A method and system for feature searching of on-vehicle display screen alignment and bonding

Abstract

The application discloses a kind of vehicle-mounted display screen alignment and fit feature search method and system, belong to display screen processing technical field, comprising: the original gray image of the object to be searched for feature is obtained;Divide n ROI regions on original gray image;Image Rcn after convolution operation is obtained;Sequential traversal is carried out to image Rcn, to obtain point P (x0, y0), the integrated edge intensity g of point P is calculated;If integrated edge intensity g≈0, edge curve is fitted according to point P;If integrated edge intensity g≠0, subpixel fitting calculation is executed to obtain the coordinates (xf, yf) of point Pf, edge curve is fitted according to final subpixel edge point Pf;Edge curve is converted to physical space coordinate system, and physical motion amount is calculated.In prior art, it is generally relied on to improve camera resolution to improve precision, which can significantly improve cost and processing speed.Compared with prior art, the application is searched for feature and positioned by subpixel level algorithm, which effectively improves efficiency and reduces cost.

Description

Technical Field

[0001] This invention relates to the field of display screen processing technology, and in particular to a method and system for finding alignment and bonding features of vehicle-mounted displays. Background Technology

[0002] With the deepening of the automotive intelligence trend, in-vehicle displays have evolved from simple information display terminals into core interactive interfaces integrating navigation, entertainment, driver assistance, and vehicle control. The richer the functions, the larger the screen size, and the more diverse the form factor, the more critical their impact on the overall vehicle quality and reliability becomes. In the bonding and assembly process, the precise alignment and bonding of the display module to the back cover directly determines not only the flatness of the displayed image and the consistency of the bezels, but also the structural strength and airtightness of the entire unit under harsh automotive-grade environments such as vibration and temperature changes. Even slight deviations in this process can lead to optical distortion or touch malfunction. Therefore, major manufacturers have listed bonding accuracy as a core control indicator for their production lines, continuously tightening tolerance standards to strive for a perfect balance between visual experience and durability.

[0003] However, the traditional path to improving accuracy is encountering real bottlenecks. The industry generally relies on increasing the optical resolution of cameras to capture finer alignment marks, thereby driving robotic arms to perform micron-level calibration. But higher-resolution industrial cameras mean a surge in image data, not only driving up hardware procurement and maintenance costs but also significantly lengthening image processing and computation cycles, slowing down the entire production line. Under the pressure of mass production, this simple "resolution-only" upgrade is clearly unsustainable—the sharp increase in costs squeezes profit margins, while the decrease in speed directly weakens production capacity advantages.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] This invention provides a method for finding alignment and fitting features of an in-vehicle display screen, comprising: acquiring an original grayscale image of the object to be featured; dividing the original grayscale image into n Regions of Interest (ROIs); performing a convolution operation on each ROI to obtain an image Rcn after the convolution operation; sequentially traversing the image Rcn to obtain the first point P(x0, y0) equal to a preset edge intensity Str, and calculating the comprehensive edge intensity g of point P; if the comprehensive edge intensity g≈0, fitting an edge curve based on point P; if the comprehensive edge intensity g≠0, performing subpixel fitting calculation to obtain the coordinates (xf, yf) of the final subpixel edge point Pf, fitting an edge curve based on the final subpixel edge point Pf; converting the edge curve to a physical space coordinate system and calculating the physical motion.

[0006] Furthermore, it also includes the following steps: Based on the features of the object to be identified and the preset parameters, divide the original grayscale image into n rectangular regions; Rotate the rectangular region so that its height is parallel to the y-axis and its width is parallel to the x-axis to obtain n ROI regions corresponding to the original grayscale image.

[0007] Furthermore, it also includes the following steps: For each ROI region, a Sobel convolution kernel of a preset size in the y-direction is used to perform a convolution operation to obtain the edge intensity of each ROI region in the y-direction; Let the region of origin (ROI) after convolution be denoted as image Rcn.

[0008] Furthermore, it also includes the following steps: For each image Rcn, traverse it sequentially first in the x direction and then in the y direction to obtain the first point P(x0,y0) that is equal to the preset edge intensity Str; For the image Rcn corresponding to point P(x0,y0), perform a convolution operation in the x direction to obtain the edge intensity map Cx of image Rcn in the x direction; For the image Rcn corresponding to point P(x0,y0), perform a convolution operation in the y direction to obtain the edge intensity map Cy of image Rcn in the y direction; Obtain the value gx of the edge intensity map Cx at coordinates (x0, y0); Obtain the value gy of the edge intensity map Cy at coordinates (x0, y0); Calculate the overall edge strength g of point P(x0,y0) based on gx and gy: .

[0009] Furthermore, if the overall edge strength g ≠ 0, the overall edge strength g is normalized to obtain the normalized edge strength. Take the previous pixel Pprev(x0,y0-1) and the next pixel Pnext(x0,y0+1) of point p in the y direction; Fit the data to point p, pixel Pprev, and pixel Pnext to obtain the formula f(v); Solve the equation To obtain the final root solution; Based on the normalized edge intensity and the final root solution, calculate the coordinates (xf, yf) of the final sub-pixel edge point Pf.

[0010] Furthermore, when point p, pixel Pprev, and pixel Pnext exhibit a non-linear relationship, the Y-coordinates of point p, pixel Pprev, and pixel Pnext are used as the independent variable u, and the pixel value is used as the dependent variable v, to apply the parabolic equation: f(v) = au 2 The formula f(v) is obtained by fitting the data using +bu+c. When point p, pixel Pprev, and pixel Pnext have a linear relationship, the difference in the Y coordinates of point p, pixel Pprev, and pixel Pnext relative to point p is used as the independent variable u, and the pixel value is used as the dependent variable v. The formula f(v) is obtained by fitting the linear formula.

[0011] Furthermore, the formula for calculating the coordinates (xf, yf) of the final sub-pixel edge point Pf is as follows:

[0012] The normalized edge strength is in the x-direction; denoted as the normalized edge strength in the y-direction.

[0013] The present invention also provides a system for finding the alignment and bonding features of an in-vehicle display screen, using the method described above.

[0014] Compared with the prior art, the present invention has the following advantages: By finding point P and calculating the final sub-pixel edge point Pf, the coordinates of the feature point are accurate to the decimal level. Thus, without increasing the camera's physical resolution, the positioning accuracy breaks through the pixel limit, thereby significantly improving the alignment capability of the fitting.

[0015] This invention compensates for insufficient hardware resolution through the aforementioned process, thus eliminating the need to increase camera resolution to improve accuracy. Even using a conventional resolution camera, sub-micron level feature coordinates can be acquired, thereby significantly reducing the procurement and maintenance costs of high-resolution industrial camera-grade data cables while ensuring accuracy. Simultaneously, the image data volume using a conventional resolution camera is relatively small, thus ensuring image processing and computation cycles and avoiding adverse effects on production line speed.

[0016] By comprehensively considering the judgment conditions of edge intensity g, the sub-pixel fitting calculation process can be actively bypassed when the corresponding conditions are met, and point P can be directly used. This avoids numerical divergence caused by sub-pixel interpolation in weak texture regions. Furthermore, it effectively reduces the overall computational load, saves computing power, and improves computational efficiency.

[0017] By determining the relationship between point p, pixel Pprev, and pixel Pnext, an accurate mathematical model can be obtained regardless of whether the edge is curved or straight, avoiding fitting failure or decreased accuracy caused by a single model. Attached Figure Description

[0018] Figure 1 Overall process diagram.

[0019] Figure 2 Image segmentation diagram. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0021] Example 1: This invention provides a method for finding alignment and bonding features of an in-vehicle display screen, referring to... Figure 1 , Figure 1 This is a schematic diagram of the process of the present invention. It includes: acquiring the original grayscale image of the object to be featured using a camera. To better capture the features, several light sources can be deployed around the object to help highlight the features.

[0022] Based on the features of the object to be featured and preset parameters, the original grayscale image is divided into n rectangular regions, each containing a portion of the features. These rectangular regions are denoted as r1, r2, ..., rn. See appendix for details. Figure 2 .

[0023] To facilitate processing, these rectangular regions are rotated so that their height is parallel to the y-axis and their width is parallel to the x-axis, in order to obtain n ROI regions corresponding to the original grayscale image, denoted as R1, R2, ..., Rn.

[0024] For each ROI region R1, R2, ..., Rn, perform convolution operations using a Sobel convolution kernel of a preset size in the y-direction to obtain the edge intensity of each ROI region in the y-direction. For example, a 3 The Sobel convolution kernel of size 3 in the y-direction is:

[0025] Let the ROI regions after convolution be denoted as Rc1, Rc2, ..., Rcn.

[0026] For each image Rc1, Rc2, ..., Rcn, perform a sequential traversal first in the x-direction and then in the y-direction to find the first point P(x0, y0) that equals the preset edge intensity Str. If the preset edge intensity Str is not an integer, the condition for finding point P is that it equals the rounded value of Str.

[0027] For the image Rcn corresponding to point P(x0,y0), use 3 A Sobel convolution kernel of size 3 is used to perform convolution operations in the x-direction to obtain the edge intensity map Cx of image Rcn in the x-direction; For the image Rcn corresponding to point P(x0,y0), use 3 A Sobel convolution kernel with a 3x ... Obtain the value gx of the edge intensity map Cx at coordinates (x0, y0); Obtain the value gy of the edge intensity map Cy at coordinates (x0, y0); Calculate the overall edge strength g of point P(x0,y0) based on gx and gy: .

[0028] If the overall edge intensity g ≠ 0, perform the following sub-pixel fitting calculation process: normalize the overall edge intensity g to obtain the normalized edge intensity. Let the normalized edge intensity in the x-direction be denoted as... The normalized edge strength in the y-direction is Specifically:

[0029] Take the previous pixel Pprev(x0,y0-1) and the next pixel Pnext(x0,y0+1) of point p in the y direction; When point p, pixel Pprev, and pixel Pnext exhibit a non-linear relationship, the y-coordinates of point p, pixel Pprev, and pixel Pnext are used as the independent variable u, and the pixel value is used as the dependent variable v, to apply the parabolic equation: f(v) = au 2 The formula f(v) is obtained by fitting the data using +bu+c. When point p, pixel Pprev, and pixel Pnext have a linear relationship, the difference in the Y coordinates of point p, pixel Pprev, and pixel Pnext relative to point p is used as the independent variable u, and the pixel value is used as the dependent variable v. The formula f(v) is obtained by fitting the linear formula.

[0030] In summary, by determining the relationship between point p, pixel Pprev, and pixel Pnext, an accurate mathematical model can be obtained regardless of whether the edge is curved or straight, avoiding fitting failure or decreased accuracy caused by a single model.

[0031] Solve the equation Several solutions are obtained. Based on the preset conditions, one of the solutions is selected as the final solution, denoted as root.

[0032] Based on the normalized edge intensity and the final root solution, the coordinates (xf, yf) of the final sub-pixel edge point Pf are calculated. The calculation formula is:

[0033] This concludes the subpixel fitting calculation process.

[0034] The coordinates (xf, yf) of the final sub-pixel edge point Pf corresponding to each rotated ROI are restored to the original image coordinates and denoted as Pn. The edge curves on the object image are fitted using the RANSAC random sample consensus algorithm for the N points P1, P2...Pn after rotation and restoration, while removing noise and maintaining the accuracy of the fitting.

[0035] If the overall edge strength g≈0, the aforementioned sub-pixel fitting calculation process will no longer be performed. Instead, point P will be placed into P1, P2...Pn to perform the edge curve fitting process.

[0036] Preferably, a threshold can be preset according to actual needs to determine the overall edge strength g. Specifically: when g ≤ threshold, g ≈ 0 is determined; when g > threshold, g ≠ 0 is determined.

[0037] In summary, by comprehensively considering the judgment conditions for edge intensity g, the sub-pixel fitting calculation process can be actively bypassed when the corresponding conditions are met, and point P can be directly used. This avoids numerical divergence caused by sub-pixel interpolation in weak texture regions. Furthermore, it effectively reduces the overall computational load, saves computing power, and improves computational efficiency.

[0038] Subsequently, the edge curves are transformed to the physical space coordinate system according to the preset camera-space coordinate system transformation relationship, and then the physical motion is calculated.

[0039] In summary, by finding point P and calculating the final sub-pixel edge point Pf, the coordinates of the feature point are accurate to the decimal level. Thus, without increasing the camera's physical resolution, the positioning accuracy breaks through the pixel limit, thereby significantly improving the alignment capability of the fitting.

[0040] On the other hand, this invention compensates for the lack of hardware resolution through the aforementioned process, thus eliminating the need to increase camera resolution to improve accuracy. Even using a conventional resolution camera, sub-micron level feature coordinates can be acquired, thereby significantly reducing the procurement and maintenance costs of high-resolution industrial camera-grade data cables while ensuring accuracy. Simultaneously, the image data volume using a conventional resolution camera is relatively small, thus ensuring image processing and computation cycles and avoiding adverse effects on production line speed.

[0041] Example 2: A system for finding alignment and bonding features of an in-vehicle display screen, using the method described in Embodiment 1.

[0042] Furthermore, it should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0043] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0044] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory (ROM) / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0045] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

[0046] It is understood that the system provided in the embodiments of the present invention corresponds to the method provided in the embodiments of the present invention, and the explanation, examples and beneficial effects of the relevant content can be referred to the corresponding parts of the above methods.

Claims

1. A method for finding alignment and bonding features of a vehicle-mounted display screen, characterized in that, include: Obtain the original grayscale image of the object whose features are to be searched; Divide the original grayscale image into n Regions of Interest (ROIs); Perform a convolution operation on each of the ROI regions to obtain the convolutional image Rcn; The image Rcn is sequentially traversed to obtain the first point P(x0,y0) that is equal to the preset edge intensity Str, and the comprehensive edge intensity g of point P is calculated. If the overall edge strength g≈0, fit the edge curve based on the point P; If the overall edge intensity g≠0, perform subpixel fitting calculation to obtain the coordinates (xf, yf) of the final subpixel edge point Pf, and fit the edge curve based on the final subpixel edge point Pf; The edge curve is transformed to a physical space coordinate system to calculate the physical motion.

2. The method for finding alignment and bonding features of a vehicle-mounted display screen as described in claim 1, characterized in that, It also includes the following steps: Based on the features of the object to be found and the preset parameters, n rectangular regions are divided on the original grayscale image; Rotate the rectangular region so that its height direction is parallel to the y-axis and its width direction is parallel to the x-axis, to obtain n ROI regions corresponding to the original grayscale image.

3. The method for finding alignment and bonding features of a vehicle-mounted display screen as described in claim 1, characterized in that, It also includes the following steps: For each ROI region, a convolution operation is performed using a Sobel convolution kernel of a preset size in the y-direction to obtain the edge intensity of each ROI region in the y-direction; The ROI region after the convolution operation is denoted as the image Rcn.

4. The method for finding alignment and bonding features of a vehicle-mounted display screen as described in claim 1, characterized in that, It also includes the following steps: For each image Rcn, perform a sequential traversal first in the x direction and then in the y direction to obtain the first point P(x0,y0) that is equal to the preset edge intensity Str; For the image Rcn corresponding to the point P(x0,y0), perform a convolution operation in the x direction to obtain the edge intensity map Cx of the image Rcn in the x direction; For the image Rcn corresponding to the point P(x0,y0), perform a convolution operation in the y direction to obtain the edge intensity map Cy of the image Rcn in the y direction; Obtain the value gx of the edge intensity map Cx at coordinates (x0, y0); Obtain the value gy of the edge intensity map Cy at coordinates (x0, y0); Calculate the combined edge strength g of point P(x0,y0) based on gx and gy: 。 5. The method for finding alignment and bonding features of a vehicle-mounted display screen as described in claim 1, characterized in that, If the overall edge strength g ≠ 0, the overall edge strength g is normalized to obtain the normalized edge strength; Take the previous pixel Pprev(x0,y0-1) and the next pixel Pnext(x0,y0+1) of the point p in the y direction; Fit the point p, the pixel Pprev, and the pixel Pnext to obtain the formula f(v); Solve the equation To obtain the final root solution; Based on the normalized edge intensity and the final root solution, calculate the coordinates (xf, yf) of the final sub-pixel edge point Pf.

6. The method for finding alignment and bonding features of a vehicle-mounted display screen as described in claim 5, characterized in that, When the relationship between point p, pixel Pprev, and pixel Pnext is non-linear, the Y-coordinates of point p, pixel Pprev, and pixel Pnext are used as the independent variable u, and the pixel value is used as the dependent variable v, to apply the parabolic equation: f(v) = au 2 The formula f(v) is obtained by fitting the data with +bu+c. When point p, pixel Pprev, and pixel Pnext are linearly related, the difference in Y coordinates of point p, pixel Pprev, and pixel Pnext relative to point p is used as the independent variable u, and the pixel value is used as the dependent variable v, to fit the formula f(v) using a linear formula.

7. The method for finding alignment and bonding features of a vehicle-mounted display screen as described in claim 1, characterized in that, The formula for calculating the coordinates (xf, yf) of the final sub-pixel edge point Pf is as follows: The The normalized edge strength is in the x-direction; The y is the normalized edge strength in the y direction.

8. A system for finding alignment and bonding features of a vehicle-mounted display screen, characterized in that, The method described in any one of claims 1 to 7 is applied.

Images