Methods for testing the straightness of structural steel
Patent Information
- Application Number
- TW114112867
- Authority / Receiving Office
- TW · TW
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-04-02
- Publication Date
- 2026-10-16
- Estimated Expiration
- 2045-04-01
AI Technical Summary
Existing methods for measuring the bending of structural steel profiles face challenges in achieving high precision due to the limitations of laser ranging devices, which are less accurate at greater distances and can only measure one edge at a time, leading to inefficiencies in measurement accuracy and efficiency.
A method utilizing an image capturing unit and processing unit to analyze an original input image of the steel profile, involving noise filtering, edge detection, contour line fitting, and baseline calculation to determine straightness, enabling simultaneous measurement of multiple edges and improving accuracy.
This approach enhances measurement accuracy and efficiency by capturing the original input image, filtering noise, and calculating straightness through polynomial regression, effectively addressing the limitations of laser-based methods.
Smart Images
Figure TWG2TA001080034_001 
Figure TWG2TA001080034_002 
Figure TWG2TA001080034_003
Abstract
Description
[Technical Field]
[0001] This invention relates to a measurement method, and more particularly to a method for testing steel. [Previous Technology]
[0002] During the processing of steel, it first undergoes a hot rolling process to form the steel into a predetermined thickness and shape, and then a cooling process to solidify the steel. However, during the cooling process, uneven temperature distribution or phase changes can easily lead to the accumulation of internal stress, causing localized expansion of the steel's volume, resulting in deformation or bending. Therefore, in the later stages of the processing, it is necessary to use devices and methods to measure the bending degree of the steel in order to control the quality of the finished steel product.
[0003] A conventional method for measuring the bending of a steel profile involves placing one end of the steel profile on a conveyor roller conveyor. The steel profile is pushed and conveyed by multiple conveyor rollers to a steel profile end positioning device, where one end is fixed. After the steel profile stabilizes, a laser ranging device is activated, emitting multiple laser beams from multiple laser light sources. The laser beams strike one edge of the steel profile and are reflected back into the laser ranging device for reception. The laser ranging device calculates the distance between the side of the steel profile and the laser ranging device by measuring the time difference between the emission and reception of the laser beams, and simultaneously transmits the measurement data to a computer. The computer converts the measurement data from the laser ranging device into a graphic and displays the bending shape of the steel profile, and calculates the straightness of the steel profile and the distance from the bending point to the end of the steel profile.
[0004] However, this method for measuring the bending of steel profiles mainly relies on laser sensing technology. Therefore, the greater the distance between the laser ranging device and the steel profile, the lower the measurement accuracy becomes, making it difficult to achieve high precision. Furthermore, the laser ranging device can only measure one edge of the steel profile at a time, and cannot simultaneously measure multiple edges of the steel profile. [Summary of the Invention]
[0005] Therefore, the object of the present invention is to provide a detection method that can at least overcome the disadvantages of the prior art.
[0006] Therefore, the method for detecting the straightness of steel profiles of the present invention is executed by a processing unit, which is signal-connected to an image capturing unit to receive and analyze an original input image of a steel profile to be detected obtained from the image capturing unit. The method includes steps (A) to (G):
[0007] In this step (A), an original product image is obtained based on the original input image.
[0008] In this step (B), noise in the original product image is filtered out to obtain a product image to be located.
[0009] In this step (C), an edge of the image of the product to be located is detected, the edge containing multiple edge points.
[0010] In this step (D), a contour line of the edge of the image of the product to be located is fitted using polynomial regression.
[0011] In this step (E), multiple edge points to be calculated are selected from multiple edge points of the edge based on the edge contour line.
[0012] In this step (F), two reference points are obtained from the waiting operation edge point, and a reference line is defined through these reference points.
[0013] In this step (G), the distance between the edge point to be calculated and the baseline is calculated to analyze the straightness of the steel section to be tested.
[0014] The advantages of this invention are as follows: This invention obtains the original input image of the steel section to be inspected through the image capturing unit, and then performs steps such as filtering noise, acquiring edge contour lines, calculating baselines, and analyzing straightness using the processing execution unit. Therefore, it can improve the problem of poor accuracy in long-distance measurement and effectively improve the accuracy of measurement. Furthermore, since the image capturing unit can capture the original input image of the steel section, the processing execution unit can obtain the straightness of multiple edges of the steel section, greatly improving measurement efficiency.
Implementation Method
[0015] The relevant technical content, features, and effects of the present invention will be clearly presented in the following detailed description of the embodiments with reference to the accompanying drawings. Furthermore, it should be noted that the drawings of the present invention are only for illustrating the structural and / or positional relationships between elements and are not related to the actual dimensions of each element.
[0016] One embodiment of the method for detecting the straightness of steel profiles of the present invention is executed by a processing unit 1. The processing unit 1 is signal-connected to an image capturing unit 2 to receive and analyze an original input image of a steel profile 9 to be detected obtained from the image capturing unit 2. Referring to FIG1, the method includes steps S1 to S7, which will be described step by step below with reference to the figures.
[0017] In step S1, the image capturing unit 2 acquires the original input image of the steel profile 9 to be inspected and a background image that does not contain the steel profile 9 to be inspected, and obtains an original product image based on the acquired original input image and the background image. In this embodiment, the image capturing unit 2 acquires multiple product images containing the steel profile 9 to be inspected along a length direction L of the steel profile 9 to be inspected, and performs image stitching on these product images to obtain the original input image.
[0018] Referring to Figure 2, in this embodiment, the steel section 9 to be tested is illustrated using an H-beam as an example. However, this is not the limitation in other embodiments of this embodiment, and can also be applied to other types of steel such as C-beams, L-beams, and I-beams. In this embodiment, the image capturing unit 2 includes multiple color cameras. However, in other embodiments of this embodiment, the image capturing unit 2 may also include at least one color or monochrome camera.
[0019] Referring to Figure 3, step S1 further includes sub-steps S11 and S12. In step S11, the image capturing unit 2 converts the original input image and the background image to grayscale. In this embodiment, the original input image and the background image obtained by the image capturing unit 2 are color images. Through the grayscale conversion in step S11, the color image can be converted into an 8-bit grayscale image, effectively reducing the amount of data required for subsequent processing.
[0020] In step S12, the image capturing unit 2 uses the background subtraction method to compare the grayscale difference between the original input image and the background image, and combines it with threshold determination to obtain the original product image after removing the background, and locate the position of the steel profile 9 to be detected.
[0021] In step S2, the processing unit 1 filters out noise from the original product image to obtain a product image to be located. The original product image obtained in step S1 still retains speckled noise similar to pepper salt caused by reflections from sawdust or water on the inspection bed, so this noise must be further filtered out in step S2. Referring to Figure 4, step S2 also includes sub-steps S21 and S22. In step S21, the processing unit 1 uses an image noise reduction method to filter out speckled noise from the original product image. The image noise reduction method uses a 7x7 mask to perform median filtering to smooth the image, thereby eliminating the speckled noise. In step S22, the processing unit 1 uses an adaptive image thresholding method, such as the Otsu algorithm, to binarize the original product image to obtain the product image to be located.
[0022] Referring to Figure 5, in step S3, the processing unit 1 detects an edge 91 of the product image to be located via morphological detection. The edge 91 includes a plurality of edge points. Referring to Figure 6, step S3 also includes sub-steps S31 and S32.
[0023] In step S31, the processing unit 1 performs morphological region filling on each region of the product image to be filled to obtain the product image after filling. For each region to be filled, the region filling can be expressed as the following formula (1), which mainly includes steps such as dilation, complementation and intersection, which can eliminate the tiny holes in the foreground part caused by factors such as reflection, water stains, and textures in the preprocessing.
[0024] …(1)
[0025] Where B is a structural element (see Figure 7), which is the complement of the image of the product to be positioned without the hole filling process, and is the element in the corresponding area to be filled. It will gradually fill the corresponding area to be filled from k=1, 2, 3 until, that is, the corresponding area to be filled is filled.
[0026] In this step S32, the processing unit 1 uses morphological boundary extraction to detect the edge 91 of the product image to be positioned after the hole is filled. The boundary extraction is to perform edge detection by boundary extraction. The principle is to erode the original area once with a 3x3 structuring element, and then use the difference set between the original area and the eroded area to find the pixel-level boundary of the original area, that is, the edge points.
[0027] In step S4, the processing unit 1 uses polynomial regression to fit an edge contour line of the edge 91 of the product image to be located. The polynomial regression system calculates a least square regression curve using the least square method, and uses the least square regression curve as the edge contour line of the edge 91.
[0028] In step S5, the processing unit 1 selects multiple edge points P to be processed from multiple edge points of the edge 91 based on the edge contour line. In this embodiment, since the least square regression curve calculated in step S4 may not pass through all such edge points, in step S5, the processing unit 1 calculates the distance from each point to the least square regression curve, calculates the standard deviation of the distance, and sets a threshold, considering data greater than the threshold as outliers. The processing unit 1 removes these outliers from the edge points to obtain the edge points P to be processed.
[0029] Referring to Figure 8, in step S6, the processing unit 1 obtains two reference points Pb from the waiting operation edge point P, and defines a baseline BL through the reference points Pb. The processing unit 1 extracts the coordinate values of the two reference points Pb of the waiting operation edge point P, and calculates a linear equation through the coordinate values of the reference points Pb, which is used as the baseline BL.
[0030] In step S7, the processing unit 1 calculates the distance between the edge point P to be calculated and the baseline BL to analyze the straightness of the steel section 9 to be tested. In this embodiment, the distance is the shortest distance between each edge point P to be calculated and the baseline BL, and the largest of the equal distances is the maximum distance d between the edge point P to be calculated and the baseline BL.
[0031] In summary, the method for detecting the straightness of steel profiles of the present invention obtains the original input image of the steel profile 9 to be detected through the image capturing unit 2, and then performs steps such as filtering noise, obtaining edge contour lines, calculating the baseline BL and analyzing the straightness through the processing execution unit. This method can greatly improve the accuracy of measurement and simplify the overall measurement architecture.
[0032] However, the above description is only an embodiment of the present invention and should not be construed as limiting the scope of the present invention. Any simple equivalent changes and modifications made in accordance with the scope of the patent application and the contents of the patent specification of the present invention shall still fall within the scope of the patent of the present invention. [Simplified Explanation of the Diagram]
[0033] Other features and effects of the present invention will be clearly presented in the embodiments with reference to the drawings, wherein:
[0034] Figure 1 is a flowchart of one embodiment of the method for detecting the straightness of steel profiles of the present invention.
[0035] Figure 2 is a schematic diagram of a steel section to be tested.
[0036] Figure 3 is a flowchart of the sub-step of step S1 in Figure 1.
[0037] Figure 4 is a flowchart of the sub-step of step S2 in Figure 1.
[0038] Figure 5 is a schematic diagram of one edge of the image of the product to be positioned, one of the steel profiles to be inspected.
[0039] Figure 6 is a flowchart of the sub-step of step S3 in Figure 1.
[0040] Figure 7 is a schematic diagram of the structural elements used in step S31 of Figure 6.
[0041] Figure 8 is a schematic diagram of a baseline of the edge of the steel section.
Claims
1. A method for detecting the straightness of a steel profile, executed by a processing unit connected to an image capturing unit, for receiving and analyzing an original input image of a steel profile to be detected obtained from the image capturing unit, the method comprising the following steps: (A) obtaining an original product image based on the original input image; (B) filtering out noise from the original product image to obtain a product image to be located; (C) detecting an edge of the product image to be located, the edge comprising multiple edge points; (D) fitting an edge contour line of the edge of the product image to be located using polynomial regression; (E) selecting multiple edge points to be calculated from the multiple edge points of the edge based on the edge contour line; (F) obtaining two reference points from the edge points to be calculated, and defining a reference line through the reference points; and (G) calculating the distance between the edge points to be calculated and the reference line to analyze the straightness of the steel profile to be detected; wherein, Step (C) includes the following sub-steps: (C-1) performing morphological hole filling processing on the image of the product to be located to obtain the image of the product to be located after hole filling; and (C-2) using morphological boundary extraction to detect the edge of the image of the product to be located after hole filling.
2. The method for detecting the straightness of a steel section as described in claim 1, wherein the processing unit further receives a background image that does not contain the steel section to be detected, wherein, Step (A) includes the following sub-steps: (A-1) converting the original input image and the background image to grayscale; and (A-2) comparing the grayscale difference between the original input image and the background image to obtain the original product image after removing the background.
3. The method for detecting the straightness of steel sections as described in claim 1, wherein, Step (B) includes the following sub-steps: (B-1) filtering out speckle noise from the original product image using an image noise reduction method; and (B-2) binarizing the original product image using an adaptive image thresholding method to obtain the product image to be located.
4. The method for detecting the straightness of steel sections as described in claim 3, wherein, In this sub-step (B-1), median filtering is used to eliminate such speckle noise.
5. The method for detecting the straightness of steel sections as described in claim 1, wherein, The image capturing unit acquires multiple product images containing the steel section to be inspected along a length direction, and stitches these product images together to obtain the original input image.