Seamless Steel Pipe Piercing Plug Detection System Based on Machine Vision

Abstract

The present invention belongs to the technical field of plug detection, and particularly relates to a seamless steel pipe piercing plug detection system based on machine vision. The present invention provides a new seamless steel pipe piercing plug detection system based on machine vision. By using machine vision technology, an optoelectronic switch and a camera assembly are arranged at the entrance of the piercing mill to collect the image of the mandrel, and the image processing unit uses a deep neural network algorithm to judge whether the plug exists or determine whether the connection gap between the plug and the mandrel is abnormal. At the same time, according to the detected result signal, the mandrel is controlled to stop suddenly. This setting detects the plug based on the machine vision method, which has the advantages of high accuracy and fast speed.

Description

Technical Field

[0001] The present invention belongs to the technical field of plug detection, and particularly relates to a seamless steel pipe piercing plug detection system based on machine vision. Background Art

[0002] During the production process of seamless steel pipes, the billet is pierced by a plug carried by a mandrel under the action of force at the piercing mill to form a hollow shell. The plug and the mandrel are connected together via a connecting device. When the plug moves towards the rolling mill opening, it will experience an acceleration and deceleration process; if the plug and the mandrel are not tightly connected, during the deceleration process, due to inertia, the plug is likely to fall off; if the plug falling off is not detected in time before the mandrel enters the piercing mill, the mandrel directly contacts the billet, which is likely to cause damage to the mandrel, resulting in unnecessary economic losses and even serious production accidents.

[0003] In the traditional production process, the connection state between the plug and the mandrel is checked manually. Manual inspection has the following problems: 1) The mandrel runs at a high speed, making it difficult for manual inspection to detect in time; 2) Workers are in a long-term tense state, and manual observation is prone to visual fatigue, resulting in missed inspections.

[0004] Meanwhile, due to the complex piercing process environment and many mechanical structures, the background of the captured mandrel image is complex, and plug detection is easily interfered by the environment; the typical temperature of the pierced billet is about 1200 degrees Celsius, and the temperature is high. In order to protect the mandrel, it is necessary to spray water on the mandrel to cool it down. Therefore, a large amount of water vapor will be formed, making image detection difficult; secondly, the mandrel runs at a high speed, and the system must complete image acquisition and output the detection result within a very short time (0.1 - 0.2 s); furthermore, even if the plug and the mandrel are in a connected state, but if the connection is not tight, when the piercing is completed and the plug exits, there is also a risk that the plug will fall off and remain in the hollow shell. Summary of the Invention

[0005] In view of the above problems, the present invention provides a new seamless steel pipe piercing plug detection system based on machine vision.

[0006] The specific technical solution of the present invention is as follows:

[0007] The present invention provides a seamless steel pipe piercing plug detection system based on machine vision, including a trigger component, a camera component, a light source, and an image processing unit;

[0008] The trigger component is arranged on one side of the entrance of the piercing mill, and is configured to detect whether there is a mandrel passing through the expected movement trajectory, and send a trigger signal to the camera component when there is a mandrel passing through;

[0009] The camera assembly is disposed above the entrance of the piercing machine, and is configured to control the light source to light up after receiving a trigger signal, delay and wait for the light source to reach the expected shooting brightness, then capture an image of the mandrel, and send the captured mandrel image to the image processing unit. The light source is disposed on the camera assembly;

[0010] The image processing unit is configured to detect and identify the image using a deep neural network algorithm, determine whether the top head exists or determine whether the connection gap between the top head and the mandrel is abnormal. When the top head does not exist or the connection gap is abnormal, an alarm is issued, and a result signal for emergency stop of the mandrel is sent to the piercing machine.

[0011] The beneficial effects achieved by the present invention:

[0012] The present invention provides a new seamless steel pipe piercing top head detection system based on machine vision. Using machine vision technology, an optoelectronic switch and a camera assembly are set at a distance from the entrance of the piercing machine to collect mandrel images, and the image processing unit uses a deep neural network algorithm to determine whether the top head exists or determine whether the connection gap between the top head and the mandrel is abnormal. At the same time, according to the detected result signal, the emergency stop of the mandrel is controlled. This setting detects the top head based on the machine vision method, and has the advantages of high accuracy and fast speed. Description of the Drawings

[0013] Figure 1 is a structural block diagram of the seamless steel pipe piercing top head detection system based on machine vision in the present invention;

[0014] Figure 2 is a picture where the top head may exist in the piercing top head detection system of the present invention;

[0015] Figure 3 is a front view of the seamless steel pipe piercing top head detection system based on machine vision in the present invention;

[0016] Figure 4 is an axonometric view of the seamless steel pipe piercing top head detection system based on machine vision in the present invention;

[0017] Figure 5 is a top view of the seamless steel pipe piercing top head detection system based on machine vision in the present invention. Detailed Embodiments

[0018] The present invention will be further described below in conjunction with the drawings and embodiments. The following embodiments are only used to explain the content of the present invention and are not used to limit the protection scope of the present invention.

[0019] The present invention provides a seamless steel pipe piercing top head detection system based on machine vision, as Figures 1 - 5As shown in the figure, it includes a trigger component 1, a camera component 2, a light source 3, and an image processing unit 4; the image processing unit is disposed within a processor, wherein the trigger component and the camera component are both in communication with the processor;

[0020] The trigger component 1 is a photoelectric switch, which is disposed on one side of the entrance of the piercing machine and is configured to detect whether there is a mandrel 5 passing through on the expected movement trajectory, and send a trigger signal (i.e., a photoelectric signal) to the camera component when there is a mandrel passing through;

[0021] The camera component 2 uses a high-definition CCD sensor, which is disposed above the entrance of the piercing machine and is configured to control the light source 3 to light up after receiving the trigger signal, and delay waiting for the light source 3 to reach the expected shooting brightness and then shoot an image of the mandrel, and send the captured mandrel image to the image processing unit 4 through a gigabit network. The light source 3 is disposed on the camera component 2; after the high-definition CCD sensor receives the photoelectric signal sent by the processor, it controls the light source to light up;

[0022] Using gigabit Ethernet to convert to gigabit optical fiber for image transmission, it has the characteristics of fast transmission speed, long transmission distance, and strong anti-interference ability;

[0023] The image processing unit 4 is configured to use a deep neural network algorithm to detect and identify the image, determine whether the top head exists or determine whether the connection gap between the top head and the mandrel is abnormal, and issue an alarm when the top head does not exist or the connection gap is abnormal, and send a result signal of mandrel emergency stop to the piercing machine.

[0024] In addition, in the present invention, the image processing unit is equivalent to a host for processing data, and it is also in communication with a display. The display real-time displays the captured image and the image processing process (such as Figure 2 shown).

[0025] The present invention provides a new seamless steel pipe piercing top head detection system based on machine vision. It adopts machine vision technology, sets a photoelectric switch and a camera component at the entrance of the piercing machine to collect mandrel images, and uses a deep neural network algorithm by the image processing unit to determine whether the top head exists or determine whether the connection gap between the top head and the mandrel is abnormal. At the same time, it controls the mandrel to stop urgently according to the detected result signal. This setting detects the top head based on the machine vision method, and has the advantages of high accuracy and fast speed.

[0026] In this embodiment, the camera component performs camera calibration before working, and calibrates the internal parameters of the camera to correct the distortion of the camera.

[0027] In this embodiment, the image processing unit is further configured to, before detecting and recognizing an image, use the natural boundary of the image width as the X variable and the minimum and maximum specifications of the top head as the natural boundary of the variable D, and establish a mapping relationship L = F(X, D) between different specifications D, unit pixel X, and physical length L by using a two-dimensional interpolation method.

[0028] In actual production, there are generally only a fixed number of specifications (represented by the outer diameter D) for the top head and the top rod available for production according to demand. Changing the production specification will change the physical distance between the top head and the top rod and the camera, and thus change the camera field of view. Since the image resolution is fixed, changing the specification will change the physical length corresponding to a single pixel. To accurately measure the physical width of the gap, it is necessary to establish a mapping relationship from the X coordinate of the image pixel to the physical length under different specifications D.

[0029] In this embodiment, the specific establishment of the mapping relationship between different specifications D, unit pixel X, and physical length adopts a bicubic interpolation method, which specifically includes the following parts:

[0030] Preparation step: Fix a scale with a suitable thickness and a length sufficient to cover the camera field of view below the camera field of view. The vertical distance from the scale to the camera is not less than the vertical distance from the minimum specification top head to the camera during production, and record the specification D. 1 ;

[0031] First sampling step: Select a horizontal baseline Y on the image so that it can fall on the scale; take M (M≥4) pixel points between the 1st pixel and the Xth pixel along the baseline as sampling points, and record the scale value corresponding to each sampling point.

[0032] Second sampling step: Lift the scale and record the current vertical distance from the scale to the camera as the specification D. 2 and collect M sample points under D 2 according to the method described in the first sampling step;

[0033] Total sample point step: Repeat the second sampling step and collect a total of N groups of sample points under different specifications D i (i = 1,..., N), and D N can cover the distance from the maximum specification top head to the camera;

[0034] Zero calibration step: Perform in-group zero calibration on each group of sampling points, specifically, subtract the scale value of the first sampling point in the group from the scale value of each sampling point in the group.

[0035] Interpolation step: Use all sampling points for interpolation.

[0036] In this embodiment, the image processing unit includes a punch head detection module, which is configured to use a deep neural network algorithm to detect an image and determine whether a punch head exists. If it does not exist, an alarm is issued, and a result signal for immediately stopping the punch rod is sent to the punching machine.

[0037] In this embodiment, the image processing unit includes a gap detection module, which is configured to, when a punch head is detected, determine whether the connection gap between the punch head and the punch rod is greater than a threshold value based on a segmentation algorithm of a fully convolutional neural network, and perform processing according to the determination result, including an encoding stage and a decoding stage, which specifically include the following parts:

[0038] In the encoding stage, when the punch head detection module detects a punch head, an ROI image is generated by taking X pixels to the left and right respectively with the right boundary of the punch head bounding box as the center, the coordinates of the ROI image in the original image are recorded, and the ROI image features are extracted by inputting them into a deep convolutional neural network.

[0039] In the decoding stage, pixel-level classification is performed using the extracted ROI image features to predict whether each pixel is a gap, and a binary map of the same size as the ROI image is predicted, where the pixel value 1 represents a gap and 0 represents the background.

[0040] Result processing step: If there is a gap in the ROI, the ROI is mapped to the coordinates in the original image, the pixel width of the gap is calculated, and the gap width is calculated in combination with the pre-established mapping relationship between unit pixels and physical lengths. If the calculated gap is greater than the threshold value, an alarm is issued, and a result signal for immediately stopping the punch rod is sent to the punching machine.

[0041] In this embodiment, the calculation of the pixel width adopts an interpolation method, which includes the following parts:

[0042] The left edge pixel XL and the right edge pixel XR of the gap are taken along the baseline Y.

[0043] Based on the left edge pixel XL and the right edge pixel XR, the pixel width L is calculated, L = L(XR, D) - L(XL, D)

[0044] In the present invention, image detection must complete two-stage determination. The first stage determines whether a punch head exists; the second stage determines the connection gap between the punch head and the punch rod. This method can accurately determine the connection state between the punch head and the punch rod, so as to stably detect the presence or absence of the punch head and accurately measure the connection gap between the punch head and the punch rod in a complex environment.

[0045] Although this specification contains many specific implementation details, these should not be construed as limiting the scope of any invention or the scope of what can be claimed, but rather as illustrations of features that may be specific to particular embodiments of a particular invention. The specific features described in this specification in the context of separate embodiments may also be implemented in combination in a single embodiment. Conversely, the various features described in the context of a single embodiment may also be implemented independently in multiple embodiments, or in any suitable sub-combination. Additionally, although features may be described above as acting in combination and even initially claimed as such, one or more features from a claimed combination may in some cases be excluded from that combination, and the claimed combination may be directed to a sub-combination or variation of a sub-combination.

[0046] Particular embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. For example, the acts recited in the claims may be performed in a different order and still achieve the desired result. As one example, in order to achieve the desired result, the processes described in the figures do not necessarily require the particular order or sequential order shown. In certain implementations, multitasking and parallel processing may be advantageous.

Claims

1. A seamless steel pipe piercing plug detection system based on machine vision, characterized in that, it includes a trigger component, a camera component, a light source, and an image processing unit; The trigger component is arranged on one side of the entrance of the piercing mill and is configured to detect whether there is a mandrel passing on the expected movement trajectory, and send a trigger signal to the camera component when there is a mandrel passing; The camera component is arranged above the entrance of the piercing mill and is configured to control the light source to light up after receiving the trigger signal, and delay waiting for the light source to reach the expected shooting brightness and then shoot an image of the mandrel, and send the shot mandrel image to the image processing unit, and the light source is arranged on the camera component; The image processing unit is configured to use a deep neural network algorithm to detect and identify the image, judge whether the plug exists or determine whether the connection gap between the plug and the mandrel is abnormal, issue an alarm when the plug does not exist or the connection gap is abnormal, and send a result signal of mandrel emergency stop to the piercing mill; The camera component performs camera calibration before working, and calibrates the internal parameters of the camera to correct the distortion of the camera; The image processing unit is further configured to, before detecting and identifying the image, take the width of the image as the natural boundary of the X variable, take the minimum specification and the maximum specification of the plug as the natural boundary of the variable D, and use the two-dimensional interpolation method to establish the mapping relationship L = F(X,D) between different specifications D, unit pixel X and physical length L.

2. The seamless steel pipe piercing plug detection system based on machine vision according to claim 1, characterized in that, The specific establishment of the mapping relationship between different specifications D, unit pixel X and physical length adopts the bicubic interpolation method, which specifically includes the following parts: Preparation step: Fix a scale ruler with appropriate thickness and sufficient length to cover the camera's field of view below the camera's field of view. The vertical distance from the scale ruler to the camera is not less than the vertical distance from the smallest specification tip to the camera during production, and record the specification D 1 ; The first sampling step, select a horizontal baseline Y on the image so that it can fall on the scale; take M (M≥4) pixel points between the 1st pixel and the Xth pixel along the baseline as sampling points, and record the scale value corresponding to each sampling point; Second sampling step: Lift the scale ruler and record the vertical distance from the current scale ruler to the camera as specification D 2 , and collect M sample points under D 2 by the method described in the first sampling step; Total sample point step, repeat the second sampling step, and total N groups of sample points under different specifications D i (i = 1,..., N), and D N can cover the distance from the largest specification plug to the camera; The zero calibration step, perform in-group zero calibration on each group of sampling points, specifically, subtract the scale value of the first sampling point in the group from the scale value of each sampling point in the group; The interpolation step, use all sampling points for interpolation.

3. The seamless steel pipe piercing plug detection system based on machine vision according to claim 1, characterized in that, The image processing unit includes a plug detection module, and the plug detection module is configured to use a deep neural network algorithm to detect the image and judge whether the plug exists. If not, an alarm is issued and a result signal of mandrel emergency stop is sent to the piercing mill.

4. The seamless steel pipe piercing plug detection system based on machine vision according to claim 3, characterized in that, The image processing unit includes a gap detection module, and the gap detection module is configured to, when a plug is detected, judge whether the connection gap between the plug and the mandrel is greater than the threshold based on the segmentation algorithm of the fully convolutional neural network, and process according to the judgment result, including an encoding stage and a decoding stage, which specifically includes the following parts: In the encoding stage, when the head detection module detects the head, an ROI image is generated by taking X pixels to the left and right respectively with the right boundary of the head bounding box as the center, and the coordinates of the ROI image in the original image are recorded and input into the deep convolutional neural network to extract the ROI image features; In the decoding stage, pixel-level classification is performed using the extracted ROI image features to predict whether each pixel is a gap, and a binary image of the same size as the ROI image is predicted, where the pixel value 1 represents a gap and 0 represents the background; In the result processing step, if there is a gap in the ROI, the ROI is mapped to the coordinates in the original image, the pixel width of the gap is calculated, and the gap width is calculated by combining the pre-established mapping relationship between the unit pixel and the physical length. If the calculated gap is greater than the threshold, an alarm is issued and a result signal for the ejector rod to stop urgently is sent to the perforator.

5. The seamless steel pipe piercing head detection system based on machine vision according to claim 4, characterized in that, The calculation of the pixel width adopts an interpolation method, including the following parts: Take the left edge pixel XL and the right edge pixel XR of the gap along the baseline Y; Calculate the pixel width L based on the left edge pixel XL and the right edge pixel XR, L = L(XR, D) - L(XL, D).

6. The seamless steel pipe piercing head detection system based on machine vision according to claim 1, characterized in that, The ejector rod image is transmitted into the image processing unit through the gigabit network.

7. The seamless steel pipe piercing head detection system based on machine vision according to claim 1, characterized in that, The camera component adopts a high-definition CCD sensor.

8. The seamless steel pipe piercing head detection system based on machine vision according to claim 1, characterized in that, The trigger component is a photoelectric switch, and the trigger signal is a photoelectric signal.

Images