Large-view-field multi-view-angle snowflake synchronous matching and three-dimensional positioning method

Abstract

The invention provides a large-view-field multi-view-angle snowflake synchronous matching and three-dimensional positioning method, and the method comprises the steps: carrying out the particle detection of snowflakes through a telecentric camera, and obtaining a single-view-angle mass center coordinate; projecting the single-view centroid coordinates of the first plane to a three-dimensional space to obtain a first projection straight line; projecting the first projection straight line to a second plane and a third plane according to geometric mapping parameters to obtain a second projection straight line and a third projection straight line; calculating Euclidean distances of all particles on the second projection straight line and the third projection straight line, and screening according to the Euclidean distances to obtain candidate particles; performing multi-view reverse mapping on the candidate particles, and determining a three-dimensional positioning cube in combination with the maximum physical size of the particles; and performing positioning verification on the three-dimensional positioning cube through the geometric mapping parameters to obtain a verification result. According to the method, on the basis of the geometric constraint model, the mass center of the single-view particle is mapped to the three-dimensional space to form the space straight line, and the three-dimensional reconstruction calculation efficiency and the snowflake measurement accuracy are improved.

Description

Technical Field

[0001] The present invention relates to the field of atmospheric science observation technology, and in particular to a large-field-of-view multi-perspective snowflake synchronous matching and three-dimensional positioning method. Background Art

[0002] The precise measurement of microphysical characteristics such as the three-dimensional shape and movement speed of snowflakes is an important basic data for atmospheric science, meteorological radar remote sensing and quantitative estimation of winter precipitation. Among the existing technologies for precise measurement of snowflakes, the two-dimensional video raindrop spectrometer (2DVD) based on dual-line array scanning has morphological splicing distortion. The orthogonal array camera solution (ZL20161095178.0) is difficult to deconstruct the microscopic fractal characteristics of snowflakes. The Multi-Angle Snowflake Camera (MASC) achieves three-dimensional reconstruction through multi-view imaging combined with the Visual Hull algorithm, but its field of view is narrow and the depth of field is very small. It can only capture a small number of snowflakes at a time and cannot meet the large-scale sampling requirements of particle size distribution (PSD) statistics at the minute level. In addition, MASC relies on the morphological features of the two-dimensional image of snowflakes (such as edges and fractal structures) for matching, and its narrow field of view makes it impossible to process high-density particles under a large field of view.

[0003] In response to the particle measurement needs in large sampling spaces, the particle multi-angle stereoscopic imaging measurement device uses three-angle telecentric cameras to expand the effective observation volume to 775cm 3 Although the system has achieved 3D reconstruction to a certain extent, it still faces many core challenges in matching and positioning: First, the mismatch problem caused by highly similar particles: Traditional methods rely on the two-dimensional morphological features of snowflakes (such as edges and fractal structures) for cross-view matching, but in large fields of view or dense snowflake scenes, a large number of particles are highly similar in morphology, resulting in intensified matching conflicts and a significant increase in the mismatch rate; Second, the bottleneck of computational efficiency under large fields of view: As the observation volume expands to 775cm 3 , the number of particles surges, and the computational complexity of the matching algorithm based on morphological feature traversal search grows exponentially, making it difficult to meet real-time requirements; third, the geometric constraints of telecentric imaging fail: the parallel light imaging characteristics of the telecentric camera eliminate the influence of distance on the imaging size, and the traditional morphological matching logic that relies on perspective projection (such as the size consistency assumption) is no longer applicable. A new geometric constraint model is urgently needed to achieve spatial association; fourth, the three-dimensional space calculation domain positioning problem: the traditional method needs to traverse the entire sampling space for calculation, facing a computing bottleneck. The increase in the observation volume leads to an exponential increase in the number of particles, further exacerbating the computational burden. Therefore, it is very necessary to design a large-field-of-view multi-perspective snowflake synchronous matching and three-dimensional positioning method. Summary of the invention

[0004] The purpose of the present invention is to provide a large-field-of-view multi-perspective snowflake synchronous matching and three-dimensional positioning method. Based on the geometric constraint model, the single-perspective particle center of mass is reversely mapped to the world coordinate system to form a spatial straight line to achieve accurate matching, thereby improving the efficiency of three-dimensional reconstruction calculation and the accuracy of snowflake measurement.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A method for synchronous matching and three-dimensional positioning of snowflakes with a large field of view and multiple viewing angles comprises the following steps:

[0007] The snowflakes are detected by three telecentric cameras to obtain three single-view centroid coordinates.

[0008] Reversely project the coordinates of the centroid of the single-view of the first plane into the three-dimensional space to obtain a first projection line;

[0009] Projecting the first projection straight line onto the second plane and the third plane respectively according to the geometric mapping parameters to obtain a second projection straight line and a third projection straight line;

[0010] Calculate the Euclidean distances of all particles on the second projection line and the third projection line, and screen them according to the Euclidean distances to obtain candidate particles;

[0011] Perform multi-view reverse mapping on the candidate particles and determine the three-dimensional positioning cube based on the maximum physical size of the particles in the two-dimensional image;

[0012] The three-dimensional positioning cube is positioned and verified through geometric mapping parameters to obtain verification results.

[0013] Optionally, three telecentric cameras are used to perform particle detection on the snowflakes respectively to obtain three single-view centroid coordinates, specifically: the snowflake image is binarized based on the local grayscale features of the snowflakes to obtain a binary image, and the geometric center coordinates of the binary image are calculated to obtain the single-view centroid coordinates.

[0014] Optionally, the expression of binarization processing is: T(u,v)=μ(u,v)-C; wherein T(u,v) is the segmentation threshold, (u,v) is the two-dimensional coordinate of the pixel, μ(u,v) is the neighborhood grayscale mean, and C is the sensitivity coefficient;

[0015] The calculation formula of the single-view centroid coordinates is: Where N is the total number of pixels in the connected area of ​​the binary image, u i is the horizontal coordinate of the ith pixel, v i is the ordinate of the i-th pixel.

[0016] Optionally, the single-view centroid coordinates of the first plane are reversely projected into three-dimensional space to obtain a first projection line. Specifically, the relationship between the three-dimensional coordinates and the two-dimensional coordinates of the single-view centroid coordinates is reversely mapped and solved according to the geometric constraint projection equation of telecentric imaging to obtain the first projection line.

[0017] Optionally, the expression of the geometric constraint projection equation is: Among them, KM 0 is the geometric mapping parameter of the first plane, (X w , Y w , Z w ) is the three-dimensional coordinate of the single-view centroid coordinate of the first plane, (u 0 ,v 0 ) is the two-dimensional coordinate of the single-view centroid coordinate of the first plane;

[0018] The expression of the first projection line is: Among them, U is the direction vector, V is the base point, and t is an arbitrary real number.

[0019] Optionally, the expression of the second projection line is: 1 =a 1 t+b 1 ;v 1 =c 1 t+d 1 ; Among them, (u 1 ,v 1 ) is the two-dimensional coordinate of the single-view centroid coordinate of the second plane, (a 1 , c 1 ) is the direction parameter of the second projection line, (b 1 , d 1 ) is the intercept parameter of the second projection line;

[0020] The expression of the third projection line is: 2 =a 2 t+b 2 ;v 2 =c 2 t+d 2 ; Among them, (u 2 ,v 2 ) is the two-dimensional coordinate of the single-view centroid coordinate of the third plane, (a 2 , c 2 ) is the direction parameter of the third projection line, (b 2 , d 2 ) is the intercept parameter of the third projection line.

[0021] Optionally, the Euclidean distance is calculated as: Among them, (u, v) is the two-dimensional coordinate of the pixel, d is the Euclidean distance, and a, b and c are the coefficients of the equation.

[0022] Optionally, performing multi-view reverse mapping on the candidate particles and determining a three-dimensional positioning cube in combination with the maximum physical size of the particles in the two-dimensional image comprises:

[0023] Obtaining a three-dimensional intersection point according to the first projection straight line and the third projection straight line;

[0024] Based on the three-dimensional intersection point and combined with the maximum physical size of the particle, the minimum enclosing cube side length is determined to obtain a three-dimensional positioning cube.

[0025] Optionally, the three-dimensional positioning cube is positioned and verified by using geometric mapping parameters to obtain verification results, including:

[0026] Project the three-dimensional intersection point onto the third plane through the third projection line to obtain a third projection point;

[0027] If the third projection point is within the particle contour of the third plane, the verification result is determined as successful positioning;

[0028] If the third projection point is not located within the particle contour of the third plane, the verification result is determined as a positioning failure.

[0029] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes provided by the present invention includes: performing particle detection on snowflakes respectively through three telecentric cameras to obtain three single-view centroid coordinates; reversely projecting the single-view centroid coordinates of the first plane to three-dimensional space to obtain a first projection line; projecting the first projection line to the second plane and the third plane respectively according to geometric mapping parameters to obtain a second projection line and a third projection line; calculating the Euclidean distance of all particles on the second projection line and the third projection line, and screening according to the Euclidean distance to obtain candidate particles; performing multi-view reverse mapping on the candidate particles, and determining the three-dimensional positioning cube in combination with the maximum physical size of the particles in the two-dimensional image; and performing positioning verification on the three-dimensional positioning cube through geometric mapping parameters to obtain verification results. Based on the geometric constraint model, the method reversely maps the centroid of the single-view particle to the world coordinate system to form a spatial line, realizes accurate matching, and improves the efficiency of three-dimensional reconstruction calculation and the accuracy of snowflake measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0031] Figure 1 It is a flow chart of the multi-view snowflake three-dimensional positioning method of the present invention;

[0032] Figure 2 It is a schematic diagram of the steps of determining a three-dimensional positioning cube of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] like Figure 1 As shown, the present invention provides a large-field multi-view snowflake synchronous matching and three-dimensional positioning method, comprising the following steps:

[0036] Step 100: Use three telecentric cameras to perform particle detection on snowflakes respectively to obtain three single-view centroid coordinates; specifically, the following steps are included:

[0037] The snowflake image is binarized by using a background separation method based on the local grayscale features of snowflakes, and the segmentation threshold is adjusted based on the local grayscale mean to segment the snowflakes and the background, thereby obtaining a binary image.

[0038] Specifically, the expression of binarization processing is:

[0039] T(u,v)=μ(u,v)-C;

[0040] Wherein, T(u,v) is the segmentation threshold, (u,v) is the two-dimensional coordinate of the pixel, μ(u,v) is the neighborhood grayscale mean, and C is the sensitivity coefficient, which is 0.4 in this embodiment;

[0041] The geometric center coordinates are calculated according to the connected areas of the binary image to obtain the single-view centroid coordinates.

[0042] Specifically, the calculation formula of the single-view centroid coordinates is:

[0043]

[0044] Where N is the total number of pixels in the connected area of ​​the binary image, u i is the horizontal coordinate of the ith pixel, v i is the ordinate of the i-th pixel.

[0045] Step 200: reversely projecting the single-view centroid coordinates of the first plane into three-dimensional space to obtain a first projection line;

[0046] Specifically, assuming that the coordinates of the centroid of a single view in the three-dimensional space WCS are (X w , Y w , Z w ), the two-dimensional coordinates in the first plane are (u 0 ,v 0 ), the geometric mapping parameter KM corresponding to the first plane 0 is a 2×4 matrix, so the expression of the geometric constraint projection equation of the first plane is: The visual column is reversely mapped by (u 0 ,v 0 ) and KM 0 Solve (X w , Y w , Z w ) process, and after solving it, we get the first projection line L in the three-dimensional space. 1 , L 1 The parameters are expressed as:

[0047]

[0048] Among them, U is the direction vector, V is the base point, and t is an arbitrary real number.

[0049] Step 300: projecting the first projection straight line onto the second plane and the third plane respectively according to the geometric mapping parameters to obtain a second projection straight line and a third projection straight line;

[0050] Specifically, L 1 The geometric mapping parameters KM of the second and third planes are respectively 1 and KM 2 Projecting onto the second and third planes, we obtain the geometric constraint projection equations of the second and third planes, respectively. The expressions are:

[0051]

[0052] Among them, (u 1 ,v 1) is the two-dimensional coordinate of the single-view centroid coordinate of the second plane, (u 2 ,v 2 ) is the two-dimensional coordinate of the single-view centroid coordinate of the third plane, X w (t), Y w (t) and Z w (t) are the parametric equations of the coordinates of the projection line in three-dimensional space. After solving them, we can get the second projection line L 2 and the third projection line L 3 , and their linear function expressions are:

[0053] L 2 :u 1 =a 1 t+b 1 ;v 1 =c 1 t+d 1 ;

[0054] L 3 :u 2 =a 2 t+b 2 ;v 2 =c 2 t+d 2 ;

[0055] Among them, u 1 and v 1 The range of u is within the image range of the second plane, 2 and v 2 The range is within the image range of the third plane, (a 1 , c 1 ) is the direction parameter of the second projection line, and the geometric mapping parameter KM of the second plane 1 OK, (b 1 , d 1 ) is the intercept parameter of the second projection line, from KM 1 and the base point of the first projection line, (a 2 , c 2 ) is the direction parameter of the third projection line, and the geometric mapping parameter KM of the third plane 2 OK, (b 2 , d 2 ) is the intercept parameter of the third projection line, from KM 2 Determined together with the base point of the first projection line.

[0056] Step 400: Calculate the Euclidean distances of all particles on the second projection line and the third projection line, and screen according to the Euclidean distances to obtain candidate particles;

[0057] Specifically, the Euclidean distance from the center of mass of all particles on all projection lines to the corresponding projection lines is calculated. The calculation formula of the Euclidean distance is:

[0058]

[0059] Where d is the Euclidean distance, ax+by+c=0 is the polar equation, a, b and c are the polar equation coefficients, and the geometric mapping parameters KM of the second and third planes are 1 and KM 2 It is derived from the basic matrix and represents the mathematical constraint relationship of the projection line in the two-dimensional plane. If d≤∈, the particle is determined as a candidate particle and the remaining particles are eliminated. The threshold ∈ in this embodiment is 2 pixels.

[0060] Step 500: Perform multi-view reverse mapping on the candidate particles, and determine the three-dimensional positioning cube in combination with the maximum physical size of the particles in the two-dimensional image;

[0061] Specific as Figure 2 As shown in the figure, the candidate particle is a determined particle, and its centroid coordinates in the first plane, the second plane and the third plane are defined as P 1 , P 2 and P 3 Through step 200, the projection line L of the candidate particle is obtained. 1 ; Then, through step 300, the projection line L of the candidate particle is obtained. 2 , and in L 2 Find the 2 The nearest point P 2 '; Similarly, point P is obtained through step 300 2 'The projection line L 3 , and according to L 1 and L 3 Get the three-dimensional intersection point P c Because L 1 With L 3 In the three-dimensional WCS, they are bound to intersect, so P c It is regarded as the center of mass of the three-dimensional positioning cube. The minimum enclosing cube side length is determined by the maximum physical size of the particle, and finally the three-dimensional positioning cube is obtained.

[0062] Step 600: Verify the positioning of the three-dimensional positioning cube through geometric mapping parameters to obtain a verification result. The specific steps are: project the three-dimensional intersection point to the third plane through the third projection line to obtain the third projection point; if the third projection point is within the particle contour of the third plane, the verification result is determined as a successful positioning; if the third projection point is not within the particle contour of the third plane, the verification result is determined as a failed positioning.

[0063] The beneficial effects of the present invention are as follows:

[0064] 1) The three-dimensional positioning of particles is realized by combining the intersection of visual columns, and the effective computational domain (three-dimensional positioning cube) for three-dimensional reconstruction of particles is determined;

[0065] 2) It abandons the reliance on the two-dimensional morphological features of snowflakes and achieves efficient matching of hundreds of particles per second in a large sampling space, thus improving the matching efficiency;

[0066] 3) The calculation domain is used to limit the range of particle three-dimensional reconstruction, which significantly reduces the calculation complexity and enables efficient batch processing.

[0067] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0068] The present invention uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only used to help understand the method and core ideas of the present invention. At the same time, for those skilled in the art, according to the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes, characterized in that: The steps include: The snowflakes are detected by three telecentric cameras to obtain three single-view centroid coordinates. Reversely project the single-view centroid coordinates of the first plane into three-dimensional space to obtain a first projection line; Projecting the first projection straight line onto the second plane and the third plane respectively according to the geometric mapping parameters to obtain a second projection straight line and a third projection straight line; Calculating the Euclidean distances of all particles on the second projection line and the third projection line, and screening according to the Euclidean distances to obtain candidate particles; Performing multi-view reverse mapping on the candidate particles, and determining a three-dimensional positioning cube in combination with the maximum physical size of the particles in the two-dimensional image; The three-dimensional positioning cube is positioned and verified by using the geometric mapping parameters to obtain a verification result.

2. The method for synchronous matching and three-dimensional positioning of snowflakes with a large field of view and multiple viewing angles according to claim 1, characterized in that: The snowflakes are respectively subjected to particle detection by three telecentric cameras to obtain three single-view centroid coordinates. Specifically, the snowflake image is binarized based on the local grayscale features of the snowflakes to obtain a binary image, and the geometric center coordinates of the binary image are calculated to obtain the single-view centroid coordinates.

3. The method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes according to claim 2, characterized in that: The expression of the binarization process is: T(u,v)=μ(u,v)-C; wherein T(u,v is the segmentation threshold, (u,v) is the two-dimensional coordinate of the pixel, μ(u,v) is the neighborhood grayscale mean, and C is the sensitivity coefficient; The calculation formula of the single-view centroid coordinates is: Where N is the total number of pixels in the connected area of ​​the binary image, u i is the horizontal coordinate of the ith pixel, v i is the ordinate of the i-th pixel.

4. The method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes according to claim 1, characterized in that: The single-view centroid coordinates of the first plane are reversely projected into three-dimensional space to obtain a first projection line. Specifically, the relationship between the three-dimensional coordinates and the two-dimensional coordinates of the single-view centroid coordinates is reversely mapped and solved by the visual column according to the geometric constraint projection equation of telecentric imaging to obtain the first projection line.

5. The method for synchronous matching and three-dimensional positioning of snowflakes with a large field of view and multiple viewing angles according to claim 4, characterized in that: The expression of the geometric constraint projection equation is: Wherein, KM0 is the geometric mapping parameter of the first plane, (X w , Y w , Z w ) is the three-dimensional coordinate of the single-view centroid coordinate of the first plane, and (u0, v0) is the two-dimensional coordinate of the single-view centroid coordinate of the first plane; The expression of the first projection line is: Among them, U is the direction vector, V is the base point, and t is an arbitrary real number.

6. The method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes according to claim 1, characterized in that: The expression of the second projection straight line is: u1=a1t+b1; v1=c1t+d1; Wherein, (u1, v1) is the two-dimensional coordinate of the single-view centroid coordinate of the second plane, (a1, c1) is the direction parameter of the second projection line, and (b1, d1) is the intercept parameter of the second projection line; The expression of the third projection straight line is: u2=a2t+b2; v2=c2t+d2; wherein, (u2, v2) are the two-dimensional coordinates of the single-view centroid coordinates of the third plane, (a2, c2) are the direction parameters of the third projection straight line, and (b2, d2) are the intercept parameters of the third projection straight line.

7. The method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes according to claim 1, characterized in that: The calculation formula of the Euclidean distance is: Among them, (u, v) is the two-dimensional coordinate of the pixel, d is the Euclidean distance, and a, b and c are the coefficients of the equation.

8. The method for synchronous matching and three-dimensional positioning of large-field multi-view snowflakes according to claim 1, characterized in that: Performing multi-view reverse mapping on the candidate particles and determining a three-dimensional positioning cube in combination with the maximum physical size of the particles in the two-dimensional image, including: Obtaining a three-dimensional intersection point according to the first projection straight line and the third projection straight line; Based on the three-dimensional intersection point and in combination with the maximum physical size of the particle, the minimum side length of the enclosing cube is determined to obtain the three-dimensional positioning cube.

9. The method for synchronous matching and three-dimensional positioning of snowflakes with a large field of view and multiple viewing angles according to claim 8, characterized in that: Performing positioning verification on the three-dimensional positioning cube by using the geometric mapping parameters to obtain a verification result, including: Projecting the three-dimensional intersection point onto the third plane through the third projection straight line to obtain a third projection point; If the third projection point is located within the particle contour of the third plane, the verification result is determined as successful positioning; If the third projection point is not located within the particle contour of the third plane, the verification result is determined as a positioning failure.

Images