Laser line scanning tire side eccentricity imaging correction method

Abstract

The invention discloses a method for correcting eccentric imaging of the side of a laser line scan tire, comprising the following steps: S1 uses a laser scanner to obtain tire information to obtain a csv file, in which each point in the file is a depth value; S2 performs denoising preprocessing and depth consistency preprocessing on the csv file, and converts the csv file into an image of real data; S3 creates a tire edge sawtooth template; S4 locates a fitting smooth curve through the tire sawtooth template, and performs eccentricity correction processing on the image; S5 tire image enhancement processing; S6 creates a tire hair template of the tire image; S7 locates the fetal hair in the image and performs a deburring operation through a template matching method; S8 adopts a line-background consistency processing method to process text areas in sections. The invention can effectively improve the contrast of the laser line scan image of the side of the tire, solves the problem of bending of the imaging picture caused by eccentricity when the tire rotates, and provides high-quality preprocessing results for character recognition on the tire surface.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a laser line scanning image processing method. Background Art

[0002] The characters on the tire carry product information and are very important to tire manufacturers [1]. However, accurate recognition of characters on the tire surface is a major problem. The judgment criteria of manual inspection methods vary from person to person, and the inspection results of different individuals may be different. Moreover, long-term work can easily cause visual fatigue of personnel, and the inspection efficiency and reliability will be greatly reduced [2]. In recent years, with the development of artificial intelligence, machine vision has been increasingly used in industrial inspection. Laser line scanning tire surface character recognition, as a type of machine vision application, has been widely used in industrial inspection in various industries. However, due to the noise and scanning range limitation of laser line scanning cameras during the scanning process, the imaging images are uneven; the rotation of the operating table causes the tire imaging images to bend and eccentric; the middle drum of the tire side is low on both sides, resulting in the phenomenon that the imaging image is bright in the middle and dark on both sides; the rotation of the operating table is not on the same plane, resulting in the overall bright and dark alternation of the image, etc., which seriously interferes with the accurate recognition of characters and greatly reduces the accuracy of machine vision in industrial inspection. Therefore, it is particularly important to invent an accurate laser line scanning tire side eccentric imaging correction processing method.

[0003] [1] Chen Yuchao. A tire mold character detection method based on machine vision[J]. Guangdong University of Technology, 2016.

[0004] [2] Ruan Yujing. A tire tread detection system based on machine vision[D]. Hangzhou Dianzi University, 2017. Summary of the invention

[0005] In view of the shortcomings in the prior art, the present invention provides a method for correcting the eccentric imaging of the side of a laser line scan tire, which can effectively improve the contrast of the laser line scan image of the side of the tire, especially the contrast of the pre-captured text and the background consistency, solve the problem of curvature of the imaging image caused by eccentricity when the tire rotates, and provide high-quality preprocessing results for character recognition on the tire surface.

[0006] The object of the present invention is achieved by: a laser line scanning tire side eccentric imaging correction processing method, comprising the following steps:

[0007] S1 uses a laser scanner to obtain tire information and obtain a csv file, in which each point is a depth value;

[0008] S2 performs denoising preprocessing and depth consistency preprocessing on the csv file, and converts the csv file into an image of real data;

[0009] S3 creates a tire edge serration template;

[0010] S4 locates and fits the smooth curve through the tire sawtooth template and performs eccentricity correction on the image;

[0011] S5 tire image enhancement processing;

[0012] S6 creates a tire hair template of the tire image;

[0013] S7 uses template matching to locate the fetal hair in the image and perform burr removal;

[0014] S8 adopts the method of processing the line and background in a consistent manner and processes the text area in sections.

[0015] As a further limitation of the present invention, step 2 is specifically:

[0016] Remove the invalid points in the csv file, that is, the values ​​between 10 -9 Below, set it to 0 and remove it, and find the deepest value max and the shallowest value min among the remaining values ​​in the csv file;

[0017] According to the noise ratio, the area around min and max is set to 0.01% and 0.001% respectively. By setting the step size to 0.05μm, the depth value is the horizontal coordinate, min is the starting point, max is the end point, and the number of pixels is the vertical coordinate, a depth histogram is drawn, and the number of pixels is S. From the two end points of the horizontal coordinate, min and max, by accumulating to S×0.01% and S×0.001% of the number of pixels, two thresholds Threshmin and Threshmax can be obtained. The depth values ​​in the csv file that are less than Threshmin are set to Threshmin, and those greater than Threshmax are set to Threshmax. Through conversion, the depth range is unified, and the csv file is converted into a real type image. This step solves the problem of uneven depth caused by high in the middle and low on both sides when scanning the tire side.

[0018] As a further limitation of the present invention, step 4 is specifically:

[0019] Using the created tire serrated edge template, after matching and positioning to obtain the coordinates of each tire serration and fitting a smooth curve, the smooth curve obtained by fitting intersects with the first serration point to obtain the standard row difference Row_ref. Then, at each intersection point, calculate the row coordinate Row of the intersection of the smooth curve and the serration intersection point. If Row > Row_ref, the adjustment value Row_adjust is obtained by Row - Row_ref, and then correction is performed by Row - Row_adjust. If Row < Row_ref, the adjustment value Row_adjust is obtained by Row_ref - Row, and then correction is performed by Row + Row_adjust. Set the gray value of the corrected pixel point, and repeat this operation until all pixel points in this row are corrected. This step solves the problem of uneven imaging caused by rotational eccentricity during tire scanning.

[0020] As a further limitation of the present invention, step 5 is specifically as follows:

[0021] In the Halcon software, taking the pixels of each row of the image as the processing object, the min_max_gray operator is used to eliminate 0.01% of the maximum and minimum gray values to obtain the minimum gray value MinLow of the remaining part; then eliminate 0.02% of the maximum and minimum gray values to obtain the minimum gray value MinHigh of the remaining part; set the background gray value row_bkg_val as the average value of the two minimum gray values, and then subtract row_bkg_val from the gray value of each row of pixels until all rows complete this operation. This step greatly enhances the contrast between the tire background and the characters, facilitating subsequent processing operations.

[0022] As a further limitation of the present invention, step 7 is specifically as follows:

[0023] According to the established fetal hair template, match and position the fetal hair and its shadow area in the tire image, eliminate the original image information of these fetal hair and its shadow area, and fill it with the average value of the surrounding pixel points. This step eliminates the interference of fetal hair on the tire characters and provides a clear image background.

[0024] As a further limitation of the present invention, step 8 is specifically as follows:

[0025] The character area to be recognized is located by template matching. After the rectangular area is generated, the min_max_gray operator is used to remove the interference of the maximum and minimum grayscale values ​​of 5%. According to the grayscale histogram selection, the minimum grayscale value MinBkg is obtained in the remaining tire image area; then all grayscale values ​​in the original image are uniformly subtracted from MinBkg, and the pixel points with grayscale values ​​less than MinBkg in the area are set to 0 and repeated to complete the background consistency processing of all rows in the area. This step improves the contrast between the background and the characters themselves in a specific area, effectively improving the accuracy of tire surface character recognition.

[0026] The present invention first obtains the reflection information csv file of the tire side through a laser scanner, removes invalid points in the csv file, and finds the deepest value max and the shallowest value min in the file; through a depth histogram, the number of pixels accumulated to the noise proportion is obtained to obtain thresholds Threshmin and Threshmax, and then threshold processing is performed on the csv file to complete depth consistency processing; secondly, a tire edge sawtooth template is created, a matching and positioning fitting smooth curve is matched, and then the first sawtooth point is selected as a reference coordinate to perform eccentricity correction processing on the image; then min_max_gray is used to obtain the average background value of each row of pixels, and each row of pixels is enhanced; then a tire hair template of the tire image is created to locate the tire hair position and then a burr removal operation is performed; finally, a text area is located through template matching, and a minimum grayscale value MinBkg is obtained by using a min_max_gray operator, and the grayscale value in the area is subtracted from MinBkg to complete the row background consistency processing.

[0027] Compared with the prior art, the beneficial effects of the present invention are as follows: in view of the noise in the laser line scan camera and the uneven imaging caused by the bending of the tire surface; the bending of the imaging picture caused by the eccentricity when the tire rotates, etc., a laser line scan image preprocessing method suitable for tire surface character recognition is proposed. The present invention can effectively improve the contrast of the laser line scan image on the tire surface, especially the contrast and background consistency of the text area, solve the problem of the bending of the imaging picture caused by the eccentricity when the tire rotates, and provide high-quality preprocessing results for character recognition on the tire surface. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0029] Figure 1 It is a flow chart of the present invention.

[0030] Figure 2 This is an image schematic diagram of converting real data after deep consistency processing of CSV files in the present invention.

[0031] Figure 3 It is a schematic diagram of the tire edge serration fitting correction in the present invention.

[0032] Figure 4 It is a schematic diagram of the eccentricity correction effect in the present invention.

[0033] Figure 5 It is a schematic diagram of the image enhancement effect in the present invention.

[0034] Figure 6 Schematic diagram of the fetal hair image removal effect in the present invention.

[0035] Figure 7 It is a schematic diagram of the effect of line background consistency in the present invention. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] like Figure 1 As shown, a laser line scanning tire side eccentricity imaging correction processing method includes the following steps.

[0038] Step 1: Use a laser scanner to obtain tire information.

[0039] Step 2: De-noising and depth consistency preprocessing of the tire laser line scan file, specifically:

[0040] The invalid points (values ​​between 10 and 10) in the csv file (the csv file is the reflection information of the tire scanned by the laser scanner, and each point in the file is the depth value) are -9 And below), set them to 0 and remove them, find the deepest value max and the shallowest value min among the remaining values ​​in the csv file;

[0041] According to the noise ratio, set the areas around min and max to 0.01% and 0.001% respectively. By setting the step size to 0.05 microns, with the depth value as the abscissa, min as the starting point, max as the ending point, and the number of pixels as the ordinate, draw a depth histogram. Count the number of pixel points as S. Starting from the two endpoints min and max of the abscissa, by accumulating to the number of pixel points of S×0.01% and S×0.001%, two thresholds Threshmin and Threshmax can be obtained. Set the depth values in the csv file that are less than Threshmin to Threshmin, and those greater than Threshmax to Threshmax. Through transformation, the depth range is unified, and the csv file is converted into a real-type image, as attached Figure 2 as shown

[0042] Step 3: Create a tire edge sawtooth template, specifically: Select the tire edge sawtooth area in the Halcon software and create an image template

[0043] Step 4: Locate and fit a smooth curve through the tire sawtooth template and perform eccentric correction on the image, specifically

[0044] After using the created tire sawtooth edge template to match and locate each tire sawtooth coordinate, fit a smooth curve. Then, use the fitted smooth curve to intersect with the first sawtooth point to obtain the standard row difference Row_ref; then calculate the row coordinate Row of the intersection of the smooth curve and the sawtooth at each intersection point; if Row>Row_ref, obtain the adjustment value Row_adjust through Row-Row_ref, and then perform correction through Row-Row_adjust; if Row<Row_ref, obtain the adjustment value Row_adjust through Row_ref-Row, and then perform correction through Row+Row_adjust. Set the gray value of the pixel points after correction, and repeatedly execute this operation until the correction of all pixel points in this row is completed, as attached Figure 3 and 4 as shown

[0045] Step 5: Tire image enhancement processing, specifically

[0046] Taking the pixels of each row of the image as the processing object, use the min_max_gray operator to remove 0.01% of the maximum and minimum gray values to obtain the minimum gray value MinLow of the remaining part; then remove 0.02% of the maximum and minimum gray values to obtain the minimum gray value MinHigh of the remaining part; set the background gray value row_bkg_val to the average of the two minimum gray values, and then subtract row_bkg_val from the gray value of each row of pixels until this operation is completed for all rows, as attached Figure 5 as shown

[0047] Step 6: Create a tire hair template for the tire image. Specifically, create a grayscale image template of the tire hair part and the shadow area in Halcon.

[0048] Step 7 uses template matching to locate the fetal hair in the image and perform burr removal operations, specifically:

[0049] According to the established fetal hair template, the fetal hair and its shadow area in the tire image are matched and located, the original image information of these fetal hair and its shadow area is removed, and the surrounding pixels are used to take the average value to fill it, as shown in the attached figure. Figure 6 shown.

[0050] Step 8 uses the same processing method for rows and backgrounds to process the text area in sections, specifically:

[0051] The character area to be recognized is located through template matching. After the rectangular area is generated, the min_max_gray operator is used to eliminate 5% (selected according to the grayscale histogram) of the interference of the maximum and minimum grayscale values, and the minimum grayscale value MinBkg is obtained in the remaining area (tire image); then all grayscale values ​​in the original image are uniformly subtracted from MinBkg, and this operation is repeated for the pixels in the area whose grayscale values ​​are less than MinBkg, completing the background consistency processing of all rows in the area.

[0052] As attached Figure 7 As shown in the figure, the final processed image solves the eccentricity problem and grayscale unevenness problem of the tire during online scanning imaging, eliminates the interference of tire hair, and has a high contrast ratio of the image character area, which greatly improves the accuracy of character recognition.

[0053] The above embodiments are only used to help understand the method and core idea of ​​the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A method for correcting the eccentric imaging of the side of a tire by laser line scanning, characterized in that, it includes the following steps: S1 Use a laser scanner to obtain tire information and get a csv file, where each point in the file is a depth value; S2 Perform denoising preprocessing and depth uniformity preprocessing on the csv file, and convert the csv file into an image of real data. The specific steps of step 2 are: Remove the invalid points in the csv file, that is, the values ​​between 10 -9 Below, set it to 0 and remove it, and find the deepest value max and the shallowest value min among the remaining values ​​in the csv file; According to the noise ratio, set the areas around min and max to 0.01% and 0.001% respectively. By setting the step size to 0.05μm, with the depth value as the abscissa, min as the starting point, max as the ending point, and the number of pixels as the ordinate, draw a depth histogram. Count the number of pixel points as S. Starting from the two endpoints min and max of the abscissa, accumulate to the number of pixel points of S×0.01% and S×0.001% respectively, and two thresholds Threshmin and Threshmax can be obtained. Set the depth values in the csv file that are less than Threshmin to Threshmin, and those greater than Threshmax to Threshmax. Through conversion, the depth range is unified, and the csv file is converted into an image of real type; S3 Create a tire edge sawtooth template; S4 Locate and fit a smooth curve through the tire sawtooth template, and perform eccentric correction processing on the image. The specific steps of step 4 are: Using the created tire sawtooth edge template, after matching and positioning to obtain the coordinates of each tire sawtooth, fit a smooth curve. Then, use the fitted smooth curve to intersect with the first sawtooth point to obtain the standard row difference Row_ref; then calculate the row coordinate Row of the intersection of the smooth curve and the sawtooth at each intersection point; if Row>Row_ref, then obtain the adjustment value Row_adjust through Row-Row_ref, and then perform correction through Row-Row_adjust; if Row<Row_ref, then obtain the adjustment value Row_adjust through Row_ref-Row, and then perform correction through Row+Row_adjust. Set the gray value of the pixel points after correction, and repeat this operation until all pixel points in this row are corrected; S5 Tire image enhancement processing; S6 Create a tuft template for the tire image; S7 Locate the tufts in the image by template matching method and perform deburring operation; S8 Adopt a processing method with consistent row background to process the text area in segments. The specific steps of step 8 are: Locate the character area to be recognized through template matching. After generating a rectangular area, eliminate the interference of 5% of the maximum and minimum gray values through the min_max_gray operator. Select according to the gray histogram, and obtain the minimum gray value MinBkg in the remaining tire image area; then subtract MinBkg from all gray values in the original image, and set the pixel points with gray values less than MinBkg in the area to 0. Repeat this operation to complete the background consistency processing of all rows in the area.

2. The method for correcting the eccentric imaging of the side of a tire by laser line scanning according to claim 1, characterized in that, the specific steps of step 5 are: In Halcon software, each row of pixels in the image is taken as the processing object. The min_max_gray operator is used to remove 0.01% of the maximum and minimum grayscale values ​​to obtain the minimum grayscale value MinLow of the rest; then 0.02% of the maximum and minimum grayscale values ​​are removed to obtain the minimum grayscale value MinHigh of the rest; the background grayscale value row_bkg_val is set to the average of the two minimum grayscale values, and then the grayscale value of each row of pixels is subtracted from row_bkg_val until this operation is completed for all rows.

3. The laser line scanning tire side eccentricity imaging correction processing method according to claim 2, It is characterized in that Step 7 is as follows: According to the established fetal hair template, the fetal hair and its shadow area in the tire image are matched and located, the original image information of these fetal hair and its shadow area is eliminated, and the average value of the surrounding pixels is used to fill it.

Images