Hot-rolled bar fixed support separation detection method based on image processing

Abstract

The invention relates to the field of fixed support separation of bars, in particular to a fixed support separation detection method of a hot-rolled bar based on image processing, which is used for solving the problems that a single-angle visual detection method is inaccurate in detection and difficult to cope with cross misalignment or end face shielding and the like in a fixed support separation detection process. According to the method, the included angle and the position of the image acquisition equipment on the left side and the right side are automatically adjusted, clear slice images at the left angle, the middle angle and the right angle are obtained, then the number of end faces, center point coordinates and the inclination angle of the bar are recognized by independently analyzing each angle image, multi-angle data are fused for space correlation comparison, and finally the bar is obtained. According to the method, the accuracy of bar quantity statistics is ensured, the actual inclination state of each bar in a three-dimensional space can be accurately restored, and finally, a plurality of early warning signals of different danger levels are generated in combination with double indexes of the bar separation quantity and the separation uniformity, so that accurate distinguishing and response to different defect degrees are realized, and false alarm and missing alarm are effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of bar stock fixed-support separation, specifically to a method for detecting fixed-support separation of hot-rolled bars based on image processing. Background Technology

[0002] In the production of hot-rolled bars, the fixed support separation is a key process before the finished product is bundled, and its technical quality directly affects the efficiency of subsequent transportation and the safety of storage.

[0003] Currently, the industry generally uses manual visual inspection or two-dimensional visual inspection to verify the separation results. Manual visual inspection relies on the experience of operators and has problems such as high labor intensity, strong subjectivity, and easy to miss detection. Traditional two-dimensional visual inspection methods usually only collect images from a single angle and count the number of bar end faces through contour recognition technology. Although it achieves automated inspection to a certain extent, it is limited by the perspective principle of two-dimensional images and cannot accurately identify the spatial posture and tilt angle of the bar.

[0004] Therefore, when bar stock is cross-over or obscured at the end face after separation, images from a single perspective are difficult to reflect the true separation state, which can easily lead to statistical errors and misjudgments of uniformity. This makes it difficult to meet the high-precision control requirements of modern hot rolling production lines for fixed-support separation quality. To address this, this application proposes a solution. Summary of the Invention

[0005] This invention automatically adjusts the angle and position of the image acquisition devices on both sides to ensure the acquisition of clear slice images from the left, center, and right angles. Then, by independently analyzing each angle image, the number of bar end faces, center point coordinates, and tilt angle are identified. Finally, multi-angle data is fused for spatial correlation comparison. This ensures the accuracy of bar quantity statistics and accurately restores the actual tilt state of each bar in three-dimensional space. It also quantitatively evaluates the degree of directional consistency of the bars during fixed-support separation. This invention solves the problems of inaccurate detection by single-angle visual detection methods and difficulty in dealing with cross-inaccuracies or end face occlusion during fixed-support separation detection. Therefore, it proposes an image processing-based fixed-support separation detection method for hot-rolled bars.

[0006] The objective of this invention can be achieved through the following technical solution: a method for detecting the fixed-support separation of hot-rolled bars based on image processing, comprising the following steps: Step 1: Perform fixed-support separation on the hot-rolled bars, and after the separation is completed, acquire images from multiple angles of the bar bundles; Step 2: Analyze the image at each angle separately to obtain the rod detection result at a single angle. The rod detection result at a single angle includes the number of slices. Step 3: Perform spatial correlation analysis on the bar detection results from a single angle, and fuse the bar detection results from multiple angles to obtain the number of separated bars and the degree of separation uniformity; Step 4: Compare the fusion detection results with the set separation results to obtain the separation compliance rate, and use the separation compliance rate to make early warning judgments and output early warnings.

[0007] In a preferred embodiment of the present invention, the method further includes an image acquisition decision module, an image separation processing module, a spatial fusion processing module, a separation comparison module, and an early warning decision module; The image acquisition decision module can perform preliminary image acquisition of the bar bundle, make shooting angle decisions based on the preliminary image acquisition results, and perform image acquisition again based on the shooting angle decisions to obtain slice images from multiple angles. The image separation and processing module acquires slice images and analyzes different slice images independently, thereby counting the number of bars contained in the slice images to obtain the number of slices; The spatial fusion processing module acquires slice images from multiple angles and the corresponding number of slices on the slice images, and spatially combines different slice images according to the acquisition angle to perform spatial prediction and restoration, thereby obtaining the number of separated bars and the degree of separation uniformity. The separation comparison module performs threshold judgment on the number of separated bars and the degree of separation uniformity, and determines whether the separation meets the standard or does not meet the standard based on the threshold judgment result. The early warning decision module statistically analyzes the separation compliance status and generates corresponding early warning outputs based on the separation compliance rate, thus outputting the early warning decision signal.

[0008] In a preferred embodiment of the present invention, the image acquisition decision module performs initial image acquisition through the central image acquisition device to obtain the position of the outer contour of the bar and records the position of the outer contour of the bar in coordinate form. The image acquisition and decision module obtains the length of the bar and the position of the bar end face based on the coordinate difference between the left and right farthest points. The image acquisition decision module generates a bar plane based on the position of the bar end face, and calculates the angle between the left image acquisition device and the left bar plane, and the angle between the right image acquisition device and the right bar plane, which are recorded as the left angle and the right angle, respectively. The image acquisition decision module performs threshold judgment on the left angle and the right angle, and adjusts the image acquisition device that is too close, so as to obtain two slice images, the left end image and the right end image, respectively.

[0009] In a preferred embodiment of the present invention, the image separation processing module marks the near-circular edge of each rod end face in the image, and takes the center point of the near-circular edge as the center of the rod. Then, a coordinate system is created with the central image acquisition device as the origin to obtain the coordinates (X) of the center point of the rod. L i Y L i Z L i ) and (X R m Y R m Z R m ); The image separation and processing module performs a reverse analysis based on the near-circular edge to obtain the tilt angle required to transform the perspective distortion of a perfect circle edge into a near-circular edge, which is recorded as α. L i and tilt angle β R m ; Where i is the number of the bar end face in the left image, L is the left end face, the maximum value of i is equal to the number of bar end faces in the left image, m is the number of the bar end face in the right image, the maximum value of m is equal to the number of bar end faces in the right image, and R is the right end face.

[0010] In a preferred embodiment of the present invention, the spatial fusion processing module compares the number of rod end faces in the left and right images to obtain signals with the same number or signals with different numbers. After obtaining signals with the same number of bars, the spatial fusion processing module records the number of bar end faces in the left or right image as the number of bar separations. After obtaining a signal with a difference in number, it generates an early warning and terminates the subsequent analysis process.

[0011] In a preferred embodiment of the present invention, the spatial fusion processing module selects the left / right end image as the anchor image and selects the corresponding overlapping center point coordinates from the image of the other end. If there are overlapping center point coordinates, they are recorded as overlapping samples; if there are no overlapping center point coordinates, they are recorded as difference samples. The spatial fusion processing module calculates the proportion of the number of differential samples in the total number of rod end faces in the left / right images to obtain the contrast difference rate. The spatial fusion processing module calculates the variance of the tilt angle of each bar end face on the left and right end face images.

[0012] In a preferred embodiment of the present invention, the spatial fusion processing module obtains the separation difference degree based on the comparison difference rate and tilt angle, and records the reciprocal of the separation difference degree as the separation uniformity degree.

[0013] In a preferred embodiment of the present invention, the separation comparison module performs a threshold judgment on the number of separated bars to determine whether the number meets the standard or the number is abnormal. The separation comparison module performs a threshold judgment on the uniformity of the bar separation to determine whether the uniformity meets the standard or the uniformity is abnormal. When the early warning decision module simultaneously obtains that the quantity and uniformity meet the standards, it generates a separation normal signal; when it simultaneously obtains that the quantity and uniformity are abnormal, it generates a low-risk early warning signal; and when it simultaneously obtains that the quantity and uniformity are abnormal, it generates a high-risk early warning signal.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. In this invention, the position of the end face of the bar is automatically calculated during the initial acquisition, and the angle and position of the image acquisition devices on the left and right sides are dynamically adjusted to ensure that clear slice images from the left, center and right angles are obtained. Then, by independently analyzing the image from each angle, the number of end faces of the bar, the coordinates of the center point and the tilt angle are identified. Then, the multi-angle data is fused for spatial correlation comparison, which not only ensures the accuracy of the bar quantity statistics, but also quantitatively evaluates the spatial consistency and neatness of the separated bars through the comparison of the center point coordinates and the analysis of the tilt angle.

[0015] 2. In this invention, by comparing whether the number of bars on the left and right end faces is consistent, it is possible to quickly determine whether there is an obvious separation anomaly. By calculating the difference in the coordinates of the center point and the distribution of the tilt angle, the actual tilt state of each bar in three-dimensional space is accurately restored. The degree of directional consistency of the bars during fixed-support separation is quantitatively evaluated, and the overall neatness of the bar binding is intuitively reflected. Finally, by combining the dual indicators of the number of separated bars and the degree of separation uniformity, multiple warning signals of different danger levels are generated, so as to achieve accurate differentiation and response to different defect degrees and effectively avoid false alarms and missed alarms. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Please refer to Figure 1 - Figure 2 As shown, the image processing-based method for detecting the fixed-support separation of hot-rolled bars includes the following steps: Step 1: Perform fixed-support separation on the hot-rolled bars, and after separation, acquire images from above the plate. Based on the initial acquisition results, determine the acquisition angles of the bars at both ends, thereby adjusting the image acquisition angles on both sides and obtaining multiple angle images of the bar bundles. Step 2: Analyze the image from each angle separately to obtain the number of bar end faces on the left and right end faces and the tilt angle of each bar. Step 3: Perform spatial correlation analysis on the bar detection results from a single angle, integrate the bar detection results from multiple angles, compare the number of bar end faces on the left and right ends to obtain the number of separated bars, and then compare the center point coordinates and tilt angles of the bars to obtain the degree of separation uniformity. Step 4: Compare the fusion detection results with the set separation results to obtain the separation compliance rate, and use the separation compliance rate to make early warning judgments and output early warnings.

[0020] Example 2: Please refer to Figure 1 - Figure 2 As shown, the hot-rolled bar fixed-support separation detection method based on image processing also includes an image acquisition decision module, an image separation processing module, a spatial fusion processing module, a separation comparison module, and an early warning decision module; The image acquisition and decision module can perform initial image acquisition through the top center image acquisition device when the bundled rod first reaches the image detection position. The acquired image is then identified by a pre-trained recognition algorithm to obtain the outer contour position of the rod. The outer contour position of the rod is recorded in coordinate form with the top image acquisition device as the origin. After acquiring the coordinates of the outer contour of the bar, the image acquisition and decision module selects the coordinates of the farthest points on the left and right sides of the outer contour of the bar, and obtains the bar length and the position of the bar end face based on the coordinate difference between the farthest points on the left and right sides. The image acquisition decision module generates a bar plane based on the position of the bar end face and calculates the angle between the left image acquisition device and the left bar plane, and the angle between the right image acquisition device and the right bar plane, which are recorded as the left angle and the right angle, respectively. The image acquisition decision module performs threshold judgment on the left angle and the right angle. If the left angle or the right angle is less than the set threshold, it is judged that the distance is too close, and the image acquisition device that is too close is moved to the outside of the image acquisition device away from the center image acquisition device. If the left angle or the right angle is greater than the set threshold, it is judged that the distance is normal, and the image acquisition device with a normal distance is not adjusted.

[0021] After moving the image acquisition device that is too close to the outside, the image acquisition decision module will make another judgment on the included angle. If the judgment results are all normal distances, the image acquisition will be acquired synchronously through the central image acquisition device and the left and right image acquisition devices, and three slice images will be obtained respectively: the left image, the central image, and the right image.

[0022] The image acquisition decision module sends three slice images to the image separation processing module; After acquiring the sliced ​​images, the image separation and processing module uses an image recognition algorithm to identify the end faces of the bars in multiple sliced ​​images to obtain the number and position of the bars. Specifically: When processing the left-end image, the image separation processing module first uses an algorithm to mark the near-circular edges of each rod end face in the image, and takes the center point of the near-circular edge as the center of the rod. Then, it creates a coordinate system with the central image acquisition device as the origin, and records the center of the rod as a coordinate point on the coordinate system to obtain the center point coordinates (X). L i Y L i Z L i ), where i is the number of the bar end face in the left image, L is the left end face, and the maximum value of i is equal to the number of bar end faces in the left end face; The image separation processing module performs a reverse analysis based on the circular edge using a perspective algorithm to obtain the tilt angle required to transform the circular edge into a circular edge through perspective distortion, which is recorded as α. L i ; The image separation processing module processes the right-end image using the same method to obtain the center point coordinates (X) of each rod end face in the right-end image. R m Y R m Z R m ) and tilt angle β R m , where m is the number of the bar end face in the right end image, the maximum value of m is equal to the number of bar end faces in the right end face, and R is the right end face; The image separation processing module sends the analysis results to the spatial fusion processing module. The spatial fusion processing module compares the number of bar end faces in the left and right images to obtain a signal of the same number or a signal of different number. After obtaining a signal of the same number, the spatial fusion processing module records the number of bar end faces in the left or right image as the number of bar separations. After obtaining a signal of different number, it generates an early warning and terminates the subsequent analysis process. The spatial fusion processing module first selects the left image as the anchor image and selects the corresponding overlapping center point coordinates from the right image. If there are overlapping center point coordinates, the bar end face corresponding to the overlapping center point coordinates is recorded as an overlapping sample. If there are no overlapping center point coordinates, the bar end face corresponding to the non-overlapping center point coordinates is recorded as a difference sample. The method for determining the coincident center coordinates is as follows: if the difference vector between the center point coordinates is less than a set threshold and the tilt angle values ​​are the same, then the coordinates are recorded as coincident. The difference vector is calculated as (X... L -X R m Y L -Y R m Z L -Z R m ); The spatial fusion processing module calculates the proportion of the number of differential samples in the total number of bar end faces in the left image and records it as the contrast difference rate. The spatial fusion processing module selects the tilt angle of each bar end face on the left and right end face images, and performs statistical calculations on the tilt angles to obtain the variance of the tilt angle on the left end face image and the variance of the tilt angle on the right end face image. The spatial fusion processing module normalizes the variance of the comparison difference rate and tilt angle, and then sums the weighted results of the normalization to obtain the separation difference degree. The reciprocal of the separation difference degree is recorded as the separation uniformity degree. By comparing the two ends of the bar, it can be determined whether the angles of multiple different bars are the same after separation, and thus judge the neatness of the bars. The separation comparison module compares the number of bars separated with the set separation standard. If the number of bars separated is the same as the set separation standard, it is recorded as meeting the standard; otherwise, an abnormal number is generated. The separation comparison module compares the uniformity of the bar separation with a set threshold. If the uniformity of separation is greater than the set threshold, it is recorded as uniformity meeting the standard; otherwise, it is recorded as uniformity abnormal. The early warning decision module generates a normal separation signal when both quantity and uniformity meet the standards, and generates a low-risk early warning signal when both quantity and uniformity are abnormal. It also generates a high-risk early warning signal when both quantity and uniformity are abnormal. This allows for different levels of early warning based on the separation status of the bars, improving system flexibility.

[0023] Thresholds, preset values, preset ranges, etc. are set for result comparison and analysis to determine good or bad. The size of these values ​​is determined by a combination of large-scale model analysis of sample data and human experience. They can also be adjusted appropriately based on seasonal or common-sense influences. Furthermore, the settings for weighting ratios, influence factors, etc., are based on the magnitude of each parameter's influence on the results. The specific values ​​are allocated to ultimately reflect the impact on the results. The settings for input and storage are also determined by a combination of large-scale model analysis of sample data and human experience. Appropriate adjustments can also be made based on seasonal or rational influence conditions.

[0024] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for detecting the fixed-support separation of hot-rolled bars based on image processing, characterized in that, Includes the following steps: Step 1: Perform fixed-support separation on the hot-rolled bars, and after the separation is completed, acquire images from multiple angles of the bar bundles; Step 2: Analyze the image at each angle separately to obtain the rod detection result at a single angle. The rod detection result at a single angle includes the number of slices. Step 3: Perform spatial correlation analysis on the bar detection results from a single angle, and fuse the bar detection results from multiple angles to obtain the number of separated bars and the degree of separation uniformity; Step 4: Compare the fusion detection results with the set separation results to obtain the separation compliance rate, and use the separation compliance rate to make early warning judgments and output early warnings.

2. The image processing-based hot-rolled bar support separation detection method according to claim 1, characterized in that, It also includes an image acquisition decision module, an image separation and processing module, a spatial fusion processing module, a separation and comparison module, and an early warning decision module; The image acquisition decision module can perform preliminary image acquisition of the bar bundle, make shooting angle decisions based on the preliminary image acquisition results, and perform image acquisition again based on the shooting angle decisions to obtain slice images from multiple angles. The image separation and processing module acquires slice images and analyzes different slice images independently, thereby counting the number of bars contained in the slice images to obtain the number of slices; The spatial fusion processing module acquires slice images from multiple angles and the corresponding number of slices on the slice images, and spatially combines different slice images according to the acquisition angle to perform spatial prediction and restoration, thereby obtaining the number of separated bars and the degree of separation uniformity. The separation comparison module performs threshold judgment on the number of separated bars and the degree of separation uniformity, and determines whether the separation meets the standard or does not meet the standard based on the threshold judgment result. The early warning decision module statistically analyzes the separation compliance status and generates corresponding early warning outputs based on the separation compliance rate, thus outputting the early warning decision signal.

3. The image processing-based hot-rolled bar support separation detection method according to claim 2, characterized in that, The image acquisition and decision module performs initial image acquisition through the central image acquisition device to obtain the position of the outer contour of the bar and records the position of the outer contour of the bar in coordinate form. The image acquisition and decision module obtains the length of the bar and the position of the bar end face based on the coordinate difference between the left and right farthest points. The image acquisition decision module generates a bar plane based on the position of the bar end face, and calculates the angle between the left image acquisition device and the left bar plane, and the angle between the right image acquisition device and the right bar plane, which are recorded as the left angle and the right angle, respectively. The image acquisition decision module performs threshold judgment on the left angle and the right angle, and adjusts the image acquisition device that is too close, so as to obtain two slice images, the left end image and the right end image, respectively.

4. The image processing-based method for detecting the fixed-support separation of hot-rolled bars according to claim 2, characterized in that, The image separation and processing module marks the near-circular edge of each rod end face in the image, and takes the center point of the near-circular edge as the center of the rod. Then, a coordinate system is created with the central image acquisition device as the origin to obtain the coordinates (X, Y, X) of the center point of the rod. L i Y L i Z L i ) and (X R m Y R m Z R m ); The image separation and processing module performs a reverse analysis based on the near-circular edge to obtain the tilt angle required to transform the perspective distortion of a perfect circle edge into a near-circular edge, which is recorded as α. L i and tilt angle β R m ; Where i is the number of the bar end face in the left image, L is the left end face, the maximum value of i is equal to the number of bar end faces in the left image, m is the number of the bar end face in the right image, the maximum value of m is equal to the number of bar end faces in the right image, and R is the right end face.

5. The image processing-based method for detecting the fixed-support separation of hot-rolled bars according to claim 2, characterized in that, The spatial fusion processing module compares the number of rod end faces in the left and right images to obtain signals with the same number or signals with different numbers. After obtaining signals with the same number of bars, the spatial fusion processing module records the number of bar end faces in the left or right image as the number of bar separations. After obtaining a signal with a difference in number, it generates an early warning and terminates the subsequent analysis process.

6. The image processing-based hot-rolled bar support separation detection method according to claim 2, characterized in that, The spatial fusion processing module selects the left / right end image as the anchor image and selects the corresponding overlapping center point coordinates from the image of the other end. If there are overlapping center point coordinates, they are recorded as overlapping samples; if there are no overlapping center point coordinates, they are recorded as difference samples. The spatial fusion processing module calculates the proportion of the number of differential samples in the total number of rod end faces in the left / right images to obtain the contrast difference rate. The spatial fusion processing module calculates the variance of the tilt angle of each bar end face on the left and right end face images.

7. The image processing-based hot-rolled bar support separation detection method according to claim 6, characterized in that, The spatial fusion processing module obtains the separation difference degree based on the comparison difference rate and tilt angle, and records the reciprocal of the separation difference degree as the separation uniformity degree.

8. The image processing-based method for detecting the fixed-support separation of hot-rolled bars according to claim 2, characterized in that, The separation comparison module performs a threshold judgment on the number of separated bars to determine whether the number meets the standard or the number is abnormal. The separation comparison module performs a threshold judgment on the uniformity of the bar separation to determine whether the uniformity meets the standard or the uniformity is abnormal. When the early warning decision module simultaneously obtains that the quantity and uniformity meet the standards, it generates a separation normal signal; when it simultaneously obtains that the quantity and uniformity are abnormal, it generates a low-risk early warning signal; and when it simultaneously obtains that the quantity and uniformity are abnormal, it generates a high-risk early warning signal.

Images