Earth infrared image reference adaptation area selection method

Abstract

The invention provides a method for selecting a reference adaptation area of an earth infrared image. The method comprises the following steps: establishing a global infrared radiation reference map and partitioning the global infrared radiation reference map; detecting a key angular point of each block region; a K-means algorithm is adopted to classify the key angular points, the classification number is determined by an improved weighted Voronoi graph, sub-regions are determined according to the classification result, it is ensured that the influences of all the sub-regions are similar, and therefore uniform distribution of features is achieved; constructing an infrared image suitability characteristic evaluation standard, counting infrared image suitability characteristic indexes of the sub-regions, and evaluating the infrared image suitability characteristic indexes; and based on the evaluation result of each sub-region, screening out a standard matching region meeting the requirement. According to the method, the infrared image reference matching area with remarkable features, rich information amount and high adaptability can be automatically generated. The technology provides solid data support for infrared image navigation, and has wide application prospects in the fields of image matching and satellite navigation.

Description

Technical Field

[0001] The invention relates to the field of satellite navigation and image processing, and in particular to a method for selecting a reference adaptation area of ​​an earth infrared image. Background Art

[0002] With the rapid development of computer vision and imaging equipment, image navigation technology plays a vital role in the fields of drones, missiles, intelligent driving, deep space exploration, and space missions. As one of the satellite navigation methods, satellite image navigation technology matches the real-time images with the pre-prepared reference map with geographic information, and combines the camera imaging model to determine the current position and attitude information of the satellite. High-quality adaptation area reference maps can provide an accurate and reliable information basis for satellite image navigation technology, which is a prerequisite for realizing image navigation.

[0003] Domestic and foreign scholars have conducted extensive research and exploration on the selection methods of matching benchmark images, which can be mainly summarized into two categories: methods based on multi-attribute decision-making and adaptive matching methods based on pattern classification theory. The method based on multi-attribute decision-making belongs to the traditional screening algorithm, which establishes a comprehensive evaluation index by combining multiple image features (such as grayscale variance, edge density, gradient features, etc.). For example, Liu Zhonghua et al. proposed a multi-scale matching template selection scheme from coarse to fine, and gave corresponding quantitative indicators; Zhang Yangyang et al. combined the hierarchical analysis method with the entropy technology method to assign weights to multiple attributes, thereby generating a comprehensive evaluation index to select matching areas; Tu Guoyong et al. used a binary tree structure model to combine multiple image attributes to achieve matching; Zhang Xiaochen et al. proposed a selection scheme based on information theory that combines image information, feature stability and regional uniqueness, which can improve the probability of self-matching.

[0004] The method based on pattern classification theory was developed after the introduction of classifiers and artificial intelligence technology in recent years. This method is usually based on multiple feature indicators and selects matching areas by setting thresholds or using autonomous learning classification algorithms. For example, Tu Guoyong et al. proposed a scene matching method based on the least squares support vector regression machine; Li Bangjie et al. established a matching area adaptability screening model based on convolutional neural networks.

[0005] Despite a lot of research and exploration, the following problems still exist:

[0006] 1. Most algorithms are for visible light images, which are easily affected by factors such as seasons and day-night changes. Infrared images have the advantage of working in all weather and all regions, but currently there are relatively few algorithms for extracting the adaptation area for infrared images.

[0007] 2. The selection strategy based on the multi-attribute decision-making method relies heavily on manual empirical judgment and the characteristics of the reference image, and has weak anti-interference ability; while the method based on pattern classification theory relies on pre-classified data for learning, which makes existing work and applications require a lot of manual intervention and are easily affected by subjective factors, resulting in low work efficiency.

[0008] 3. Most current algorithms evaluate the adaptability of images through comprehensive indicators, but few studies focus on the comprehensive algorithmic process of extracting one or more adaptation areas from an image. Summary of the invention

[0009] In order to solve the problems in the prior art, the present invention proposes a method for selecting a reference adaptation zone of an earth infrared image. Starting from the content of the reference image, the present invention focuses on the correlation and inherent connection between the features of the adaptation zone, reduces the interference of human subjective factors, and realizes the autonomous extraction of the reference adaptation zone. In order to further extract the adaptation zone with significant features, rich information and excellent adaptability, the present invention combines a multi-scale comprehensive screening algorithm and proposes corresponding quantitative evaluation indicators.

[0010] The present invention provides a method for selecting a reference adaptation area of ​​an earth infrared image, comprising the following steps:

[0011] Step 1: Obtain an infrared radiation image of the target area as an infrared radiation reference map, divide the infrared radiation reference map into several block areas, and perform steps 2 to 6 for each block area;

[0012] Step 2: Detect the key corner points of the current block area and cluster the key corner points to obtain k cluster center points;

[0013] Step 3: Use the weighted Voronoi diagram method to calculate the influence area of ​​each cluster center point in the current block area, and adjust the number of cluster classifications k according to the area ratio of each influential area;

[0014] Step 4: Repeat steps 2-3 until the area ratios of the influential regions in the current block area are all within the range of 0.8 to 1.2; solve the rectangular envelope of each influential region finally obtained, and the envelope range is the corresponding sub-region;

[0015] Step 5: Count the infrared image adaptability characteristic indicators of each sub-region; the infrared image adaptability characteristic indicators include grayscale standard deviation Std, information entropy H and Frieden grayscale entropy F;

[0016] Step 6: Establish an infrared image adaptability feature evaluation standard. The sub-region that meets the evaluation standard is the benchmark adaptation area of ​​the current block area. The infrared image adaptability feature evaluation standard is: grayscale standard deviation Std≥15, information entropy H≥5, and Frieden grayscale entropy F≤3.

[0017] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0018] (1) The present invention proposes a method for selecting a reference adaptation area for an earth infrared image. By simplifying the reference image preparation problem into a classification problem, the image feature analysis, clustering, and weighted Voronoi diagram methods are used to automatically determine the number of classifications and classify similar key corner points to ensure the correlation and inherent connection of feature points in the matching adaptation area, making the feature distribution more scientific and uniform. This method significantly reduces the dependence on manual labor, improves work efficiency, and weakens the influence of subjective factors.

[0019] (2) The weighted Voronoi diagram and multiple image feature indices are used to quantitatively evaluate and analyze the selected sub-regions. The selected matching reference image has the advantages of large amount of information, obvious features, and strong anti-interference ability, and successfully realizes the preparation of high-quality earth infrared image navigation reference map. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flow chart for the implementation of the present invention;

[0021] Figure 2 are the key corner points detected by the Harris algorithm in the present invention;

[0022] Figure 3 It is the result of K-means and weighted Voronoi diagram classification in the present invention;

[0023] Figure 4 It is the sub-region to be selected extracted by the present invention. DETAILED DESCRIPTION

[0024] The present invention is further described and illustrated below in conjunction with specific embodiments. The embodiments are merely exemplary of the present disclosure and do not define the scope of limitation. The technical features of each embodiment of the present invention may be combined accordingly without conflicting with each other.

[0025] like Figure 1 As shown, the specific implementation steps of the present invention include the following:

[0026] Step 1: Based on Planck's blackbody radiation law, construct the earth's background infrared radiation model and generate a global infrared radiation benchmark map.

[0027] In this embodiment, based on Planck's blackbody radiation law, combined with MODIS remote sensing images, global land cover data (GLCNMO) and MODTRAN software, an 8-14μm band mid-infrared radiation model of the earth is constructed. Planck's blackbody radiation law describes the relationship between the spectral radiation intensity B(λ,T) and the wavelength λ at any temperature T, and its calculation formula is:

[0028]

[0029] The Earth's background infrared radiation model in the 8-14μm band is expressed as follows:

[0030]

[0031] Among them I 8-14 is the radiation intensity in the 8-14μm band, ∈ is the surface emissivity, τ(λ) is the atmospheric transmittance, R rfl (λ) is the atmospheric radiation and solar radiation reflected downward, R at (λ)↑ is the atmospheric radiation reflected upward.

[0032] Based on the earth's background infrared radiation model, the infrared radiation intensity of each location point on the earth is obtained, and an infrared radiation image of the earth is constructed according to the infrared radiation intensity of each location point. The infrared radiation image is the global infrared radiation reference map.

[0033] Step 2: Perform a preliminary block division on the global infrared radiation benchmark map, dividing it into several block areas with larger spatial scales.

[0034] In this embodiment, the global infrared radiation reference map obtained in step 1 is divided along the longitude and latitude directions with a step length of 10°. The spatial coverage of each block area is 20°×20°. The block areas that only include the ocean are eliminated. Figure 2 An image of one of the regions is shown.

[0035] Step 3: Use feature point detection algorithm to identify key corner points in each block area, and use K-means clustering method to classify these key corner points.

[0036] In this embodiment, the Harris corner point extraction algorithm is used to detect key corner points in each block area. The basic principle of this algorithm is to find pixels with large grayscale changes in a fixed window and mark these pixels as key corner points; see Figure 2 As shown, the detected key corners are marked in green.

[0037] The basic principle of the K-means clustering algorithm is to select k points in the data set as cluster centers and classify the data closest to these centers according to the Euclidean distance. By continuously iteratively updating the positions of the cluster centers, the similarity of sample data of the same type is maximized. The objective function of the K-means clustering algorithm is:

[0038]

[0039] Where P = {p1,…,p n} represents the n key corner points extracted by the Harris corner detection algorithm, C = {c1,…,c k} are k cluster center points, and the initial value of each cluster center point is the center point of the current block area, z ij represents the key corner point p i With cluster center c j The ownership relationship is specifically expressed as:

[0040]

[0041] Cluster center c j The update formula is:

[0042]

[0043] By continuously iteratively updating the cluster center point c j , so that the objective function F(z,C) is minimized, thus completing the clustering process. Figure 3 As shown, take the number of categories k = 4, the K-means algorithm will Figure 2 The key corner points in the image are divided into four categories. Different categories are represented by different colors in the image, and the cluster center point of each category is marked with a black "×".

[0044] Step 4: Use the weighted Voronoi diagram to calculate the influence of each category, and autonomously adjust the number of categories of the K-means clustering algorithm to make the feature distribution of each partition more scientific and balanced.

[0045] In this embodiment, the cluster center points of each category are used as sites, and the weighted Voronoi diagram is used to calculate the influence area V (c j ), the specific formula is as follows:

[0046]

[0047] Among them, ∩ j≠i {.} means finding the intersection of all sets satisfying j≠i, S = {s1,…,s q} represents the set of all pixels in the image, is pixel s p and cluster center c j The weighted Euclidean distance between p ,c j ) is pixel s p To cluster center c j The Euclidean distance, w j is the weight of the jth classification, and the weight of each classification is proportional to the number of key corner points contained in each classification. The purpose of introducing weights is to balance the uneven distribution of feature points caused by terrain and location differences.

[0048] In order to obtain a set of adaptation areas with uniform distribution of feature points, the number of clusters can be determined by comparing the area of ​​influence of each category in the block area. When the influence ratio of each category in the block area is close to 1, it means that the influence of each category in the block area is similar and the features are evenly distributed. The mathematical expression of the influence ratio is:

[0049]

[0050] When maxr ij >>1, it means that the number of categories needs to be increased, so the number of categories is increased by 1, K-means clustering is performed again, and the influence of each category is calculated; when maxr ij When <<1, it means that the number of clusters needs to be reduced. Then reduce the number of categories by 1, re-perform K-means clustering and calculate the influence of each category. Through multiple calculations, r ij ≈1. Figure 3 As shown in the figure, after multiple calculations, the key corner points of the block area are divided into four categories. The influence areas controlled by the cluster centers of the four categories are divided by black lines in the figure. The influence ratio between each category ranges from 0.8 to 1.2. To facilitate subsequent calculations, the rectangular envelope of each influence area is solved to obtain four sub-areas to be selected. The four sub-areas are as follows: Figure 4 shown.

[0051] Step 5: Establish an infrared image adaptability feature evaluation standard, perform statistics on the infrared image adaptability feature indicators of each sub-region, and obtain the evaluation results.

[0052] The image adaptability characteristic indicators in this embodiment include grayscale standard deviation Std, information entropy H and Frieden grayscale entropy F.

[0053] The evaluation criteria for the adaptability characteristics of infrared images are grayscale standard deviation Std≥15, information entropy H≥5, and Frieden grayscale entropy F≤3.

[0054] The grayscale standard deviation Std reflects the degree of grayscale change of the sub-region as a whole. The larger the grayscale standard deviation, the more drastic the grayscale change and the greater the amount of information contained in the image. The calculation formula is:

[0055]

[0056] Where a and b are the number of rows and columns of pixels in the sub-region, and f(i,j) represents the pixel value located in the i-th row and j-th column in the sub-region. is the average gray value of the sub-region.

[0057] The information entropy H reflects the average amount of information contained in the sub-region. The larger the information entropy, the richer the image content and the more features. Its calculation formula is:

[0058]

[0059] in, is the ratio of the pixel value in the i-th row and j-th column in the sub-region to the total pixel value, and lg represents the logarithm.

[0060] The Frieden grayscale entropy F describes the distribution information of the grayscale value of the sub-region, reflecting the sharpness and clarity of the image features. The larger the Frieden grayscale entropy, the smoother the image grayscale changes and the less information. Its calculation formula is:

[0061]

[0062] Among them, p ij The definition of is the same as that in information entropy, and e is a natural constant.

[0063] calculate Figure 3 The image adaptability feature indicators of each sub-region in the shown area are shown in Table 1.

[0064] Table 1: Image adaptability characteristic indexes of this embodiment

[0065]

[0066] Step 6: Based on the comprehensive evaluation results of each sub-region, select the infrared reference matching map that meets the requirements.

[0067] From the candidate adaptation images generated in step 4, a sub-region that meets the infrared image adaptability feature evaluation standard is selected as the infrared reference adaptation area of ​​the current region.

[0068] According to the calculated characteristic values ​​of each partition in Table 1, the results show that sub-regions 1 to 4 all meet the above evaluation criteria, so these sub-regions can be used as reference adaptation areas of infrared images.

[0069] Combining the reference adaptation areas of all block areas can form a reference map with high quality and high matching accuracy, which can be used for image matching technology in the field of satellite navigation.

[0070] The present invention proposes a method for selecting a reference adaptation area for an earth infrared image based on K-means clustering and weighted Voronoi diagram. This technology simplifies the reference image selection process into a classification problem, uses the key corner points extracted by the Harris corner detection algorithm as the classification object, and combines the improved weighted Voronoi diagram to achieve autonomous selection of the adaptation area, ensure the correlation of feature points in the adaptation reference area, reduce the influence of human subjective factors, and improve work efficiency.

[0071] In addition, by combining multiple image feature information (such as grayscale standard deviation, information entropy, Frieden grayscale entropy), matching areas with significant features, rich information, and prominent structure are further extracted, thereby improving the quality of the reference image and the accuracy of matching. The present invention enriches the selection method of infrared image matching reference areas, provides important data support for infrared image navigation technology, and has broad application prospects in the fields of image matching and satellite navigation.

[0072] The above content describes the technical features of the present invention in detail through embodiments, but the specific forms described in the embodiments should not be regarded as limiting the content of the present invention. Various modifications, variations or alternatives based on the core idea of ​​the present invention should all fall within the scope of protection of this patent. The scope of protection of the present invention shall be subject to the claims and shall not be limited by the specific embodiments.

Claims

1. A method for selecting a reference adaptation area for an earth infrared image, characterized in that: include: Step 1: Obtain an infrared radiation image of the target area as an infrared radiation reference map, divide the infrared radiation reference map into several block areas, and perform steps 2 to 6 for each block area; Step 2: Detect the key corner points of the current block area and cluster the key corner points to obtain k cluster center points; Step 3: Use the weighted Voronoi diagram method to calculate the influence area of ​​each cluster center point in the current block area, and adjust the number of cluster classifications k according to the area ratio of each influential area; Step 4: Repeat steps 2-3 until the area ratios of the influential regions in the current block area are all within the range of 0.8 to 1.2; solve the rectangular envelope of each influential region finally obtained, and the envelope range is the corresponding sub-region; Step 5: Count the infrared image adaptability feature indicators of each sub-region; Step 6: Establish an infrared image adaptability feature evaluation standard, and the sub-region that meets the evaluation standard is the reference adaptation area of ​​the current block area.

2. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 1, characterized in that: The target area is the entire world; and obtaining the infrared radiation image of the target area is specifically: Combined with MODIS remote sensing images, the earth background infrared radiation model in the 8-14μm band is established: I 8-14 =∫8 14 τ(λ)[∈B(λ,T)+(1-∈)R rfl (l)]+R at (λ)↑dλ Among them I 8-14 is the infrared radiation intensity in the 8-14μm band, ∈ is the surface emissivity, τ(λ) is the atmospheric transmittance, R rfl (λ) is the atmospheric radiation and solar radiation reflected downward, R at (λ)↑ is the atmospheric radiation reflected upward; The infrared radiation intensity of each point on the earth is obtained based on the earth's background infrared radiation model, and the infrared radiation image of the earth is constructed according to the infrared radiation intensity of each point on the earth.

3. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 1, characterized in that: Before executing steps 2-6, remove the tiles that only include ocean.

4. The method for selecting a reference adaptation area for an earth infrared image according to claim 1, characterized in that: The detecting of the key corner points of the current block area specifically comprises performing key corner point detection on the current block area using a Harris corner point extraction algorithm.

5. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 1, characterized in that: The key corner points are clustered as follows: K-means clustering algorithm is used to select k key corner points in the current block area as cluster center points, each key corner point and its closest cluster center point are classified into one category according to the Euclidean distance, and the position of the cluster center point is continuously updated iteratively to minimize the objective function F(z, C) of the K-means clustering algorithm. The objective function F(z, C) is: where p i represents the i-th key corner point, c j is the jth cluster center, z ij Represents the key corner point p1 and the cluster center point c j The ownership relationship, if ||p i -c j || 2 =min 1≤j≤k ||p i -c j || 2 , then z ij =1, otherwise z ij =0; cluster center c j Each iteration of is updated according to the following formula: in is the summation symbol.

6. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 1, characterized in that: The weighted Voronoi diagram method is used to calculate the influence area of ​​each classification in the current block area, and the calculation formula is: Among them, V(c j ) represents the cluster center point c j The area of ​​influence, j≠i {.} means finding the intersection of all sets satisfying j≠i, s p represents the pth pixel in the reference image, q is the number of pixels in the reference image, is pixel s p and cluster center c j The weighted Euclidean distance between p ,c j ) is pixel s p To cluster center c j The Euclidean distance, w j is the weight of the jth category, and the weight of each category is proportional to the number of key corner points contained in each category.

7. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 1, characterized in that: The method of adjusting the number of clustering categories k according to the area ratio of each influential region is specifically as follows: The area ratio of each influential region is When the maximum value of the influence ratio of each category is maxr ij >1.2, the number of categories k is increased by 1; when maxr ij <0.8, the number of categories k is reduced by 1.

8. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 1, characterized in that: The infrared image adaptability characteristic indicators include grayscale standard deviation Std, information entropy H and Frieden grayscale entropy F; The calculation formula of the grayscale standard deviation Std is: Where a and b are the number of rows and columns of pixels in the sub-region, and f(i,j) represents the pixel value located in the i-th row and j-th column in the sub-region. is the average gray value of the sub-region; The calculation formula of the information entropy H is: in, is the ratio of the pixel value in the i-th row and j-th column of the sub-region to the total pixel value, and lg represents the logarithm; The calculation formula of the Frieden grayscale entropy F is: Among them, e is a natural constant.

9. The method for selecting a reference adaptation area of ​​an earth infrared image according to claim 8, characterized in that: The infrared image adaptability feature evaluation criteria are: grayscale standard deviation Std≥15, information entropy H≥5, and Frieden grayscale entropy F≤3.

Images