Visual counting monitoring system and method for section steel bales

Abstract

The invention discloses a visual counting monitoring system for section steel bales. The visual counting monitoring system comprises an industrial camera, an intelligent analysis module and an audible and visual alarm. Wherein the industrial cameras are respectively arranged at the F1 roller way and the F2 roller way and are used for respectively acquiring profile steel end face images at the F1 roller way and the F2 roller way, and the output ends of the industrial cameras are connected to the intelligent analysis module; and the intelligent analysis module is used for identifying the profile steel end face images at the F1 roller way and the F2 roller way based on a visual identification mode and driving the audible and visual alarm to work or not based on an identification result. Compared with a traditional artificial vision method and other methods, the method is low in cost and convenient to install, operate and maintain, all-weather operation can be achieved, and enterprises can be effectively helped to save the operation and maintenance cost.

Description

Technical Field

[0001] The invention relates to the field of machine vision detection and early warning of steel sections, and in particular to a visual counting monitoring system and method for steel section bundles. Background Art

[0002] Metallurgical enterprises have long used artificial vision to count the number of finished steel products, but artificial vision has disadvantages such as low efficiency, high cost and prone to false detection, and can no longer meet the growing demand for steel counting. More importantly, there are a large number of hoisting and packaging equipment in the finished product area, which poses potential harm to statisticians who work in the finished product area for a long time, and the company's operation and maintenance pressure is relatively high.

[0003] In recent years, with the development of machine vision technology and computing acceleration hardware, various machine vision systems and methods have played an important role in human production and life. Machine vision systems have the advantages of small size, convenient maintenance, all-weather and non-contact operation, which can effectively make up for the shortcomings of artificial vision and meet the actual needs of industrial production. Therefore, using industrial cameras to collect images of the finished steel area and designing visual algorithms to detect the number of steel sections will become a trend.

[0004] The prior art discloses an H-beam counting algorithm based on openCV, with the publication number CN115641310A. Although the patent discloses an H-beam counting algorithm, the patent focuses on counting the H-beams after bundling, and fails to provide early counting warning before the bundling operation occurs. In actual use, although the method disclosed can calculate the number of H-beams and issue a warning, the bundling operation of the steel has been completed at this time. If there is a problem, it needs to be "unpacked and bundled" again, and the warning value is relatively limited. Therefore, the algorithm still has defects when it cannot be applied to the counting and monitoring of steel bundles. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a visual counting monitoring system and method for steel bundles. The use of a machine vision system and method can effectively overcome the shortcomings of artificial vision, count the steel bundles in advance and issue abnormal alarms before they are bundled, effectively saving the company's operation and maintenance costs and reducing the impact of bundle abnormalities on subsequent secondary processing and sales.

[0006] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is: a visual counting monitoring system for steel bundles, including an industrial camera, an intelligent analysis module and an audible and visual alarm; wherein the industrial cameras are respectively arranged at the F1 roller and the F2 roller, and are used to collect the end surface images of the steel at the F1 roller and the F2 roller, respectively, and the output end thereof is connected to the intelligent analysis module; the intelligent analysis module is used to identify the end surface images of the steel at the F1 roller and the F2 roller based on a visual recognition method and drive the audible and visual alarm to work or not based on the recognition result.

[0007] The intelligent analysis module is built with a visual counting unit and a steel section falling off identification unit, which are used to respectively detect the number of steel sections in the steel section image and detect and determine whether there is abnormal falling off on the end surface of the steel section.

[0008] At least one industrial camera is arranged at the F1 roller table, and is arranged on the side of the F1 roller table to realize the shooting of the end face image of the steel section.

[0009] At least two industrial cameras are arranged at the exit of the F2 roller conveyor, and are arranged at both sides of the F3 roller conveyor respectively to capture the end surface images of the steel sections at both sides of the F2 roller conveyor.

[0010] The sound and light alarm is installed on site in the steel finished product area.

[0011] A monitoring method for a steel bundle visual counting monitoring system comprises the following steps:

[0012] Step 1: The industrial camera installed on the side of the F1 roller table waits for the shooting trigger signal. When the trigger signal is captured, the end face image of the steel section is captured and transmitted to the intelligent analysis module;

[0013] Step 2, the intelligent analysis module analyzes and gives the number of steel sections based on the visual counting method, and controls the triggering of the sound and light alarm based on the number of steel sections;

[0014] Step 3, the industrial cameras installed on both sides of the F2 roller table wait for the shooting trigger signal. When the trigger signal is captured, the end face images of the steel sections are simultaneously shot from the left and right sides of the F3 roller table. The intelligent analysis module analyzes whether the steel sections in the image are falling off based on the steel section falling off identification method, and controls the triggering of the sound and light alarm based on the steel section falling off condition.

[0015] The trigger signal of the industrial camera installed at the F1 roller conveyor is the PLC signal sent by the PLC controller when "F1 stopper falls, F1 roller conveyor starts".

[0016] The intelligent analysis module has pre-developed and saved a visual counting method, which is a steel section detection and counting method based on deep learning, and includes: collecting a set number of steel section end face images to produce a training set, marking the image ground-truth and training the deep learning model; then, deploying the trained model to the intelligent analysis module; when the module calls the visual counting method, it outputs the number of steel sections and the detection box in the image based on the model.

[0017] The trigger signal of the industrial camera installed at the F2 roller table is the PLC signal when "F2 guard frame rises and F2 roller table starts".

[0018] The steel section falling off identification method is a steel section detection method based on deep learning. This method determines whether the steel section has fallen off by detecting the image Y coordinate of each steel section detection frame; when the image Y coordinate of the steel section detection frame is smaller than the image Y coordinate of other steel section detection frames and reaches the set threshold, it is determined that the steel section has fallen off.

[0019] The advantages of the present invention are as follows: the use of machine vision systems and methods can effectively overcome the shortcomings of artificial vision, count in advance before steel bundles are bundled and issue abnormal alarms, effectively saving the company's operation and maintenance costs and reducing the impact of bundle abnormalities on subsequent secondary processing and sales. This solution has the advantages of small size, convenient maintenance, all-weather and non-contact operation. By collecting and analyzing the image features of steel bundles and designing a reasonable visual detection method for on-site deployment, it can effectively replace artificial vision to achieve accurate and efficient counting of steel, thereby saving operation and maintenance costs for the company and improving the quality of steel bundles. Using a visual counting system and method for steel bundles disclosed in the present invention, it is possible to promptly detect whether there is an abnormality in the quantity of a batch of steel before the bundling operation is performed; if there is an abnormality, an alarm is triggered in time to take measures to reduce the defective bundling rate.

[0020] Compared with traditional artificial vision methods and other methods, the present invention has low cost, is more convenient to install and maintain, and can operate around the clock, which can effectively help enterprises save operation and maintenance costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The following is a brief description of the contents expressed in the drawings of the present specification and the marks in the drawings:

[0022] Figure 1 It is a schematic diagram of the arrangement principle of the monitoring system of the present invention;

[0023] Figure 2 The figure is a flow chart of the monitoring method of the present invention.

[0024] The marks in the above figures are: 1. Industrial camera; 2. Intelligent analysis module; 3. Sound and light alarm. DETAILED DESCRIPTION

[0025] The specific implementation of the present invention will be further explained in detail below by describing the optimal embodiment with reference to the accompanying drawings.

[0026] The existing technology uses the method of manual visual counting, which has the disadvantage that manual vision working in the finished steel product area for a long time will inevitably have false detections and missed detections, and can no longer meet the growing demand for counting steel bundles. In response to this, the present application provides a method with the advantages of small size, convenient maintenance, all-weather and non-contact operation. By collecting and analyzing the image features of steel bundles and designing a reasonable visual detection method for on-site deployment, it can effectively replace manual vision to achieve accurate and efficient counting of steel, thereby saving operation and maintenance costs for enterprises and improving the quality of steel bundles. At the same time, it can realize counting and identification before bundling and give abnormal alarms, and at the same time identify the falling phenomenon of steel before bundling and issue an alarm. The specific plan is introduced as follows:

[0027] like Figure 1 , 2 As shown, a visual counting and monitoring system for steel bundles of the present application includes an industrial camera, an intelligent analysis module and an audible and visual alarm;

[0028] The industrial cameras are respectively arranged at the F1 roller and the F2 roller, and are used to collect the end surface images of the steel sections at the F1 roller and the F2 roller, respectively, and the output ends thereof are connected to the intelligent analysis module; the F1 roller and the F2 roller are components of the steel section bundling system, and are used to control bundling respectively, wherein the F1 roller is the odd-numbered layer grouping mechanism for steel section bundling, and the F2 roller is the even-numbered layer grouping mechanism for steel section bundling, and the odd-numbered and even-numbered layer grouping of bundling are controlled by the F1 roller and the F2 roller. Installing the recognition camera for identification at the steel section grouping position can accurately detect abnormalities and alarms in advance, and the operator can handle them quickly.

[0029] The intelligent analysis module is used to identify the end surface images of the steel sections at the F1 roller and the F2 roller based on visual recognition and to drive the sound and light alarm to work or not based on the recognition results.

[0030] The intelligent analysis module has built-in visual counting unit and steel section shedding recognition unit, which are used to detect the number of steel sections in the steel section image and to detect and judge whether there is abnormal shedding on the end face of the steel section. The visual counting unit is implemented by a visual counting algorithm, which is a steel section detection and counting method based on deep learning, including: collecting a set number of steel section end face images to make a training set, marking the image ground-truth and training the deep learning model; then, deploying the trained model in the intelligent analysis module; when the module calls the visual counting method, it outputs the number of steel sections and the detection frame in the image based on the model. The steel section shedding recognition unit is implemented by a steel section detection algorithm based on deep learning, which determines whether there is a shedding phenomenon by detecting the image Y coordinate of each steel section detection frame; when the image Y coordinate of the steel section detection frame is less than the image Y coordinate of other steel section detection frames and reaches the set threshold, it is judged that the steel section has fallen off.

[0031] At least one industrial camera is installed at the F1 roller, which is installed on the side of the F1 roller to capture the end face image of the steel section. Based on the end face image of the steel section obtained by taking a photo at this position, the intelligent analysis module calls the visual counting unit to identify the number of steel sections in the image, and then compares and judges based on the steel section data and the quantity threshold. When the number of steel sections identified exceeds the set threshold range, it is judged as an abnormal number of steel sections, and the sound and light alarm is driven to sound an alarm, thereby avoiding the defect of not discovering the abnormal number. Because the F1 roller is the starting position for the steel section bundle grouping, counting and statistics at this position can accurately detect abnormalities and alarms in advance, and operators can handle them quickly.

[0032] There are at least two industrial cameras installed at the exit of the F2 roller. For example, two cameras are installed on both sides of the F3 roller to capture the end face images of the steel sections on both sides of the F2 roller. Based on the end face images of the steel sections obtained by taking photos at this position, the intelligent analysis module calls the steel section shedding recognition unit to identify the steel section shedding in the image. When the steel section shedding abnormality is identified, the corresponding alarm signal is output and the sound and light alarm is driven to emit an alarm sound, so as to timely discover the steel section shedding phenomenon and alarm in time. Steel section shedding means that in the process of steel section packaging, even-numbered layers need to use magnetic suspension, and sometimes the magnetic suspension sometimes fails during the production process, causing one end of the steel section to fall off. Because the F2 roller is the starting point for the even-numbered layer grouping of the bundling, it is most appropriate to set it at the F2 roller when performing steel section shedding detection.

[0033] In order to make the alarm more timely and effective, the sound and light alarm and intelligent analysis module are installed at the finished steel product area, which is the bundling work site. When an abnormal alarm is automatically discovered, it can be handled in time to avoid the time-consuming and labor-intensive defects of unpacking after bundling. Reminders are issued before bundling and timely processing is carried out.

[0034] This solution is based on the monitoring method of the above-mentioned steel bundle visual counting monitoring system, and includes the following steps:

[0035] Step 1: The industrial camera installed on the side of the F1 roller table waits for the shooting trigger signal. When the trigger signal is captured, the end face image of the steel section is captured and transmitted to the intelligent analysis module;

[0036] Step 2, the intelligent analysis module analyzes and gives the number of steel sections based on the visual counting method, and controls the triggering of the sound and light alarm based on the number of steel sections;

[0037] Step 3, the industrial cameras installed on both sides of the F2 roller table wait for the shooting trigger signal. When the trigger signal is captured, the end face images of the steel sections are simultaneously shot from the left and right sides of the F3 roller table. The intelligent analysis module analyzes whether the steel sections in the image are falling off based on the steel section falling off identification method, and controls the triggering of the sound and light alarm based on the steel section falling off condition.

[0038] The trigger signal of the industrial camera installed at the F1 roller is the PLC signal sent by the PLC controller when "F1 stopper falls, F1 roller starts". The PLC controller is the control center of the entire roller operation. It is used to control the F1, F2 rollers and the entire steel bundling control. Therefore, the PLC controller needs to send a signal during the steel bundling process. The trigger control of the industrial camera can be achieved by obtaining the Kongzi signal in the PLC controller, and the purpose of interlocking control can be achieved. In the PLC-controlled steel bundling system, the F1 roller is the odd-numbered layer grouping mechanism for steel bundling, and the F2 roller is the even-numbered layer grouping mechanism for steel bundling. The PLC signal "F1 stopper falls, F1 roller starts" in the PLC controller indicates that the grouping of this steel bundling is completed and transmitted to the next control node. This signal can trigger the industrial camera to start counting.

[0039] The intelligent analysis module has pre-developed and saved a visual counting method, which is a steel section detection and counting method based on deep learning, and includes: collecting a set number of steel section end face images to produce a training set, marking the image ground-truth and training the deep learning model; then, deploying the trained model to the intelligent analysis module; when the module calls the visual counting method, it outputs the number of steel sections and the detection box in the image based on the model.

[0040] The trigger signal of the industrial camera installed at the F2 roller is the PLC signal when "F2 stopper rises, F2 roller starts". When the PLC signal in the process of PLC-controlled steel bundling is "F2 stopper falls, F2 roller starts", it means that the grouping of this steel bundling is completed and transported to the next control node. Therefore, it is necessary to detect the falling of the bundled steel at this time, triggering the industrial cameras on both sides of F2 to realize image acquisition and subsequent recognition, judgment and alarm.

[0041] The PLC controller performs bundling control based on the F1 and F2 rollers. For example, if a bundle of angle steels (8 layers, 5 in odd layers and 4 in even layers) needs to be bundled, the entire procedure is as follows: the F1 roller is grouped into 5 groups and transported to the first layer of the baler via the transport roller; the F2 roller is grouped into 4 groups and transported to the F3 roller position via the transport roller, and then lifted to the second layer of the baler by the magnetic crane; the above actions are repeated to complete the bundling of the steel sections. Therefore, industrial cameras are set on the corresponding F1 and F2 rollers to achieve the corresponding identification and detection purposes.

[0042] The steel section falling off identification method is a steel section detection method based on deep learning. This method determines whether the steel section has fallen off by detecting the image Y coordinate of each steel section detection frame; when the image Y coordinate of the steel section detection frame is smaller than the image Y coordinate of other steel section detection frames and reaches the set threshold, it is determined that the steel section has fallen off.

[0043] The present invention utilizes a machine vision system and method to effectively overcome the shortcomings of artificial vision, count in advance before steel bundles are bundled and issue abnormal alarms, effectively saving the company's operation and maintenance costs and reducing the impact of bundle abnormalities on subsequent secondary processing and sales.

[0044] The system of the present invention is composed of an industrial camera, an intelligent analysis module, and an audible and visual alarm. The industrial camera is installed on the side of the F1 roller and the side of the F2 roller; the intelligent analysis module is connected to the industrial camera, and the audible and visual alarm is connected to the intelligent analysis module and installed on site in the finished steel product area.

[0045] The industrial cameras installed on the sides of the F1 and F2 roller tables take pictures of the end faces of the steel sections according to the trigger signal. The intelligent analysis module saves the visual counting method and the steel section shedding identification method, which can detect the number of steel sections in the image and determine whether there is any abnormal shedding on the end faces of the steel sections. If there is an abnormality, the on-site sound and light alarm will be triggered; otherwise, it will remain silent. The attached figure is only a schematic diagram. In actual use, it should be modified and optimized according to the actual working conditions on site.

[0046] Step 1: The industrial camera installed on the side of the F1 roller table waits for the shooting trigger signal. When the trigger signal is captured, the end face image of the steel section is captured and transmitted to the intelligent analysis module.

[0047] Step 2: The visual counting method saved in the intelligent analysis module analyzes and gives the number of steel sections, and triggers the sound and light alarm as needed.

[0048] Step 3, the industrial cameras installed on both sides of the F2 roller table wait for the shooting trigger signal. When the trigger signal is captured, the end face images of the steel section are taken from both sides at the same time. The steel section shedding recognition method saved in the intelligent analysis module analyzes whether the steel section in the image has shedding, and triggers the sound and light alarm to alarm according to the situation;

[0049] Step 4: Analyze and give the number of steel sections using the visual counting method saved in the intelligent analysis module, and trigger the sound and light alarm as needed. Return to step 1. Steel section bundling is a repetitive process, so at this time it returns to step 4 and then feeds back to step 1.

[0050] In step 1, only one industrial camera is installed on the side of the F1 roller table, and the trigger signal refers to the PLC signal when "F1 stopper falls and F1 roller table starts".

[0051] In step 2, the visual counting method is a steel section detection and counting method based on deep learning. First, a certain number of steel section end face images are collected to make a training set, the image ground-truth is marked and the deep learning model is trained; then, the model is deployed in the intelligent analysis module; finally, when the module calls the visual counting method, the number of steel sections and the detection box in the image are output.

[0052] In step 3, an industrial camera is installed on both sides of the F2 roller. The trigger signal refers to the PLC signal when "F2 guard rises and F2 roller starts". This signal triggers the industrial cameras deployed on both sides of the F2 roller at the same time. The steel section falling off recognition method is a steel section detection method based on deep learning. The training set used by this method is consistent with the visual counting method, but this method adds a falling off recognition mechanism. The principle of this mechanism is: for steel sections that have not fallen off normally, the image Y coordinates of each steel section detection frame will be relatively close due to the fixed roller height. When a certain steel section falls off, the image Y coordinate of the steel section detection frame will be significantly smaller than the image Y coordinate of other steel section detection frames. Based on this, a difference threshold can be set to determine whether the steel section has fallen off.

[0053] like Figure 1 As shown, the system includes an industrial camera 1, an intelligent analysis module 2, and an audible and visual alarm 3. The industrial cameras 1 are installed on the sides of the F1 roller and the F2 roller respectively; each industrial camera 1 is equipped with an intelligent analysis module 2, which is connected to the industrial camera 1. Each intelligent analysis module 2 is equipped with an audible and visual alarm 3, which is connected to the intelligent analysis module 2.

[0054] As attached Figure 2 As shown, in step 1, an industrial camera is installed on the side of the F1 roller table, and the industrial camera is triggered by the PLC signal when "the F1 stopper falls and the F1 roller table starts".

[0055] Step 2: The visual counting method is a steel section detection and counting method based on deep learning. Collect the end face images of the steel section to make a training set, and use the Label Img tool to mark the image ground-truth to train the deep learning model; the model can be a series of target detection deep learning models such as YOLO and SSD. Deploy the trained model in the intelligent analysis module. When the image is transmitted to the intelligent analysis module, the module will call the visual counting method to detect the number of steel sections in the image and output the value and detection box.

[0056] Step 3, install an industrial camera on both sides of the F2 roller. The trigger signal refers to the PLC signal when "F2 guard rises and F2 roller starts". This signal will trigger the industrial cameras deployed on both sides of the F2 roller at the same time. The steel section falling off identification method is a steel section detection method based on deep learning. The training set used by this method is consistent with the visual counting method, but this method adds a falling off identification mechanism. The principle of this mechanism is: for steel sections that have not fallen off normally, the image Y coordinates of each steel section detection frame will be relatively close due to the fixed roller height. When a certain steel section falls off, the image Y coordinate of the steel section detection frame will be significantly smaller than the image Y coordinate of other steel section detection frames. Based on this, a difference threshold can be set to determine whether the steel section has fallen off. In specific implementation, the bubble sort method can be used to first sort the Y coordinates of each detection frame; then, starting from the detection frame with the largest Y value, it is compared with the detection frame with the smallest Y value in turn; if the absolute value of the difference between the Y value of a certain detection frame and the detection frame with the smallest Y value is greater than the threshold A, it means that the steel section corresponding to the detection frame has fallen off.

[0057] The specific implementation of step 4 is similar to step 2. After the counting and alarming functions are completed, the purpose of repeating the content of step 2 is that the steel bundle is a repetitive process, so the counting, identification and alarming also need to be repeated.

[0058] Obviously, the specific implementation of the present invention is not limited to the above-mentioned methods. As long as various non-substantial improvements are made using the method concept and technical solution of the present invention, they are all within the protection scope of the present invention.

Claims

1. A visual counting and monitoring system for steel bundles, characterized by: It includes an industrial camera, an intelligent analysis module and an audible and visual alarm; wherein the industrial cameras are respectively arranged at the F1 roller and the F2 roller, and are used to collect the end surface images of the steel sections at the F1 roller and the F2 roller, respectively, and the output end thereof is connected to the intelligent analysis module; the intelligent analysis module is used to identify the end surface images of the steel sections at the F1 roller and the F2 roller based on a visual recognition method and drive the audible and visual alarm to work or not based on the recognition result.

2. A visual counting and monitoring system for steel bundles as claimed in claim 1, characterized in that: The intelligent analysis module is built with a visual counting unit and a steel section falling off identification unit, which are used to respectively detect the number of steel sections in the steel section image and detect and determine whether there is abnormal falling off on the end surface of the steel section.

3. A visual counting and monitoring system for steel bundles as claimed in claim 1, characterized in that: At least one industrial camera is arranged at the F1 roller table, and is arranged on the side of the F1 roller table to realize the shooting of the end face image of the steel section.

4. A visual counting and monitoring system for steel bundles as claimed in claim 1, characterized in that: At least two industrial cameras are arranged at the exit of the F2 roller conveyor, and are arranged at both sides of the F3 roller conveyor respectively to capture the end surface images of the steel sections at both sides of the F2 roller conveyor.

5. The visual counting and monitoring system for steel bundles according to any one of claims 1 to 4, characterized in that: The sound and light alarm is installed on site in the steel finished product area.

6. A monitoring method for a visual counting monitoring system for steel bundles according to any one of claims 1 to 5, characterized in that: include Step 1: The industrial camera installed on the side of the F1 roller table waits for the shooting trigger signal. When the trigger signal is captured, the end face image of the steel section is captured and transmitted to the intelligent analysis module; Step 2, the intelligent analysis module analyzes and gives the number of steel sections based on the visual counting method, and controls the triggering of the sound and light alarm based on the number of steel sections; Step 3, the industrial cameras installed on both sides of the F2 roller table wait for the shooting trigger signal. When the trigger signal is captured, the end face images of the steel sections are simultaneously captured from the left and right sides of the F3 roller table. The intelligent analysis module analyzes whether the steel sections in the image have fallen off based on the steel section falling off identification method, and controls the triggering of the sound and light alarm based on the falling off situation of the steel sections.

7. The monitoring method of the visual counting monitoring system for steel bundles according to claim 6, characterized in that: The trigger signal of the industrial camera installed at the F1 roller conveyor is the PLC signal sent by the PLC controller when "F1 stopper falls, F1 roller conveyor starts".

8. The monitoring method of the visual counting monitoring system for steel bundles according to claim 6 or 7, characterized in that: The intelligent analysis module has pre-developed and saved a visual counting method, which is a steel section detection and counting method based on deep learning, and includes: collecting a set number of steel section end face images to produce a training set, marking the image ground-truth and training the deep learning model; then, deploying the trained model to the intelligent analysis module; when the module calls the visual counting method, it outputs the number of steel sections and the detection box in the image based on the model.

9. The monitoring method of the visual counting monitoring system for steel bundles according to claim 6 or 7, characterized in that: The trigger signal of the industrial camera installed at the F2 roller table is the PLC signal when "F2 guard frame rises and F2 roller table starts".

10. The monitoring method of the visual counting monitoring system for steel bundles according to claim 6 or 7, characterized in that: The steel section falling off identification method is a steel section detection method based on deep learning. This method determines whether the steel section has fallen off by detecting the image Y coordinate of each steel section detection frame. When the image Y coordinate of the steel section detection frame is smaller than the image Y coordinate of other steel section detection frames and reaches the set threshold, it is determined that the steel section has fallen off.

Images