. the remote sensing image splicing method based on human visual characteristics as claimed in claim 6, is characterized in that, utilizes the increment matrix P that obtains in each iterative process, obtains next time according to (I+P)H→H The transformation matrix H in the iterative process is substituted into the formula (I+P)H→H to iteratively solve to obtain the final transformation matrix H. 8. the remote sensing image splicing method based on human visual characteristics as claimed in claim 7, is characterized in that, the coordinate transformation relational expression between described reference image and image to be spliced is: $\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]<span\; class="notranslate">\; ~</span>\left[\begin{array}{ccc}{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 3</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 4</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 5</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 6</span>}}\\ {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 7</span>}}& {\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; 8</span>}}& \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ {\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}\\ \mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right]$ Substituting the final transformation matrix H into the above formula, the image to be stitched is subjected to projective transformation, and the stitched image is obtained.