CN118746257A - High-precision extraction method and system for optical slice center based on weighted dimensionality reduction decomposition - Google Patents

Abstract

The present invention discloses a method and system for extracting the center of a light slice with high precision based on weighted dimensionality reduction decomposition, and relates to the field of calibration of a line structured light vision measurement system in two-dimensional/three-dimensional vision measurement. The method and system for extracting the center of a light slice with high precision based on weighted dimensionality reduction decomposition mainly include: decomposing the "image layer space" with dimensionality reduction to obtain a grayscale vector, analyzing and calculating the grayscale vector to obtain the sub-pixel coordinates of the decomposition direction of the "image layer space", matching the sequence number in the vertical direction of the decomposition direction of the current grayscale vector, obtaining the center coordinates of the light slice corresponding to the grayscale vector, traversing and calculating the center coordinates of the light slice corresponding to all grayscale vectors, and obtaining the center of the light slice of the light strip pixel-level image. The implementation of the method and system for extracting the center of a light slice with high precision based on weighted dimensionality reduction decomposition provided by the present invention can improve the real-time performance and precision of extracting the center of a light slice of a single-line laser scene.

Description

High-precision extraction method and system for optical slice center based on weighted dimensionality reduction decomposition

Technical Field

The present invention relates to the field of calibration of line structured light vision measurement system in two-dimensional/three-dimensional vision measurement, and more specifically to a method and system for high-precision extraction of light slice center based on weighted dimensionality reduction decomposition.

Background Art

The optical slice laser scanning 3D measurement method is an important active vision technology. It is based on the principle of 3D laser scanning technology and has the advantages of wide measurement range, high measurement efficiency, and small equipment hardware size. It is widely used in national key development fields such as aerospace, automotive intelligent manufacturing, and ship intelligent manufacturing.

The principle of light-slice laser scanning 3D measurement is to project a light-slice single-line laser onto a heterogeneous machined surface to generate a light-slice contour that represents the corresponding cross section of the heterogeneous machined surface. After the imaging system acquires the light-slice contour, the sub-pixel center coordinates of the light-slice contour in the image space are first extracted, and the light-slice center coordinates in the image space are mapped to the physical space (calibration) by calibration to obtain the true height data of the entire light-slice contour. Specifically, through one-dimensional translation, high-precision 3D point cloud data of the heterogeneous machined surface can be obtained by unidirectional scanning perpendicular to the projection surface of the light-slice single-line laser. Therefore, extracting the sub-pixel center coordinates of the light-slice contour in the image space is an important prerequisite for the light-slice laser scanning 3D measurement method, and the calculation of the sub-pixel center coordinates of the light-slice contour in the image space with high efficiency, high precision, and high robustness is the key to ensuring the real-time, high precision, and high robustness of the line structured light vision measurement system.

Traditional methods for extracting the center of a light slice mainly include the Steger algorithm. The Steger algorithm is based on the Hessian matrix. It obtains the normal direction of the light slice contour in the image space through the Hessian matrix, and then obtains the sub-pixel position of the center of the light slice contour by finding the center point in the normal direction. This method has high robustness and high accuracy, but the method has a large amount of calculation, and it is difficult to achieve rapid extraction of the center line of the light slice contour, and it is difficult to meet the application occasions with high real-time requirements such as rapid measurement of three-dimensional reconstruction.

J. Zhang proposed a method for extracting the center line of the three-dimensional reconstruction center line structured light stripes with high precision. This method processes the image and extracts its sub-pixel center through the secondary optimization algorithm of the threshold segmentation method and the grayscale centroid method. The method based on J. Zhang can not only realize multi-line laser measurement, but also extract the center line of the line structured light stripes with high real-time performance. However, this method requires the use of Gaussian filtering to process the image, which is not only time-consuming but also has a negative edge blurring effect. In addition, the edge pixel points often change, so it is difficult to achieve high-precision and high-stability extraction of the center coordinates of the light stripes.

The patent "Method and device for extracting the center of line structured light based on Gaussian super Laplace distribution" (patent number: ZL202311135784.4) proposes a method for extracting the center of line structured light based on Gaussian super Laplace distribution. It uses Gaussian filtering to process the structured light image. It can remove the noise that affects the fitting accuracy without destroying the original shape and structure of the line structured light. This method has a certain noise resistance and meets certain accuracy requirements. However, in the method of determining the fitting area within the boundary area of the light strip in this method, the fitting interval is determined by expanding the maximum grayscale value of each column by twice the width of the line structured light strip on both sides. Because in the extraction of the structured light center, each column may have multiple and discontinuous maximum grayscale value points, which is easy to produce large errors, so this method has poor anti-interference and low accuracy.

Scholars such as Y.Gong, G.Liu, H.Huang and T.Song proposed an improved laser centerline extraction algorithm based on inner propulsion. This method first determines the normal vector of the laser line based on the center point fitting tangent equation, then calculates the sub-pixel coordinates of the center point normal vector, and finally uses linear smoothing to make the laser center point smoother. Compared with the grayscale centroid method, the root mean square error of this method is improved by 0.337 pixels, which increases the robustness of centerline extraction. However, the threshold-type light strip center point search algorithm proposed by the scholar obtains the center point coordinates through the center method, which will reduce the accuracy. However, this method has a large amount of calculation, and the time-consuming problem has not been well solved.

In summary, traditional centerline extraction methods are generally divided into two categories. The first category is the geometric center method represented by the threshold method. This type of method has strong real-time performance, strong anti-interference ability, high accuracy, but is more time-consuming. Although some scholars later performed image stereo correction and used Gaussian filtering to pre-process the image to reduce errors, it is still impossible to achieve rapid detection of complex objects. The second type of method is the energy center method represented by the grayscale centroid method. This type of method is suitable for multi-line laser measurement scenes. The structured light center is found through the contour information of the structured light. Therefore, this type of method is more used to match the multi-line laser center. It has high real-time performance but is greatly affected by noise and has low accuracy. In addition, the conventional use of the grayscale centroid method is to process two-dimensional scenes, while the multi-threshold domain weighted method mentioned in this embodiment processes the one-dimensional scene after the dimensionality reduction decomposition of the two-dimensional scene, which improves the real-time performance of the structured light center measurement. Later, some scholars proposed a secondary optimization algorithm for the grayscale centroid method, which has a certain improvement in accuracy compared to the grayscale centroid method, but still cannot meet the high-precision structured light center measurement. Therefore, the current line structured light vision measurement system has low real-time performance, low accuracy and low robustness.

Summary of the invention

The purpose of the present invention is to provide a method and system for high-precision extraction of light slice centers by weighted dimensionality reduction decomposition, which can improve the real-time performance and precision of light slice center extraction in a single-line laser scene.

The present invention provides a high-precision extraction method for the center of a light slice based on weighted dimensionality reduction decomposition, comprising: S1: acquiring a light strip pixel-level image, and obtaining a dimensionality reduction decomposition direction of an "image layer space" of the light strip pixel-level image and a vertical direction of the dimensionality reduction decomposition direction of the "image layer space" of the light strip pixel-level image according to the light strip pixel-level image; S2: graying the light strip pixel-level image, and decomposing the light strip pixel-level image according to the dimensionality reduction decomposition direction of the "image layer space" of the light strip pixel-level image, to obtain a gray vector of the "image layer space" of the light strip pixel-level image; S3: obtaining a sub-pixel coordinate of a decomposition direction of the "image layer space" of the gray vector according to the gray vector of the "image layer space" of the light strip pixel-level image; S4: matching a sequence number of a vertical direction of the decomposition direction of a current gray vector according to the sub-pixel coordinate of the decomposition direction of the "image layer space" of the gray vector, to obtain a light slice center coordinate corresponding to the gray vector; S5: traversing and calculating the light slice center coordinates corresponding to all gray vectors, to obtain a light slice center of the light strip pixel-level image.

Furthermore, step S1 of the above-mentioned high-precision extraction method of the light slice center based on weighted dimensionality reduction decomposition specifically includes: S11: acquiring a light strip pixel-level image, and obtaining the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space" according to the light strip pixel-level image; S12: obtaining the angle between the main direction of the light slice and the row direction of the "image layer space" and the angle between the main direction of the light slice and the column direction of the "image layer space" according to the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space"; S13: when the angle between the main direction of the light slice and the row direction of the "image layer space" is not less than When the angle between the main direction of the light slice and the column direction of the "image layer space" is less than the angle between the main direction of the light slice and the column direction of the "image layer space", the row direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the column direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the column direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image.

Furthermore, step S3 of the above-mentioned method for high-precision extraction of light slice center based on weighted dimensionality reduction decomposition specifically includes: obtaining the sub-pixel coordinates of the decomposition direction of the "image layer space" of the grayscale vector according to the grayscale vector of the "image layer space" of the light strip pixel-level image, such as the formula:

Wherein, _Pk is the sub-pixel coordinate in the decomposition direction of the “image layer space” of the kth grayscale vector, n is the number of sub-pixel coordinates existing in the retained area, _Pki is the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _Aki is the weight corresponding to the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _rki is the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, a is the first preset coefficient for calculating the grayscale value weight, b is the second preset coefficient for calculating the grayscale value weight, c is the third preset coefficient for calculating the grayscale value weight, Th is the grayscale baseline constant, T is the preset grayscale threshold, ^σ2 (.) is the variance function, Max(.) is the maximum value function, _VG is the total grayscale value of the entire image, _RA (T) is the probability that the pixel is classified into class A, _VA (T) is the average grayscale of the pixels assigned to class A, and _RB is the average grayscale of the pixels assigned to class A. (T) is the probability that the pixel is assigned to class B, _VB (T) is the average grayscale of the pixels assigned to class B, _Ri is the probability that the grayscale of the pixel is i, and _gi is the number of pixels with grayscale i in the image.

Further, step S4 of the above-mentioned method for high-precision extraction of the center of a light slice based on weighted dimensionality reduction decomposition specifically includes: according to the sub-pixel coordinates of the decomposition direction of the "image layer space" of the grayscale vector, matching the serial number in the vertical direction of the decomposition direction of the current grayscale vector, obtaining the coordinates of the center of the light slice corresponding to the grayscale vector, such as the formula:

P _k (R _k , C _k ) = (P _k , k)

Among them, _Pk ( _Rk , _Ck ) is the coordinate of the center of the light slice corresponding to the k-th grayscale vector, _Rk is the coordinate of the decomposition direction of the center of the light slice in the "image layer space", _Ck is the vertical coordinate of the decomposition direction of the center of the light slice in the "image layer space", _Pk is the sub-pixel coordinate of the decomposition direction of the "image layer space" of each k-th grayscale vector, and k is the serial number in the vertical direction of the decomposition direction of the current grayscale vector.

Furthermore, step S5 of the above-mentioned method for high-precision extraction of light slice centers based on weighted dimensionality reduction decomposition specifically includes: traversing and calculating the light slice center coordinates corresponding to all grayscale vectors to obtain the light slice center of the light strip pixel-level image, as shown in the formula:

{m}＝{P _k (R _k ,C _k )|k＝1,2,...,H}

Wherein, {m} is the center of the light slice of the light bar pixel level image, and H is the number of grayscale vectors.

The present invention also provides a system, comprising the following modules: an image acquisition module, configured to: acquire a light strip pixel-level image, and obtain a dimensionality reduction decomposition direction of the "image layer space" of the light strip pixel-level image and a vertical direction of the dimensionality reduction decomposition direction of the "image layer space" of the light strip pixel-level image according to the light strip pixel-level image; a vector acquisition module, configured to: grayscale the light strip pixel-level image, and decompose the light strip pixel-level image according to the dimensionality reduction decomposition direction of the "image layer space" of the light strip pixel-level image to obtain a grayscale vector of the "image layer space" of the light strip pixel-level image; decompose the coordinates The module is configured as follows: according to the grayscale vector of the "image layer space" of the light strip pixel-level image, the sub-pixel coordinates of the decomposition direction of the grayscale vector of the "image layer space" are obtained; the coordinate matching module is configured as follows: according to the sub-pixel coordinates of the decomposition direction of the "image layer space" of the grayscale vector, the serial number of the vertical direction of the decomposition direction of the current grayscale vector is matched to obtain the light slice center coordinates corresponding to the grayscale vector; the image light slice center module is configured as follows: the light slice center coordinates corresponding to all grayscale vectors are traversed and calculated to obtain the light slice center of the light strip pixel-level image.

Furthermore, the image acquisition module of the above system is specifically configured as follows: acquiring a light strip pixel-level image, and obtaining the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space" according to the light strip pixel-level image; obtaining the angle between the main direction of the light slice and the row direction of the "image layer space" and the angle between the main direction of the light slice and the column direction of the "image layer space" according to the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space"; when the angle between the main direction of the light slice and the row direction of the "image layer space" is not less than the angle between the main direction of the light slice and the column direction of the "image layer space", the angle between the main direction of the light slice and the row direction of the "image layer space" is obtained. direction, the column direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the row direction of the light stripe pixel level image is taken as the perpendicular direction of the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image; when the angle between the main direction of the light slice and the row direction of the "image layer space" is smaller than the angle between the main direction of the light slice and the column direction of the "image layer space", the row direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the column direction of the light stripe pixel level image is taken as the perpendicular direction of the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image.

Furthermore, the decomposition coordinate module of the above system is specifically configured as follows: according to the grayscale vector of the "image layer space" of the light strip pixel-level image, the decomposition direction sub-pixel coordinate of the "image layer space" of the grayscale vector is obtained, as shown in the formula:

Wherein, _Pk is the sub-pixel coordinate in the decomposition direction of the “image layer space” of the kth grayscale vector, n is the number of sub-pixel coordinates existing in the retained area, _Pki is the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _Aki is the weight corresponding to the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _rki is the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, a is the first preset coefficient for calculating the grayscale value weight, b is the second preset coefficient for calculating the grayscale value weight, c is the third preset coefficient for calculating the grayscale value weight, Th is the grayscale baseline constant, T is the preset grayscale threshold, ^σ2 (.) is the variance function, Max(.) is the maximum value function, _VG is the total grayscale value of the entire image, _RA (T) is the probability that the pixel is classified into class A, _VA (T) is the average grayscale of the pixels assigned to class A, and _RB is the average grayscale of the pixels assigned to class A. (T) is the probability that the pixel is assigned to class B, _VB (T) is the average grayscale of the pixels assigned to class B, _Ri is the probability that the grayscale of the pixel is i, and _gi is the number of pixels with grayscale i in the image.

Furthermore, the coordinate matching module of the above system is specifically configured as follows: according to the sub-pixel coordinates of the decomposition direction of the "image layer space" of the grayscale vector, the serial number in the vertical direction of the decomposition direction of the current grayscale vector is matched to obtain the center coordinates of the light slice corresponding to the grayscale vector, as shown in the formula:

P _k (R _k , C _k ) = (P _k , k)

Among them, _Pk ( _Rk , _Ck ) is the coordinate of the center of the light slice corresponding to the k-th grayscale vector, _Rk is the coordinate of the decomposition direction of the center of the light slice in the "image layer space", _Ck is the vertical coordinate of the decomposition direction of the center of the light slice in the "image layer space", _Pk is the sub-pixel coordinate of the decomposition direction of the "image layer space" of each k-th grayscale vector, and k is the serial number in the vertical direction of the decomposition direction of the current grayscale vector.

Furthermore, the image light slice center module of the above system is specifically configured as follows: traversing and calculating the light slice center coordinates corresponding to all grayscale vectors to obtain the light slice center of the light strip pixel-level image, as shown in the formula:

{m}＝{P _k (R _k ,C _k )|k＝1,2,...,H}

Wherein, {m} is the center of the light slice of the light bar pixel level image, and H is the number of grayscale vectors.

The method and system for high-precision extraction of light slice centers based on weighted dimensionality reduction decomposition provided by the present invention have the following beneficial effects:

A dimensionality reduction decomposition model of the two-dimensional grayscale matrix of the "image layer space" is constructed. By decomposing the two-dimensional grayscale matrix of the "image layer space" into an ordered "grayscale vector pool", the problem of extracting the two-dimensional coordinates of the center of the light slice in the "image layer space" is transformed into a problem of high-precision calculation of the center point of the vector in the "grayscale vector pool" and matching the vector sequence number, which reduces the difficulty of extracting the center of the light slice.

The grayscale distribution characteristic mechanism of each grayscale vector in the "grayscale vector pool" is studied, and a high-precision calculation model of the grayscale center point of the vector in the "grayscale vector pool" is constructed to improve the accuracy of light slice center extraction;

Based on the orderliness of the "grayscale vector pool", the sequence number of the vector in the "grayscale vector pool" can be matched with the grayscale center point of the vector through a loop traversal of the "grayscale vector pool". Compared with the two-dimensional matrix operation (convolution, refinement, morphological algorithm), the amount of calculation for extracting the center of the light slice is reduced, thereby improving the real-time performance of the extraction.

BRIEF DESCRIPTION OF THE DRAWINGS

The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

FIG1 is a flow chart of a method for high-precision extraction of optical slice centers based on weighted dimensionality reduction decomposition provided by the present invention;

FIG2 is a schematic diagram of the direction of dimensionality reduction decomposition of the “image layer space” provided by the present invention;

FIG3 is a schematic diagram of dimensionality reduction decomposition of the “image layer space” provided by the present invention in the column direction (perpendicular to the main direction of the light slice);

FIG4 is a schematic diagram of grayscale distribution of a grayscale vector provided by the present invention;

Among them, (a) is a schematic diagram of the grayscale distribution of the grayscale vector V ₁ , and (b) is a schematic diagram of the grayscale distribution of the grayscale vector V _K ;

FIG5 is a schematic diagram of a “grayscale-weight mapping curve” provided by the present invention;

Figure 6 is a schematic diagram of the implementation method of the grayscale vector serial number matching method in the "grayscale vector pool" taking V _K as an example provided by the present invention.

DETAILED DESCRIPTION

In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

Fig. 1 is a schematic diagram of a method for extracting light slice centers with high precision based on weighted dimensionality reduction decomposition in this embodiment. In this embodiment, the method for extracting light slice centers with high precision based on weighted dimensionality reduction decomposition includes:

S1: Obtain a light stripe pixel-level image, and obtain a dimensionality reduction decomposition direction of the “image layer space” of the light stripe pixel-level image and a vertical direction of the dimensionality reduction decomposition direction of the “image layer space” of the light stripe pixel-level image according to the light stripe pixel-level image;

Specifically, step S1 includes: S11: acquiring a light strip pixel-level image, and obtaining a main direction of a light slice, a row direction of an "image layer space" of the light strip pixel-level image, and a column direction of an "image layer space" according to the light strip pixel-level image; S12: obtaining an angle between the main direction of the light slice and the row direction of the "image layer space" and an angle between the main direction of the light slice and the column direction of the "image layer space" according to the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space"; S13: when the angle between the main direction of the light slice and the row direction of the "image layer space" is not less than the angle between the main direction of the light slice and the column direction of the "image layer space", the angle between the main direction of the light slice and the row direction of the "image layer space" is obtained. When the angle between the main direction of the light slice and the row direction of the "image layer space" is less than the angle between the main direction of the light slice and the column direction of the "image layer space", the row direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the row direction of the light stripe pixel level image is taken as the vertical direction of the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image; when the angle between the main direction of the light slice and the row direction of the "image layer space" is less than the angle between the main direction of the light slice and the column direction of the "image layer space", the row direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the column direction of the light stripe pixel level image is taken as the vertical direction of the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image;

S2: graying the light stripe pixel-level image, decomposing the light stripe pixel-level image according to the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel-level image, and obtaining the grayscale vector of the "image layer space" of the light stripe pixel-level image;

S3: according to the grayscale vector of the “image layer space” of the pixel-level image of the light strip, the sub-pixel coordinates of the decomposition direction of the grayscale vector of the “image layer space” are obtained;

Specifically, step S3 includes: according to the grayscale vector of the "image layer space" of the pixel-level image of the light strip, obtaining the sub-pixel coordinates of the decomposition direction of the "image layer space" of the grayscale vector, such as the formula:

Wherein, _Pk is the sub-pixel coordinate in the decomposition direction of the “image layer space” of the kth grayscale vector, n is the number of sub-pixel coordinates existing in the retained area, _Pki is the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _Aki is the weight corresponding to the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _rki is the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, a is the first preset coefficient for calculating the grayscale value weight, b is the second preset coefficient for calculating the grayscale value weight, c is the third preset coefficient for calculating the grayscale value weight, Th is the grayscale baseline constant, T is the preset grayscale threshold, ^σ2 (.) is the variance function, Max(.) is the maximum value function, _VG is the total grayscale value of the entire image, _RA (T) is the probability that the pixel is classified into class A, _VA (T) is the average grayscale of the pixels assigned to class A, and _RB is the average grayscale of the pixels assigned to class A. (T) is the probability that a pixel is assigned to class B, V _B (T) is the average grayscale of pixels assigned to class B, R _i is the probability that the grayscale of a pixel is i, and g _i is the number of pixels with grayscale i in the image;

S4: According to the sub-pixel coordinates of the decomposition direction of the "image layer space" of the gray vector, the serial number of the vertical direction of the decomposition direction of the current gray vector is matched to obtain the center coordinates of the light slice corresponding to the gray vector;

Specifically, step S4 includes: according to the sub-pixel coordinates of the decomposition direction of the "image layer space" of the gray vector, matching the serial number of the vertical direction of the decomposition direction of the current gray vector, and obtaining the light slice center coordinates corresponding to the gray vector, such as the formula:

P _k (R _k , C _k ) = (P _k , k)

Wherein, P _k (R _k , C _k ) is the coordinate of the center of the light slice corresponding to the k-th grayscale vector, R _k is the coordinate of the decomposition direction of the center of the light slice in the "image layer space", C _k is the vertical coordinate of the decomposition direction of the center of the light slice in the "image layer space", P _k is the sub-pixel coordinate of the decomposition direction of the "image layer space" of each k-th grayscale vector, and k is the serial number of the vertical direction of the decomposition direction of the current grayscale vector;

S5: traverse and calculate the light slice center coordinates corresponding to all grayscale vectors to obtain the light slice center of the light strip pixel-level image;

Specifically, step S5 includes: traversing and calculating the light slice center coordinates corresponding to all grayscale vectors to obtain the light slice center of the light strip pixel level image, such as the formula:

{m}＝{P _k (R _k ,C _k )|k＝1,2,...,H}

Wherein, {m} is the center of the light slice of the light bar pixel level image, and H is the number of grayscale vectors.

This embodiment provides a system, including the following modules:

The image acquisition module is configured to: acquire the light strip pixel level image, and obtain the "image layer space" dimensionality reduction decomposition direction of the light strip pixel level image and the vertical direction of the "image layer space" dimensionality reduction decomposition direction of the light strip pixel level image according to the light strip pixel level image;

Specifically, the image acquisition module of the above system is configured as follows: acquiring a light strip pixel-level image, and obtaining the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space" according to the light strip pixel-level image; obtaining the angle between the main direction of the light slice and the row direction of the "image layer space" and the angle between the main direction of the light slice and the column direction of the "image layer space" according to the main direction of the light slice, the row direction of the "image layer space" of the light strip pixel-level image, and the column direction of the "image layer space"; when the angle between the main direction of the light slice and the row direction of the "image layer space" is not less than the angle between the main direction of the light slice and the column direction of the "image layer space", the angle between the main direction of the light slice and the row direction of the "image layer space" is obtained. , the column direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the row direction of the light stripe pixel level image is taken as the vertical direction of the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image; when the angle between the main direction of the light slice and the row direction of the "image layer space" is smaller than the angle between the main direction of the light slice and the column direction of the "image layer space", the row direction of the light stripe pixel level image is taken as the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and the column direction of the light stripe pixel level image is taken as the vertical direction of the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image;

The vector acquisition module is configured to: grayscale the light stripe pixel level image, decompose the light stripe pixel level image according to the dimensionality reduction decomposition direction of the "image layer space" of the light stripe pixel level image, and obtain the grayscale vector of the "image layer space" of the light stripe pixel level image;

The decomposition coordinate module is configured to: obtain the decomposition direction sub-pixel coordinates of the "image layer space" of the grayscale vector according to the grayscale vector of the "image layer space" of the pixel-level image of the light strip;

Specifically, the decomposition coordinate module of the above system is configured as follows: according to the grayscale vector of the "image layer space" of the pixel-level image of the light strip, the decomposition direction sub-pixel coordinate of the "image layer space" of the grayscale vector is obtained, as shown in the formula:

Wherein, _Pk is the sub-pixel coordinate in the decomposition direction of the “image layer space” of the kth grayscale vector, n is the number of sub-pixel coordinates existing in the retained area, _Pki is the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _Aki is the weight corresponding to the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, _rki is the grayscale value of the i-th sub-pixel coordinate point in the “image layer space” decomposition direction of the kth grayscale vector, a is the first preset coefficient for calculating the grayscale value weight, b is the second preset coefficient for calculating the grayscale value weight, c is the third preset coefficient for calculating the grayscale value weight, Th is the grayscale baseline constant, T is the preset grayscale threshold, ^σ2 (.) is the variance function, Max(.) is the maximum value function, _VG is the total grayscale value of the entire image, _RA (T) is the probability that the pixel is classified into class A, _VA (T) is the average grayscale of the pixels assigned to class A, and _RB is the average grayscale of the pixels assigned to class A. (T) is the probability that a pixel is assigned to class B, V _B (T) is the average grayscale of pixels assigned to class B, R _i is the probability that the grayscale of a pixel is i, and g _i is the number of pixels with grayscale i in the image;

The coordinate matching module is configured to: match the sub-pixel coordinates of the decomposition direction of the grayscale vector in the "image layer space" to the serial number in the vertical direction of the decomposition direction of the current grayscale vector, and obtain the center coordinates of the light slice corresponding to the grayscale vector;

Specifically, the coordinate matching module of the above system is configured as follows: according to the sub-pixel coordinates of the decomposition direction of the "image layer space" of the grayscale vector, the serial number of the vertical direction of the decomposition direction of the current grayscale vector is matched to obtain the center coordinates of the light slice corresponding to the grayscale vector, as shown in the formula:

P _k (R _k , C _k ) = (P _k , k)

Wherein, P _k (R _k , C _k ) is the coordinate of the center of the light slice corresponding to the k-th grayscale vector, R _k is the coordinate of the decomposition direction of the center of the light slice in the "image layer space", C _k is the vertical coordinate of the decomposition direction of the center of the light slice in the "image layer space", P _k is the sub-pixel coordinate of the decomposition direction of the "image layer space" of each k-th grayscale vector, and k is the serial number of the vertical direction of the decomposition direction of the current grayscale vector;

The image light slice center module is configured as follows: traversing and calculating the light slice center coordinates corresponding to all grayscale vectors to obtain the light slice center of the light strip pixel-level image;

Specifically, the image light slice center module of the above system is configured as follows: traverse and calculate the light slice center coordinates corresponding to all grayscale vectors to obtain the light slice center of the light strip pixel-level image, as shown in the formula:

{m}＝{P _k (R _k ,C _k )|k＝1,2,…,H}

Wherein, {m} is the center of the light slice of the light bar pixel level image, and H is the number of grayscale vectors.

In some embodiments, the above-mentioned method for high-precision extraction of light slice centers based on weighted dimensionality reduction decomposition can also be implemented by the following steps:

Step 1. Select the dimensionality reduction decomposition direction of the "image layer space", assuming that the resolution of the "image layer space" is W×H:

1) Determine the main direction of the light slice;

2) Analyze the angle α between the main direction of the light slice and the row direction of the “image layer space”;

3) Analyze the angle β between the main direction of the light slice and the column direction of the “image layer space”;

4) Compare the size of α and β. When α≥β, perform dimensionality reduction decomposition on the “image layer space” in the column direction; when α＜β, perform dimensionality reduction decomposition on the “image layer space” in the row direction, and grayscale the current image. Next, take the dimensionality reduction decomposition on the “image layer space” in the column direction as an example. Similarly, the dimensionality reduction decomposition on the “image layer space” in the row direction is similar. The schematic diagram of the dimensionality reduction decomposition direction selection of the “image layer space” is shown in Figure 2;

Step 2. Decompose the “image layer space” in the column direction:

Assuming that the resolution of the “image layer space” is W×H, the “image layer space” can be decomposed into a “grayscale vector pool” consisting of H grayscale vectors {V _k |k＝1,2...H} according to column vector dimensionality reduction, as shown in FIG3 ; taking the grayscale vector V _k as an example, the corresponding light slice center coordinates on the grayscale vector are:

PA _k (R _k , C _k ) = (P _k , k)

Where _Rk is the column coordinate of the center of the light slice in the "image layer space"; _Ck is the row coordinate of the center of the light slice in the "image layer space"; _Pk is the sub-pixel coordinate of the gray vector _Vk in the "gray vector pool" in the "image layer space" column direction; k is the vector number of the gray vector _Vk in the row direction;

Step 3. High-precision calculation method of the center point of the grayscale vector of the "grayscale vector pool" is implemented:

After the dimensionality reduction segmentation of the "image layer space", the extraction of the center of the light slice in the "image layer space" is transformed from a two-dimensional matrix operation problem into a center calculation problem of the grayscale curve of H grayscale vectors. Considering the gradient characteristics of the light slice grayscale distribution with the highest center and the lowest on both sides, and in order to suppress the grayscale interference of ambient light, this embodiment proposes a vector adaptive threshold algorithm, which sets the grayscale baseline constant Th based on each vector. The calculation method of Th is as follows:

Let g _i be the number of pixels with gray level i in the image, then g ₀ ,g ₁ ,g ₂ , then the probability that the grayscale of any point in the grayscale diagram is i is:

And there are:

Assume that the threshold is T, and the points in the image are divided into two categories, A and B, according to their corresponding grayscale, where A∈(0,T) and B∈(T+1,255); then, for any point selected, the probability of the point being classified into category A is _RA (T), and the average grayscale of the points assigned to category A is _VA (T):

Similarly, the probability of a point being assigned to class B is _RB (T), and the average grayscale of the points assigned to class B is _VB (T):

The total gray value V _G of the entire image:

So we know the variance is:

σ ² (T)＝ _RA (T)(V _A (T)-V _G ) ² +R _B (T)(V _B (T)-V _G ) ²

The grayscale baseline constant Th can be further determined by the following formula.

Further, we discuss two cases:

1) In the first case, there is no light slice distribution on the grayscale vector:

Taking the grayscale vector V ₁ in Figure 3 as an example, Figure 4(a) is a schematic diagram of the grayscale distribution of V _1. Since V ₁ has no light slice grayscale distribution, the entire grayscale distribution is lower than the grayscale baseline constant Th. In this case, it is considered that there is no effective center point on V ₁ , that is, there is no light slice distribution on the grayscale vector V ₁ .

2) In the second case, there is a light slice distribution on the grayscale vector:

Taking the grayscale vector V _K in Figure 3 as an example, Figure 4(b) is a grayscale distribution diagram of the grayscale vector V _K. Since there is a light slice distribution on V _K , it is considered that there is an effective center point on V _K. Therefore, the images whose grayscale distribution diagram of the grayscale vector V _K is greater than the grayscale baseline constant Th are retained, and the images whose grayscale distribution diagram is less than Th are removed, and processed by the following formula;

Where r _ki is the gray value of the sub-pixel coordinate point in the column direction of the i-th “image layer space” in the gray distribution diagram of the gray vector V _K ;

According to the grayscale mapping principle, different weights A _ki are set for different grayscale values r _ki ; the higher the grayscale value, the higher the reliability of the sub-pixel coordinate point in the column direction of the "image layer space", so the higher the grayscale value, the larger the specified A _ki , so in order to increase the proportion of high brightness, this embodiment proposes a "grayscale-weight mapping curve", as shown in Figure 5, this example in this embodiment is only one embodiment of the mapping, the curve is not limited to this one, and can be changed according to the scene and conditions. The function is as follows, and r _ki ∈ (Th, 255), in this example Th = 200:

Wherein, A _ki is the weight corresponding to the grayscale value of the grayscale value of the column direction sub-pixel coordinate point of the i-th “image layer space” in the grayscale distribution diagram of the grayscale vector V _K ;

Furthermore, the sub-pixel coordinate P _k in the column direction of the “image layer space” of the gray vector V _K can be calculated by the following formula:

In the formula, n is the number of sub-pixel coordinates in the reserved area, P _ki is the i-th sub-pixel coordinate point in the column direction of the "image layer space" in the grayscale distribution diagram of the grayscale vector V _K , and P _k is the sub-pixel coordinate R _K in the column direction of the "image layer space" of the center of the optical slice of V _K ;

Step 4. Grayscale vector serial number matching method in the "grayscale vector pool" is implemented:

Based on steps 1, 2, and 3, the calculation of the center points of all grayscale vectors in the "grayscale vector pool" can be realized by looping. Taking _Vk as an example, while obtaining the sub-pixel coordinates _Rk of the light slice center of _Vk , the index number of the row direction of the current grayscale vector is matched, and the matching of the light slice center coordinates ( _Pk , k) is realized by combining the following formula. FIG6 is a schematic diagram of the grayscale vector sequence matching method in the "grayscale vector pool" taking Vk as an example:

P _k (R _k , C _k ) = (P _k , k)

After traversing and calculating the H grayscale vectors in the "grayscale vector pool", the set {m} can be obtained, and finally a high-real-time extraction model for the center coordinates of the light slice in the "image layer space" is constructed:

{m}＝{P _k (R _k ,C _k )|k＝1,2,...,H}

At the same time, since the grayscale baseline constant Th method is used to calculate the center point, the influence of noise interference such as ambient light on the accuracy of light slice center extraction is suppressed.

The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims (10)

1. The high-precision extraction method of the optical slice center based on the weighted dimension reduction decomposition is characterized by comprising the following steps of:

S1: acquiring a light bar pixel level image, and obtaining the vertical direction of the dimension reduction decomposition direction of the image layer space of the light bar pixel level image and the dimension reduction decomposition direction of the image layer space of the light bar pixel level image according to the light bar pixel level image;

s2: graying the light bar pixel level image, and decomposing the light bar pixel level image according to the dimension-reducing decomposition direction of the image layer space of the light bar pixel level image to obtain a gray vector of the image layer space of the light bar pixel level image;

s3: according to the gray vector of the image layer space of the light bar pixel level image, sub-pixel coordinates of the gray vector in the decomposition direction of the image layer space are obtained;

s4: according to the sub-pixel coordinates of the gray level vector in the decomposition direction of the image layer space, matching the sequence number of the current gray level vector in the vertical direction of the decomposition direction to obtain the center coordinates of the optical slice corresponding to the gray level vector;

S5: and traversing the center coordinates of the optical slice corresponding to all the gray vectors to obtain the center of the optical slice of the optical stripe pixel level image.

2. The method for extracting the optical slice center with high precision based on weighted dimension reduction decomposition according to claim 1, wherein step S1 specifically comprises:

s11: acquiring a light bar pixel level image, and acquiring a light slice main direction, an image layer space row direction and an image layer space column direction of the light bar pixel level image according to the light bar pixel level image;

S12: obtaining an included angle between the main direction of the light slice and the row direction of the image layer space and an included angle between the main direction of the light slice and the column direction of the image layer space according to the main direction of the light slice, the row direction of the image layer space of the light strip pixel-level image and the column direction of the image layer space;

S13: when the included angle between the main direction of the light slice and the row direction of the image layer space is not smaller than the included angle between the main direction of the light slice and the column direction of the image layer space, the column direction of the light bar pixel level image is taken as the dimension-reducing decomposition direction of the image layer space of the light bar pixel level image, and the row direction of the light bar pixel level image is taken as the vertical direction of the dimension-reducing decomposition direction of the image layer space of the light bar pixel level image; when the included angle between the main direction of the light slice and the row direction of the image layer space is smaller than the included angle between the main direction of the light slice and the column direction of the image layer space, the row direction of the light bar pixel level image is taken as the dimension reduction decomposition direction of the image layer space of the light bar pixel level image, and the column direction of the light bar pixel level image is taken as the vertical direction of the dimension reduction decomposition direction of the image layer space of the light bar pixel level image.

3. The method for extracting the optical slice center with high precision based on weighted dimension reduction decomposition according to claim 1, wherein step S3 specifically comprises: according to the gray vector of the image layer space of the light bar pixel level image, the sub-pixel coordinates of the decomposing direction of the image layer space of the gray vector are obtained, such as the formula:

σ²(T)＝R_A(T)(V_A(T)-V_G)²+R_B(T)(V_B(T)-V_G)²

Wherein P _k is the sub-pixel coordinate of the decomposition direction of the "image layer space" of the kth gray level vector, n is the number of sub-pixel coordinates existing in the reserved area, P _ki is the sub-pixel coordinate point of the decomposition direction of the ith "image layer space" of the kth gray level vector, A _ki is the weight corresponding to the gray level value of the sub-pixel coordinate point of the decomposition direction of the ith "image layer space" of the kth gray level vector, R _ki is the gray level value of the sub-pixel coordinate point of the decomposition direction of the ith "image layer space" of the kth gray level vector, a is the 1 st preset coefficient for calculating the gray level weight, B is the 2 nd preset coefficient for calculating the gray level weight, c is the 3 rd preset coefficient of the calculated gray value weight, th is the gray base line constant, T is the preset gray threshold value, σ ² (-) is the variance function, max (-) is the maximum function, V _G is the gray total value of the whole image, R _A (T) is the probability that the pixel point is classified into class a, V _A (T) is the average gray level of the pixel point assigned to class a, R _B (T) is the probability that the pixel point is classified into class B, V _B (T) is the average gray level of the pixel point assigned to class B, R _i is the probability that the gray level of the pixel point is i, and g _i is the number of pixels in the image with gray level i.

4. The method for extracting the optical slice center with high precision based on weighted dimension reduction decomposition according to claim 1, wherein step S4 specifically comprises: according to the sub-pixel coordinates of the resolution direction of the image layer space of the gray level vector, matching the sequence number of the vertical direction of the resolution direction of the current gray level vector to obtain the center coordinates of the optical slice corresponding to the gray level vector, wherein the formula is as follows:

P_k(R_k,C_k)＝(P_k,k)

Wherein, P _k(R_k,C_k) is the optical slice center coordinate corresponding to the kth gray-scale vector, R _k is the optical slice center decomposition direction coordinate of the "image layer space", C _k is the vertical direction coordinate of the optical slice center decomposition direction of the "image layer space", P _k is the decomposition direction subpixel coordinate of the "image layer space" of each of the kth gray-scale vectors, and k is the number of the vertical direction of the decomposition direction of the current gray-scale vector.

5. The method for extracting the optical slice center with high precision based on weighted dimension reduction decomposition according to claim 1, wherein step S5 specifically comprises: traversing the center coordinates of the optical slice corresponding to all the gray vectors to obtain the center of the optical slice of the optical stripe pixel level image, wherein the formula is as follows:

{m}＝{P_k(R_k,C_k)|k＝1,2,...,H}

Wherein { m } is the center of the optical slice of the optical stripe pixel level image, and H is the number of gray vectors.

6. A system, the system comprising the following modules:

An image acquisition module configured to: acquiring a light bar pixel level image, and obtaining the vertical direction of the dimension reduction decomposition direction of the image layer space of the light bar pixel level image and the dimension reduction decomposition direction of the image layer space of the light bar pixel level image according to the light bar pixel level image;

A vector acquisition module configured to: graying the light bar pixel level image, and decomposing the light bar pixel level image according to the dimension-reducing decomposition direction of the image layer space of the light bar pixel level image to obtain a gray vector of the image layer space of the light bar pixel level image;

A decomposition coordinates module configured to: according to the gray vector of the image layer space of the light bar pixel level image, sub-pixel coordinates of the gray vector in the decomposition direction of the image layer space are obtained;

The coordinate matching module is configured to: according to the sub-pixel coordinates of the gray level vector in the decomposition direction of the image layer space, matching the sequence number of the current gray level vector in the vertical direction of the decomposition direction to obtain the center coordinates of the optical slice corresponding to the gray level vector;

an image light slice center module configured to: and traversing the center coordinates of the optical slice corresponding to all the gray vectors to obtain the center of the optical slice of the optical stripe pixel level image.

7. The system of claim 6, wherein the image acquisition module is specifically configured to:

acquiring a light bar pixel level image, and acquiring a light slice main direction, an image layer space row direction and an image layer space column direction of the light bar pixel level image according to the light bar pixel level image;

Obtaining an included angle between the main direction of the light slice and the row direction of the image layer space and an included angle between the main direction of the light slice and the column direction of the image layer space according to the main direction of the light slice, the row direction of the image layer space of the light strip pixel-level image and the column direction of the image layer space;

When the included angle between the main direction of the light slice and the row direction of the image layer space is not smaller than the included angle between the main direction of the light slice and the column direction of the image layer space, the column direction of the light bar pixel level image is taken as the dimension-reducing decomposition direction of the image layer space of the light bar pixel level image, and the row direction of the light bar pixel level image is taken as the vertical direction of the dimension-reducing decomposition direction of the image layer space of the light bar pixel level image; when the included angle between the main direction of the light slice and the row direction of the image layer space is smaller than the included angle between the main direction of the light slice and the column direction of the image layer space, the row direction of the light bar pixel level image is taken as the dimension reduction decomposition direction of the image layer space of the light bar pixel level image, and the column direction of the light bar pixel level image is taken as the vertical direction of the dimension reduction decomposition direction of the image layer space of the light bar pixel level image.

8. The system of claim 6, wherein the decomposition coordinate module is specifically configured to: according to the gray vector of the image layer space of the light bar pixel level image, the sub-pixel coordinates of the decomposing direction of the image layer space of the gray vector are obtained, such as the formula:

σ²(T)＝R_A(T)(V_A(T)-V_G)²+R_B(T)(V_B(T)-V_G)²

Wherein P _k is the sub-pixel coordinate of the decomposition direction of the "image layer space" of the kth gray level vector, n is the number of sub-pixel coordinates existing in the reserved area, P _ki is the sub-pixel coordinate point of the decomposition direction of the ith "image layer space" of the kth gray level vector, A _ki is the weight corresponding to the gray level value of the sub-pixel coordinate point of the decomposition direction of the ith "image layer space" of the kth gray level vector, R _ki is the gray level value of the sub-pixel coordinate point of the decomposition direction of the ith "image layer space" of the kth gray level vector, a is the 1 st preset coefficient for calculating the gray level weight, B is the 2 nd preset coefficient for calculating the gray level weight, c is the 3 rd preset coefficient of the calculated gray value weight, th is the gray base line constant, T is the preset gray threshold value, σ ² (-) is the variance function, max (-) is the maximum function, V _G is the gray total value of the whole image, R _A (T) is the probability that the pixel point is classified into class a, V _A (T) is the average gray level of the pixel point assigned to class a, R _B (T) is the probability that the pixel point is classified into class B, V _B (T) is the average gray level of the pixel point assigned to class B, R _i is the probability that the gray level of the pixel point is i, and g _i is the number of pixels in the image with gray level i.

9. The system of claim 6, wherein the coordinate matching module is specifically configured to: according to the sub-pixel coordinates of the resolution direction of the image layer space of the gray level vector, matching the sequence number of the vertical direction of the resolution direction of the current gray level vector to obtain the center coordinates of the optical slice corresponding to the gray level vector, wherein the formula is as follows:

P_k(R_k,C_k)＝(P_k,k)

Wherein, P _k(R_k,C_k) is the optical slice center coordinate corresponding to the kth gray-scale vector, R _k is the optical slice center decomposition direction coordinate of the "image layer space", C _k is the vertical direction coordinate of the optical slice center decomposition direction of the "image layer space", P _k is the decomposition direction subpixel coordinate of the "image layer space" of each of the kth gray-scale vectors, and k is the number of the vertical direction of the decomposition direction of the current gray-scale vector.

10. The system of claim 6, wherein the image light slice center module is specifically configured to: traversing the center coordinates of the optical slice corresponding to all the gray vectors to obtain the center of the optical slice of the optical stripe pixel level image, wherein the formula is as follows:

{m}＝{P_k(R_k,C_k)|k＝1,2,...,H}

Wherein { m } is the center of the optical slice of the optical stripe pixel level image, and H is the number of gray vectors.