An image preprocessing method for laser vector ranging

Abstract

The application discloses an image preprocessing method for laser vector ranging, which comprises the following steps: segmenting and extracting an effective image captured by a camera; converting the effective image into a gray image; and calculating the edge of the image after the generation of the gray image is completed. The application mainly utilizes the unique requirements of a sensor and prior information of hardware components to provide a reliable image preprocessing method, and provides reliable input for a subsequent ellipse recognition algorithm.

Description

Technical fields:

[0001] This invention relates to the field of non-contact measurement, and presents a principle-based measurement algorithm for a high-precision sensor. Based on a special laser (ring) light source and an image moment feature recognition method, vector measurement is performed on the measured plane, that is, simultaneously obtaining the normal (direction) information and distance information of the measured plane. Specifically, it relates to the field of non-contact measurement, providing an image preprocessing method for vector laser ranging sensors. The basic principle of this sensor is: designing a special ring laser emission light source to concentrate the laser beam energy and controllable its shape (ring), then capturing the projection of the laser beam onto the surface of the measured object, and subsequently performing image processing and moment feature extraction on the captured image. The extracted shape information is used to calculate the vector information of the measured plane, which includes the normal (direction) information and distance information of the measured plane. Before image feature extraction, a multi-step preprocessing operation is performed on the image to obtain the edge information of the image, which serves as the input image for ellipse feature recognition. Background technology:

[0002] Existing methods / sensors for measuring normal angles are mainly divided into two types: contact and non-contact. Contact angle measurement requires high-precision, high-reliability tooling and fixtures, along with angle measuring equipment, resulting in limited accuracy and complex applications. Non-contact angle measurement primarily uses gyroscopes, which are inertial elements. However, inertial sensors generally suffer from slow response, low-frequency zero-point drift, and high difficulty in software compensation, limiting their application to moving parts and making them less useful in manufacturing. Furthermore, current sensors used for environmental sensing obtain either distance or normal angle, but not both simultaneously (referred to as vector ranging in this paper). If the application requires simultaneous acquisition of spatial information of the measured object, i.e., vector ranging data, multiple sensors or combinations of multiple similar sensors must be used, i.e., sensor fusion. These methods are generally only applicable to specific applications and lack versatility; their design and manufacturing processes are complex, essentially equivalent to developing a dedicated device.

[0003] Existing non-contact measurement methods, including Time-of-Flight (TOF) and phase-based ranging, all rely on the propagation characteristics of light; angle measurement primarily uses inertial sensors. These methods are relatively mature and reliable. In recent years, with the improvement of computer hardware computing power, computer vision measurement methods have gradually been commercialized.

[0004] Currently, sensors used for environmental sensing only acquire distance or normal angle, not both simultaneously. In some applications, obtaining both distance and normal angle information of the measured object simultaneously requires custom design, including multi-sensor configurations, complex specialized tooling, mechanical mechanisms, and relatively complex sensor fusion algorithms. Custom designs are only for specific applications and are difficult to make universal; the design and manufacturing process is complex, equivalent to developing a dedicated piece of equipment. While computer vision measurement methods can acquire perceptual information from all angles, the need for processing large amounts of redundant data and complex image processing algorithms results in high computational requirements, complex configurations, and high costs, making it difficult to achieve general-purpose, mass-producible, large-scale market applications. Summary of the Invention:

[0005] To address the shortcomings of existing technologies, this invention provides an image preprocessing method for laser vector ranging. This invention offers a convenient method for acquiring vector information of the object under test based on laser and image acquisition. The image from the photosensitive camera is processed to obtain the data required by the sensor. A ring laser beam illuminates the surface of the object under test, forming an ellipse. Through image processing, the geometric information of the ellipse (center, direction, characteristic axes, etc.) is acquired. This geometric information is used to output the final sensor measurement data. In the fields of computer vision and image processing, there are many methods for shape recognition; however, due to the diversity of nature and the influence of different natural environments on photosensitive components, absolutely reliable and robust algorithms are almost nonexistent.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] An image preprocessing method for laser vector ranging includes the following steps: segmenting and extracting the effective image captured by the camera; converting the effective image into a grayscale image; and calculating the image edges after the grayscale image is generated.

[0008] As a further aspect of the present invention, the step of segmenting and extracting the effective image captured by the camera includes: calculating the maximum center offset, where the side is located at the furthest point of the sensor's ranging range, d. max =d0+Δd max At that time, the pixel distance between the center of the ellipse and the center point of the camera image is the maximum center offset N. p，max d0 is the distance between the calibration surface and the laser; the maximum semi-major axis of the ellipse is calculated based on the effective image captured by the camera, where the maximum value of the semi-major axis is: The calculation of the segmentation range for a square image uses the circumscribed square region of the effective circle as the effective region. Therefore, the image segmentation range is defined as follows: the segmentation range is centered at the image center, with a side length equal to the diameter of the effective circle. A square, where: Ceil is the positive rounding calculation.

[0009] As a further aspect of the present invention, the step of converting the effective image into a grayscale image includes: using the sum of the RGB components of the color image to characterize the energy of the pixel, i.e., Pow. ij =r ij +g ij +b ij The input color image is represented as an n×n matrix, where the RGB components corresponding to pixel (i, j) are (r... ij g ij b ij Feature color extraction based on residuals defines the feature component residuals as follows:

[0010] δ=|r ij +k g r ij +k b r ij -(r ij +g ij +b ij )|=|(k g +k b )r ij -(g ij +b ij The grayscale pixel value of this pixel is: in These are the color component coefficients;

[0011] Then the corresponding color image (r) ij g ij b ij The grayscale pixel value is defined as:

[0012]

[0013] As a further aspect of the present invention, the calculation of grayscale images includes: brightness calibration and feature color extraction.

[0014] As a further aspect of the present invention, the step of calculating the image edge after generating the grayscale image includes: after generating the grayscale image, performing image filtering; using the Sobel algorithm to calculate the gradient and perform interpolation calculation to obtain the gradient magnitude of pixels in the same direction.

[0015] As a further aspect of the present invention, an improved circular mean filter is used to denoise and filter the grayscale image. For a pixel position (i0, j0), its pixel value is... Its output after circular mean filtering is: in N R Let R be the number of pixels within a circle centered at (i0, j0).

[0016] As a further aspect of the present invention, the Sobel operator correlation matrix template is defined as:

[0017]

[0018] First, calculate the gradient values ​​in the x and y directions:

[0019] The adjacent matrix is ​​defined as:

[0020] Calculate the magnitude and direction angle of the gradient:

[0021] As a further aspect of the present invention, the step of performing interpolation calculation to obtain the gradient magnitude of pixels in the same direction includes: for the gradient magnitude matrix IA grad Point a on ij The corresponding direction angle Θ ij Calculate the points passing through point a. ij The direction angle is Θ ij The straight line, and a ij The gradient magnitudes of the two intersection points of the sides of the square formed by adjacent pixels are obtained by interpolation: a v1 and a v2 ;

[0022] The gradient magnitude segmentation matrix centered on the current pixel is:

[0023]

[0024] For the direction angle Θ ij a v1 and a v2 The calculation is as follows:

[0025] When 0≤Θ ij When <45°:

[0026]

[0027] When 45°≤Θ ij When <90°:

[0028]

[0029] When 90°≤Θ ij When <135°:

[0030]

[0031] When 135°≤Θ ij When <180°:

[0032]

[0033] If a ij At the same time greater than a v1 and a v2 , then I edge The pixel value e at coordinate (i, j) in the image ij =1, otherwise it is e ij =0.

[0034] The present invention has the following beneficial effects:

[0035] This invention mainly utilizes the unique requirements of sensors and the prior information of hardware components to provide an original and reliable image preprocessing method, providing reliable input for subsequent ellipse recognition algorithms.

[0036] To more clearly illustrate the structural features and effects of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. Attached image description:

[0037] Figure 1 and Figure 2 This is a schematic diagram of the segmentation method mentioned in this invention.

[0038] Figure 3 This is a schematic diagram of laser energy sampling in this invention.

[0039] Figure 4 This is a schematic diagram of the process for calculating grayscale images in this invention.

[0040] Figure 5 and Figure 6 This is a schematic diagram illustrating the principle of tilt angle measurement in this invention.

[0041] Figure 7 This is a schematic diagram showing the relationship between the incident angle at different angles and the imaging of the photosensitive camera in this invention.

[0042] Figure 8 These are schematic diagrams of effective images captured by the camera in this invention.

[0043] Figure 9 , Figure 10 These are schematic diagrams showing the resolution of a single-sensor camera and a dual-sensor camera used in this invention.

[0044] Figure 11 and Figure 12 This is a schematic diagram of the distance measurement principle in this invention. Detailed implementation method:

[0045] The present invention will now be further described in conjunction with the accompanying drawings and relevant knowledge, and will be described clearly and completely. Obviously, the described applications are only some embodiments of the present invention, and not all embodiments.

[0046] Reference in this invention Figures 5-12 As shown, a tilt measurement method based on a ring laser is proposed. Based on the characteristics of the laser source and the moment feature information of the captured image, the vector information from the laser source to the measured plane is calculated. The vector information includes the distance from the laser source to the plane and the normal angle between the laser source and the measured plane.

[0047] The system includes a laser source and two cameras. The laser is positioned at the center of the device, emitting a ring-shaped light spot. The cameras are positioned on either side of the laser, on the same plane as it. The ring-shaped light spot emitted by the laser strikes the surface of the object being measured. The cameras illuminate the light spot at different angles, resulting in different spot shapes. When the angle of incidence between the object's surface and the laser spot varies, the spot exhibits different elliptical shapes. For a fixed-position laser and cameras, different spot shapes correspond to different normal angles to the object's surface.

[0048] The laser emits a ring of light, assuming its radius is r0, the angle between the laser's incident angle and the normal to the object being measured is α, and the angle between the photosensitive camera and the laser's normal is β, where β is predetermined at the factory. Then:

[0049] The length of the major axis (variable axis) of the ellipse formed by the ring light on the surface of the object being measured is:

[0050] L s =2r0 / cosα

[0051] The length of this ellipse along the major axis (variable axis) of the image taken by the right camera is:

[0052]

[0053] Similarly, the length of the major axis (variable axis) of the ellipse in the image formed by the left camera is:

[0054]

[0055] L in this invention lens，1 and L lens，2 The values ​​are obtained through calculations using a photosensitive camera.

[0056] If a single camera is used to calculate the normal angle (e.g., the right camera), then the angle α between the object being measured and the laser direction is:

[0057]

[0058] Further optimization, if two cameras are used simultaneously in the calculation, then:

[0059]

[0060] in, In this invention, regardless of whether a single camera or a dual camera is used, the ellipse parameter k used in this invention mainly refers to the ratio of the major axis to the minor axis of the ellipse.

[0061] In this invention, the recognition results of a single camera or two cameras can be used. The difference between the two lies in the resolution (accuracy). The system resolution (accuracy) analysis is performed using single-camera and dual-camera calculations under different camera-laser angles β.

[0062] Taking β = 45° as an example, in the case of a single camera, when α changes from 0° to 10°, the measurement input... The change rate is 17.68% when the value changes from 0.707 to 0.832; however, for the dual-camera operation under the same conditions, the change rate of the measurement input k from 1 to 0.5 is 50%, which is much higher than 17.68%, resulting in an improvement of nearly 3 times in system sensitivity.

[0063] In this invention, a schematic diagram of the distance measurement between the laser and the object being measured is shown. This invention uses the right camera as an example for calibration and calculation. The calculation process for the left camera is the same, specifically including...

[0064] (1) Image center calibration

[0065] Before the sensor leaves the factory, distance calibration is performed and a calibration surface is set. When the center of the ring light is in the center of the camera image, the distance between the laser exit and the calibration surface is recorded as d0.

[0066] (2) Distance calculation

[0067] Example image from the right camera Figure 8 As shown, when a ring laser illuminates the side of the object, the number of pixels from the center of the ellipse to the center of the image is N. p Therefore, according to geometric relations, we can obtain:

[0068] The distance from the laser output to the measured plane is:

[0069]

[0070] Where C is a parameter related to the camera's light sensitivity performance, which can be obtained through experimental calibration or through camera configuration.

[0071] Reference Figures 1-12As described above, based on the sensor principle of this invention, due to the concentrated energy of laser light, only a portion of the pixels within the effective ranging range of a photosensitive camera image contain measurement information. This invention utilizes the sensor's measurement range (angle and distance) and known environmental factors to perform effective image segmentation and extraction.

[0072] Reference Figures 1-2 As shown, the laser emits a ring of light, assuming its ring radius is r0. The positions of the laser and the photosensitive camera are fixed, and the angle between the normals of the photosensitive camera and the laser is β.

[0073] An image preprocessing method for laser vector ranging includes the following steps: segmenting and extracting the effective image captured by the camera; converting the effective image into a grayscale image; and calculating the image edges after the grayscale image is generated.

[0074] Further optimization includes the effective image segmentation and extraction captured by the camera, which includes:

[0075] Maximum center offset calculation: When the distance between the calibration surface and the laser is d0, the center of the ellipse captured by the camera is located at the center of the image pixel, meaning the normal vector of the camera's imaging plane passes through the center of the laser spot on the calibration surface. When the distance is d0, the maximum offset is calculated by moving the ellipse to the farthest point of the sensor's ranging range. max =d0+Δd max At that time, the pixel distance (number of pixels) between the center of the ellipse and the center point of the camera image is the maximum center offset N. p，max ;

[0076] Calculation of the major axis of the ellipse: The sensor's angular range is α∈[0~α max ], where α is the angle between the incident angle of the laser and the normal of the object being measured.

[0077] Reference Figure 8 As shown, the maximum value of the semi-major axis of the ellipse is:

[0078] Calculation of the segmentation range of a square image: The ranging range of the sensor d∈[d0-Δd] max ,d0+Δ dmax ,], Angle measurement range α∈[0, α max If the effective image (ellipse) captured by the camera lies within a circle centered at the image center with a radius of , then the effective image (ellipse) captured by the camera lies within a circle centered at the image center with a radius of . The area within the circle is called the effective circle. Where: Ceil is a positive rounding calculation, and the result is the smallest integer value not less than the calculated quantity. C is a parameter related to the camera's light sensitivity performance, which can be obtained through experimental calibration or through camera configuration.

[0079] Note: R p，eff The unit is pixel distance, which is the number of pixels.

[0080] From an image processing perspective, images are typically calculated and processed in matrix form. To facilitate subsequent image processing, the circumscribed square region of the effective circle is used as the effective area of ​​this sensor invention. Therefore, the image segmentation range is defined as: centered on the image center, with a side length equal to the diameter of the effective circle, i.e. A square.

[0081] In a preferred embodiment of the present invention, the process of converting a valid image into a grayscale image includes, after obtaining the valid image, converting the color image into a grayscale image to facilitate subsequent image filtering and recognition. The present invention utilizes the monochromaticity of laser (single spectrum, narrow band range) and the concentrated and controllable energy, enabling laser image calibration under controllable conditions (light environment, controlled environment). Therefore, compared to conventional image processing methods, the present invention utilizes prior information about the laser source (laser energy and spectral characteristics) to generate a grayscale image. This further "suppresses" useless pixels (mainly background and noise points), thereby simplifying subsequent image filtering, edge extraction, and other tasks, ensuring sensor accuracy.

[0082] The input color image is represented as an n×n matrix, where the RGB components corresponding to pixel (i, j) are (r... ij g ij b ij The processed grayscale image has the same size as the input color image, and the grayscale value of each pixel is p. ij .

[0083] Specifically, it includes:

[0084] Energy filtering, in the sensor involved in this invention, involves concentrated laser energy that must be guaranteed to have a certain energy level in the image. Therefore, the energy of the laser-irradiated pixel will not fall below a certain threshold. This threshold value is related to the emitted laser energy and the propagation medium. This invention uses the sum of the RGB components of the color image to characterize the energy of the pixel, i.e., Pow. ij =r ij +g ij +b ij .

[0085] The experimental method determines the energy threshold of the laser. The laser beam is placed against a black background in a dark environment, and the plane to be measured is positioned at its maximum distance range (e.g., ...). Figure 1 The system takes pictures and samples the pixels at the laser irradiation point to obtain the averaged energy data. Figure 3 This is an actual captured image of a ring laser. The average laser energy at the maximum distance range is Pow. thmin Since other light sources exist in actual working environments, the lower limit Pow of the above experiments is used in this invention. th，minThis refers to the laser energy threshold acquired at the maximum distance range. For the generation of grayscale images, if the energy of a pixel is lower than Pow... thmin If so, then the value of that pixel in the grayscale image will be assigned to 0.

[0086] It also includes feature color extraction based on residuals, utilizing the spectral characteristics of lasers to further enhance the effective pixels of the ring laser.

[0087] First, in a laboratory environment, the laser aperture is adjusted to converge at a single point as much as possible. Assuming the laser shines against a black background, the RGB energy components of the generated laser image are normalized and then expressed as r... ref g ref b ref With infrared laser (r ref Taking a relatively large color component as an example, the color component coefficients are defined as follows:

[0088] In sensor applications, the camera captures the pixels p of an image. i，j Its RGB component energy is r ij g ij b ij Based on the color component coefficients, the characteristic component residuals are defined as follows:

[0089] δ=|r ij +k g r ij +k b r ij -(r ij +g ij +b ij )|=|(k g +k b )r ij -(g ij +b ij )|

[0090] The grayscale pixel value of this pixel is:

[0091] Corresponding color image (r) ij g ij b ij The grayscale pixel value is defined as:

[0092]

[0093] To effectively ensure the stability and robustness of the recognition, the color component coefficients were calibrated offline under different lighting (white light) conditions and against a black background.

[0094] Under controlled lighting conditions in the laboratory, different lighting conditions Γi For i∈[0,l], obtain the color component coefficients. and

[0095] In this invention, the sensor is equipped with an ambient light sensing device, which can sense the brightness Γ of the ambient light in real time, i.e., the illumination condition Γ. i This invention classifies ambient light intensity into four categories (l=4), as shown in Table 1:

[0096] Table 1

[0097]

[0098] In Table 1, the data in columns 2 and 3 are provided by laboratory testing conditions. When calculating grayscale images, the following methods are used: Figure 4 As shown.

[0099] If environmental sensing devices cannot be used due to limitations in operating conditions, the data in columns 2 and 3 of the table above can still be obtained under controlled laboratory conditions. Based on the sensor's operating environment, the general component coefficients are obtained using a weighted averaging method as follows:

[0100]

[0101]

[0102] Where ∑ i W i =1. Regarding the weight W i It is determined based on prior information about the sensor's operating environment.

[0103] This invention also includes image edge calculation, specifically: the invention uses the same process as the Canny algorithm for edge calculation. Generally, the Canny algorithm includes steps such as Gaussian filtering, gradient calculation, non-maximum suppression, and double gate processing.

[0104] Considering the characteristics of the sensors involved, this invention does not employ a Gaussian filtering method, but instead uses an improved circular mean filter. The subsequent gradient calculation steps and non-maximum suppression algorithm steps are the same as those in the Canny algorithm. Since the ellipse recognition algorithm mentioned in the next section requires further filtering of the obtained image edges, and this filtering process largely overlaps with the dual-gate value processing process in the existing Canny algorithm, this invention does not perform dual-gate value processing.

[0105] Image filtering is necessary after grayscale image generation. There are many types of image filtering methods, such as mean filtering, median filtering, Gaussian filtering, circular filtering, and Laplace operator filtering, selected according to different application requirements. In the sensor application involved in this invention, the purpose of filtering is to suppress noisy pixels and background pixels, enhance specific shapes or textures, and ensure the reliability of subsequent edge extraction and shape recognition algorithms.

[0106] To highlight the object to be identified in the sensor image, namely the smooth and continuous characteristics of the ellipse, this invention employs an improved circular mean filter to denoise and filter the grayscale image generated above. The advantages of this filter are mainly twofold: (1) it "repairs" the dots on the image, making it easier to identify the center of the laser beam. For example... Figure 7 (2) To repair the noise in the imaging of a continuous smooth pattern (especially the pattern boundary), so that its edge, i.e. the derivative or gradient of the image signal, is closer to the real boundary, thus ensuring the reliability of the subsequent edge extraction and ellipse recognition algorithm.

[0107] For a pixel at position (i0, j0), its pixel value is Its output after circular mean filtering is:

[0108] in N R Let R be the number of pixels within a circle centered at (i0, j0). In this application, the value of R is related to the energy distribution of the ring laser. To ensure reliable identification of the center, we take... C W The diameter is the center point diameter of the laser beam.

[0109] Note 1: For points located at the edge of the image, their edge pixels can be padded with zeros.

[0110] Note 2: R in the above formula is in pixels.

[0111] Gradient calculation, specifically including: the filtered n×n grayscale image is: I input The value of each pixel is p ij .

[0112] The gradient image magnitude is defined as IA grad The size is n×n, and the value of each pixel is a. ij .

[0113] The gradient image orientation angle is defined as ID. grad The size is n×n, and the value of each pixel is Θ. ij .

[0114] This invention uses the Sobel algorithm to calculate its gradient.

[0115] The Sobel operator correlation matrix template is defined as:

[0116]

[0117] First, calculate the gradient values ​​in the x and y directions (horizontal and vertical directions):

[0118]

[0119] The adjacent matrix is ​​defined as:

[0120] The second step is to calculate the magnitude and direction angle of the gradient:

[0121]

[0122] It also includes nonmaximum suppression calculation, and outputs the edge calculation result I. edge A binary image of size n×n, where the value of each pixel is e. ij .

[0123] For input image I input For each pixel in the image, if the magnitude of its gradient is greater than the magnitudes of the gradients of its two neighboring pixels in the same direction, it indicates that the gradient is a maximum value in this direction (direction angle), and the probability that it is located on the edge of the image is extremely high. Therefore, this point is considered to be an edge point e. ij =1. Conversely, it indicates that it may be due to noise points, which can be suppressed, i.e., e i Set to 0. This is because the gradient direction angle Θ... ij It is not necessarily an integer multiple of 45°, so interpolation calculation is required to obtain the gradient magnitude of pixels in the same direction.

[0124] For the gradient magnitude matrix IA grad Point a on ij The corresponding direction angle Θ ij Calculate the points passing through point a. ij The direction angle is Θ i The straight line, and a ij The gradient magnitudes of the two intersection points of the sides of the square formed by adjacent pixels are obtained by interpolation: a v1 and a v2 .

[0125] The gradient magnitude segmentation matrix centered on the current pixel is:

[0126]

[0127] For the direction angle Θ ija v1 and a v2 The calculation is as follows:

[0128] When 0≤Θ ij When <45°:

[0129]

[0130] When 45°≤Θ ij When <90°:

[0131]

[0132] When 90°≤Θ ij When <135°:

[0133]

[0134] When 135°≤Θ ij When <180°:

[0135]

[0136] If a ij At the same time greater than a v1 and a v2 , then I edge The pixel value e at coordinate (i, j) in the image ij =1, otherwise it is e ij =0.

[0137] The calculated I above edge Ellipse recognition is performed using an image as input. The ellipse recognition method employs an arc segment-based recognition approach.

[0138] This invention provides an image preprocessing method based on laser light sources and image features. A ring-shaped laser beam illuminates the surface of the object being measured, forming an ellipse. Image processing methods are used to acquire the geometric information of the ellipse (center, direction, characteristic axes, etc.), which is then used to output sensor measurement data. This invention primarily utilizes the unique requirements of the sensor and prior information from the hardware components to provide an original and reliable image preprocessing method, offering reliable input for subsequent ellipse recognition algorithms.

[0139] The technical principles of the present invention have been described above with reference to specific embodiments, which are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments; all technical solutions falling within the scope of the present invention's concept are within its protection scope. Those skilled in the art can conceive of other specific embodiments of the present invention without creative effort, and these embodiments will all fall within the protection scope of the present invention.

Claims

1. An image preprocessing method for laser vector ranging, characterized in that, The steps include: segmenting and extracting the valid image captured by the camera; converting the valid image into a grayscale image; After generating the grayscale image, the image edges are calculated. The steps for segmenting and extracting the effective image captured by the camera include: calculating the maximum center offset, and determining the edge of the image by which the edge is located at the furthest point in the sensor's ranging range. At that time, the pixel distance between the center of the ellipse and the center point of the camera image is the maximum center offset. , The distance between the calibration surface and the laser; the maximum semi-major axis of the ellipse is calculated based on the effective image captured by the camera, where the maximum value of the semi-major axis is... The calculation of the segmentation range for a square image uses the circumscribed square region of the effective circle as the effective region. Therefore, the image segmentation range is defined as follows: the area centered on the image center, with a side length equal to the diameter of the effective circle. The square, where ceil is the positive rounding calculation, r0 is the radius of the ring laser, β is the angle between the normal of the camera and the laser, and Δd max For the maximum ranging deviation, ω is the maximum value of the angle between the laser incident angle and the normal of the object being measured; c is the camera's photosensitive performance parameter.

2. The image preprocessing method for laser vector ranging as described in claim 1, characterized in that, The step of converting a valid image to a grayscale image includes: using the sum of the RGB components of the color image to characterize the energy of a pixel, i.e. The input color image is represented as The matrix has RGB components corresponding to pixel (i, j) as follows: , , Feature color extraction based on residuals: The feature component residuals are defined according to the color component coefficients as follows: The grayscale pixel value of the pixel is: ; Then the corresponding color image ( , , The pixel value of a grayscale image is defined as follows: , where r ref b ref These are the normalized RGB component values ​​of a black background under laser irradiation in a laboratory environment. Pow is the color component coefficient. th,min g is the laser energy threshold at the maximum distance range. ref It is the normalized value of the green component of the reference pixel under laser illumination.

3. The image preprocessing method for laser vector ranging as described in claim 2, characterized in that, The calculation of grayscale images includes: brightness calibration and feature color extraction.

4. The image preprocessing method for laser vector ranging as described in claim 3, characterized in that, The steps for calculating image edges after grayscale image generation include: image filtering after grayscale image generation; gradient calculation and interpolation calculation using the Sobel algorithm to obtain the gradient magnitude of pixels in the same direction.

5. The image preprocessing method for laser vector ranging as described in claim 4, characterized in that, An improved circular mean filter is used to denoise and filter grayscale images for a specific pixel location. Its pixel value The output after circular mean filtering is: ,in , For Let R be the center of the circle, and R be the number of pixels within the radius.

6. The image preprocessing method for laser vector ranging as described in claim 5, characterized in that, The Sobel operator correlation matrix template is defined as: ， ； First, calculate the gradient values ​​in the x and y directions: ; The adjacent matrix is ​​defined as: ; Calculate the magnitude and direction angle of the gradient: ; .

7. The image preprocessing method for laser vector ranging as described in claim 6, characterized in that, The steps for interpolating and obtaining the gradient magnitude of pixels in the same direction include: for the gradient magnitude matrix Points on The corresponding direction angle ; Calculate the points passed through The direction angle is The straight line, and The gradient magnitudes of the two intersection points of the sides of the square formed by adjacent pixels are obtained by interpolation: ; The gradient magnitude segmentation matrix centered on the current pixel is: ； For direction angle , and The calculation is as follows: when hour: ； when hour: ； when hour: ； when hour: ； if At the same time greater than and ,So Pixel value at coordinates (i, j) in the image Conversely .

Images