A real-time video monitoring intelligent analysis alarm system and method for stream inoculant

Abstract

The present application relates to the field of steel processing and artificial intelligence cross technology, and specifically discloses a real-time video monitoring intelligent analysis alarm system and method for stream following inoculant. The real-time video monitoring intelligent analysis alarm system for stream following inoculant realizes real-time monitoring of the flow line in the pouring stage of casting production, positioning of the molten steel pouring area and the inoculant inoculation area, and identification of the molten steel in the molten steel pouring area and the inoculant in the inoculant inoculation area. Then, the real-time state is classified according to the identification results of the molten steel and the inoculant, and an alarm is given when the state is determined to be abnormal, reminding the staff to handle it in time. The present application not only solves the problem of defective parts caused by uneven inoculation of inoculant in the existing pouring technology in the casting production process, but also solves the problem that some castings cannot be inoculated due to the blockage of the inoculant nozzle in the existing pouring technology in the casting production process, thereby reducing the workload of the staff, improving the production qualification rate of the castings, and reducing the production loss of the castings.

Description

Technical Field

[0001] This invention belongs to the interdisciplinary field of steel processing and manufacturing and artificial intelligence, and includes a real-time video monitoring, intelligent analysis and alarm system and method for inoculants. Background Technology

[0002] The casting production process is generally divided into six stages: mold processing, sand mixing, molding (core making), smelting, pouring, and sand removal. Among them, the pouring stage is a crucial link in the entire casting production process. In this stage, molten metal material is introduced into a sand box with a well-adjusted mold to guide its formation.

[0003] To further improve the quality of castings, inoculants need to be added to the molten metal during the pouring stage. Adding an appropriate amount of inoculant can control the temperature, gas content, and microstructure of the molten steel, thereby improving the performance of the castings. This process is called in-flow inoculation.

[0004] In-flow inoculation is a commonly used inoculation technology. It involves adding the inoculant along with the molten steel during pouring, using a manual funnel or automatic induction device, to complete the inoculation process. In-flow inoculation can effectively control the uniformity of the inoculant distribution in the molten steel. However, in the casting production process, the metering of the inoculant used is difficult to control, and the temperature of the molten steel and the pouring time are hard to manage. This often results in the inoculant melting too early, preventing subsequent pours from being inoculated and causing uneven inoculation. Incomplete melting of the inoculant, leading to insufficient inoculation, premature melting, or incomplete melting, all result in inconsistent casting quality, and in severe cases, scrap and wasted inoculant. Furthermore, during frequent inoculation processes, blockage of the inoculant nozzle frequently occurs. If not detected in time, this can lead to insufficient inoculation in subsequent castings, failure to solidify, and other problems, severely reducing overall production efficiency and causing economic losses. Summary of the Invention

[0005] The purpose of this invention is to address the problems of insufficient and uneven inoculation during casting production. A real-time video monitoring and intelligent analysis alarm system and method for inoculating the inoculant is designed. Real-time video data acquisition of the target area of ​​the inoculant is achieved using image acquisition equipment. The real-time video monitoring and intelligent analysis alarm system for inoculating the inoculant is installed on the controller, enabling real-time monitoring of the casting production line; locating the molten steel pouring area and the inoculant inoculation area in the monitoring image of the pouring area; identifying and analyzing the state of molten steel in the molten steel pouring area and the inoculant inoculation area; and determining and alarming abnormal states during the pouring process.

[0006] To address the problems existing in the prior art, the present invention adopts the following technical solution:

[0007] A real-time video monitoring, intelligent analysis, and alarm system for in-flow inoculants includes an image acquisition device, an industrial controller, and software for the real-time video monitoring, intelligent analysis, and alarm system for in-flow inoculants embedded in the industrial controller.

[0008] Image acquisition equipment is used to monitor and acquire the status of the casting machine, molten steel, in-flow inoculant, and in-flow inoculant nozzle;

[0009] The industrial controller includes an audible and visual alarm 8 and an industrial computer 9; the industrial computer 9 is used to run the real-time video monitoring and intelligent analysis alarm system software for in-flow inoculants; when the real-time video monitoring and intelligent analysis alarm system software for in-flow inoculants transmits an alarm signal to the audible and visual alarm 8, the audible and visual alarm 8 executes the alarm event.

[0010] The software for the real-time video monitoring, intelligent analysis, and alarm system for inoculated progesterone includes a monitoring module, a positioning module, an identification module, an analysis module, and an alarm module.

[0011] The monitoring module is used to monitor the casting production line in real time.

[0012] The positioning module is used to receive the location information of the pouring area in the casting production line screen; the positioning module finds and marks the position of the pouring machine and the position of the inoculant nozzle in the monitoring image;

[0013] The identification module is used to receive the molten steel pouring position and inoculant inoculation position information from the positioning module, and then perform molten steel identification within the molten steel pouring position and inoculant identification within the inoculant inoculation position;

[0014] The analysis module is used to perform steel condition analysis and inoculant condition analysis. The steel condition includes two states: "in pouring" and "not poured"; the inoculant condition includes two states: "in inoculation" and "not inoculated".

[0015] When the molten steel is in the "pouring" state and the inoculant is in the "not inoculated" state, it is determined to be an abnormal state of "inoculant nozzle blockage".

[0016] When the probiotic is in a "probiotic insufficiency" state, it is determined to be an "insufficient probiotic" abnormal state;

[0017] The alarm module is used to trigger an alarm after determining an abnormal state; when any abnormal state occurs, an alarm event is triggered, an alarm signal is issued, and the audible and visual alarm 8 is invoked to execute the audible and visual alarm event.

[0018] Preferably, in the positioning module, the location information of the pouring area includes "in the pouring area" and "not in the pouring area";

[0019] When in the "in the pouring area" state, the target tracking method is used to track and mark the positions of the pouring machine and the inoculant nozzle in the pouring area. When either the pouring machine or the inoculant nozzle leaves the pouring area, the tracking stops and the "not in the pouring area" signal is output.

[0020] When in the "in the pouring zone" state, determine and mark the molten steel pouring position and the inoculant inoculation position.

[0021] Preferably, the recognition module is also used to separately count the pixel area of ​​the molten steel recognition result and the inoculant recognition result, and establish a coordinate system with real time to obtain the real-time curve of "molten steel area - time" and the real-time curve of "inoculant area - time" respectively.

[0022] Preferably, the analysis module is also used to analyze and obtain the real-time curve of "inoculant cycle - inoculant dosage" based on the real-time curves of "molten steel area - time" and "inoculant area - time".

[0023] Preferably, in the alarm module, an alarm signal is output when 25 consecutive frames of abnormal status signals are received from the analysis module; if the accumulated abnormal status signals do not reach 25 consecutive frames, no alarm signal is output.

[0024] Preferably, the software for the real-time video monitoring, intelligent analysis, and alarm system for in-flow inoculants also includes a data management module;

[0025] The data management module is used to store screenshots, time, production line, type, and remarks when an alarm event occurs.

[0026] Preferably, the image acquisition device includes a camera 5 and a sealing cover. The camera 5 is positioned parallel to the casting production line. The camera 5 is installed in the sealing cover, and a glass lens is installed in the front opening of the sealing cover. A sealed connection is formed between the sealing cover and the glass lens.

[0027] Preferably, an air-cooled cooler 7 is fixed at the rear end inside the sealed cover, and the camera 5 is fixed at the front end of the air-cooled cooler 7.

[0028] Preferably, the method includes the following steps:

[0029] S1. Data Acquisition; Real-time raw flow data of the casting production line is acquired using image acquisition equipment;

[0030] S2. Real-time data is collected and processed through the monitoring module;

[0031] S3. Calibrate the pouring area; calibrate the pouring area on the monitoring screen;

[0032] S4. Determine whether the pouring machine and inoculant nozzle are located within the pouring area; after locating and marking the positions of the pouring machine and inoculant nozzle in the monitoring image, output the position status signal;

[0033] S5. Target tracking of the casting machine and inoculant nozzle; output position status signal;

[0034] S6. Positioning of molten steel and inoculant; marking the molten steel pouring position and inoculant inoculation position according to the position status signal;

[0035] S7. Identify molten steel and inoculant in the positioning area; use image masking during identification to mask the inoculant area when identifying molten steel; mask the molten steel area when identifying inoculant.

[0036] S8. Statistical analysis and visualization of identification results; plotting real-time curves of "time-molten steel area" and "time-inoculant area";

[0037] S9. Results data analysis; determining the state of molten steel and inoculant based on threshold values;

[0038] S10. Abnormal status data filtering and alarm; an alarm event is triggered when the molten steel and inoculant are in abnormal states.

[0039] S11. Abnormal status data management, abnormal data review.

[0040] Preferably, in step S7, a de-flaming algorithm is used in the molten steel identification. The de-flaming algorithm collects the molten steel detection results of the current frame and the molten steel detection results of the previous frame, and performs an image AND operation on the two results to extract the common molten steel area in the two frames, thereby achieving the de-flaming effect.

[0041] The beneficial effects of this invention are as follows:

[0042] 1. In this invention, users can use a real-time video monitoring and intelligent analysis alarm system and method for inoculants to observe the real-time status of molten steel and inoculants during the production process.

[0043] 2. In this invention, users can promptly detect abnormal conditions of the inoculant through real-time monitoring and alarm of a real-time video monitoring and intelligent analysis alarm system and method for inoculant, alerting users to handle the situation in a timely manner, reducing the workload of staff, improving the production qualification rate of castings, and reducing production losses of casting products.

[0044] 3. In this invention, the positioning module uses a template matching algorithm and a target tracking algorithm to determine the location of molten steel and the inoculant in the image, thereby achieving rapid positioning of the molten steel and inoculant targets.

[0045] 4. In this invention, the use of an inoculant detection template solves the problem of mutual interference caused by the close proximity of molten steel and inoculant.

[0046] 5. In this invention, a steel and inoculant detection algorithm is used to solve the problem of misidentifying the flames around the molten steel as molten steel, thereby improving the accuracy of the steel and inoculant identification results.

[0047] 6. In this invention, the intelligent algorithm used identifies molten steel and inoculant. The overall method is efficient and rapid, achieving millisecond-level detection and clear and smooth identification of video stream data.

[0048] 7. In this invention, a delayed judgment method is used to determine the state of molten steel and inoculant, which reduces the false alarm rate and improves the overall accuracy of the identification results.

[0049] 8. In this invention, an audible and visual alarm method is used, which is suitable for factory areas with high noise and complex environments, and can intuitively notify users to handle alarm events in a timely manner.

[0050] 9. In this invention, a stainless steel sealing cover is used to seal the zoom camera, the air-cooled refrigerator, and the glass lens. This protects the vulnerable components inside the stainless steel sealing cover from accidental impacts and also protects the camera lens from damage caused by splashing hot molten steel.

[0051] 11. In this invention, an air-cooled refrigerator is used, which allows compressed air to circulate and achieve a cooling effect through the interlayer ventilation, and forms an air curtain in front of the camera, giving the image acquisition equipment the ability to withstand high temperatures and prevent dust.

[0052] 12. This invention provides an intelligent interpretation method for real-time production status of core process locations during the pouring stage of casting production. It solves the problem of defective parts caused by uneven inoculant development in existing pouring technologies, and the problem of incomplete inoculation in some castings due to inoculant nozzle blockage. This reduces the workload of construction personnel and improves the production qualification rate of castings. The template matching method solves the problem of difficult target positioning of molten steel and inoculant; the intelligent interpretation algorithm solves the problem of difficulty in separating molten steel, inoculant, and background; and the delayed judgment mode improves algorithm accuracy and reduces false alarm rate, achieving real-time tracking and monitoring of the inoculant status during casting production.

[0053] 13. In the production of certain special types of castings, the inoculant used may react violently with the molten steel, resulting in a large number of flames at the inoculation site. This may lead to some flames being mistaken for molten steel, affecting subsequent processes. The flame removal algorithm used in this invention achieves flame removal through post-processing of the molten steel detection results, effectively solving the problem of flames in the molten steel area affecting molten steel identification and ensuring the stability of the molten steel identification results. Attached Figure Description

[0054] Figure 1 An external view of an embodiment of the automatic analysis and alarm system for in-flow probiotics according to the present invention is shown.

[0055] Figure 2 The image acquisition area of ​​an embodiment of an automatic analysis and alarm system for in-flow probiotics according to the present invention is shown in the left view.

[0056] Figure 3 A logic diagram of the positioning module of an embodiment of an automatic analysis and alarm system for in-flow probiotics according to the present invention is shown.

[0057] Figure 4 A logic diagram of the alarm module of an embodiment of an automatic analysis and alarm system for in-flow probiotics according to the present invention is shown.

[0058] Figure 5 A flowchart illustrating the actual operation of an embodiment of the automatic analysis and alarm system for in-flow probiotics according to the present invention is shown.

[0059] Figure 6 An example of a progesterone "area-time" curve is shown in the data from an embodiment of an automatic progesterone analysis and alarm system according to the present invention.

[0060] Figure 7 An example of a "area-time" curve of molten steel from data in an embodiment of an automatic analysis and alarm system for inoculants according to the present invention is shown.

[0061] Figure 8(a) shows the steel detection results without using the descaling algorithm in the data of an embodiment of the automatic analysis and alarm system for inoculants in the flow according to the present invention.

[0062] Figure 8(b) shows the steel detection results using the descaling algorithm in the data of an embodiment of the automatic analysis and alarm system for inoculants in the flow according to the present invention.

[0063] In the diagram: 1. Inoculant pouring area; 2. Pouring machine working area; 3. Inoculant production location; 4. Production conveyor belt; 5. Camera; 6. Stainless steel sealing cover; 7. Air-cooled refrigerator; 8. Audible and visual alarm; 9. Industrial control computer; 10. System input device; 11. LCD display; 12. Inoculant storage silo; 13. Inoculant discharge pipe; 14. Pouring machine. Implementation

[0064] The present invention will be further described below with reference to the embodiments and accompanying drawings.

[0065] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described below with reference to the accompanying drawings and embodiments: Example

[0066] like Figure 1 As shown in the figure, this embodiment discloses the hardware part of a real-time video monitoring and intelligent analysis alarm system for inoculated progenitors, including an image acquisition device and an industrial controller.

[0067] The image acquisition equipment includes a 400W pixel variable-focus high-definition camera 5 installed beside the production conveyor belt 4, used to acquire images within the monitored area (inoculant pouring area 1, pouring machine working area 2, and inoculant production position 3); an air-cooled refrigerator 7 to reduce the operating temperature of the camera and ensure its normal operation; and a stainless steel sealed cover 6 to protect the camera 5 from accidental collisions, dust accumulation, and other factors that could affect its normal use. The image acquisition equipment can be installed and fixed by wall mounting, hanging, or pole mounting.

[0068] The stainless steel sealing cover 6 is cylindrical in shape, with dimensions of 360mm in length, 140mm in width, and 110mm in height. In this embodiment, a bracket is used to mount the sealing cover and equipment such as the camera. A fixed bracket is attached to the upper end of the stainless steel sealing cover 6, through which the image acquisition equipment is installed on one side of the casting production line. The installation position of the image acquisition equipment allows the camera to capture clear images of the casting machine, molten steel, inoculant, and inoculant nozzle on the production line.

[0069] The camera 5 and the air-cooled cooler 7 are placed inside a stainless steel sealed cover, with the camera 5 positioned above the air-cooled cooler 7. When the camera 5 is in operation, the air-cooled cooler 7 delivers compressed air to form a circulating airflow within the stainless steel sealed cover 6 to achieve a cooling effect. At the same time, it removes dust from the stainless steel sealed cover 6 through the air outlet, giving the image acquisition equipment high temperature resistance and dust prevention capabilities.

[0070] Figure 2The image acquisition device's acquisition area is shown in a left view. This area includes a conveyor belt 4 storing casting molds, a casting machine 15 storing molten steel, a storage tank 13 storing inoculant, and an inoculant nozzle 14 responsible for inoculant flow. Whenever the conveyor belt 4 moves to the corresponding inoculant production position 3, molten steel is released, along with the inoculant. During video acquisition, it should be ensured that… Figure 2 The casting machine, molten steel, inoculant feeding pipe, and inoculant can be clearly photographed without being obstructed by any objects.

[0071] like Figure 1 The industrial controller includes an industrial computer 9 for operating the real-time video monitoring, intelligent analysis, and alarm system for the inoculant; an audible and visual alarm 8 for executing alarm events issued by the system; an LCD display 11 for displaying the monitoring screen and the human-machine interface of the system; and a system input device 10 for transmitting system operation commands to the operator.

[0072] In this embodiment, the industrial computer 9 is 292mm long, 93mm wide, and 290mm high, with a hard disk memory of 2TB, which can continuously store one month's worth of real-time monitoring data. The industrial computer is installed in the main control room and is used to operate the real-time video monitoring, intelligent analysis, and alarm system for the in-flow inoculant.

[0073] The audible and visual alarm 8 is 174mm long, 90mm wide, and 278.5mm high. It includes an alarm indicator light and an alarm buzzer. The alarm indicator light flashes red, and the alarm buzzer executes a cyclical alarm command. Both the alarm buzzer and the alarm indicator light can be manually turned on or off. The audible and visual alarm 8 is connected to an industrial control computer 9 via a USB interface and the RS485 communication protocol. It is controlled by a real-time video monitoring and intelligent analysis alarm system. When the system sends an alarm signal, the audible and visual alarm 8 notifies the operator by flashing LEDs.

[0074] The LCD monitor 11 is 440.7mm long, 179.9mm wide, and 269.2mm high. It displays the monitoring screen and is also responsible for displaying the system's human-machine interface.

[0075] The above are the detailed parameters of the hardware component of a real-time video monitoring, intelligent analysis, and alarm system for in-flow inoculants, as described in this invention patent. Example

[0076] This embodiment discloses a real-time video monitoring and intelligent analysis alarm system software for in-flow inoculants, which is mounted on the industrial control computer 9 in Embodiment 1. The LCD display 11 can fully display the real-time video monitoring and intelligent analysis alarm system software for in-flow inoculants.

[0077] The real-time video monitoring, intelligent analysis, and alarm system software for inoculants is embedded in the industrial control computer 9. It includes a monitoring module responsible for real-time monitoring, recording, and playback; a positioning module responsible for locating and tracking molten steel and inoculant targets; an identification module responsible for identifying molten steel and inoculant targets; an analysis module responsible for analyzing the real-time status of molten steel and inoculant; an alarm module responsible for issuing alarm events and triggering audible and visual alarms when the real-time status of molten steel and inoculant is abnormal; and a data management module responsible for managing alarm event information, storing and retrieving video recordings.

[0078] While patent WO2007 / 088285 does not present its software system details, in this invention, software design and implementation are the core of realizing system functionality and the key to ensuring accuracy. The monitoring module is used for:

[0079] 1. Real-time monitoring of the casting production line;

[0080] 2. Real-time monitoring video recording;

[0081] 3. Video playback of the casting production line;

[0082] 4. Enable and disable real-time recording, real-time preview, screenshots, playback of historical video data, pause, fast forward, rewind, and screenshots.

[0083] Compared with patent WO2007 / 088285, this invention adds a positioning module that can automatically locate the position of the casting machine and the position of the inoculant nozzle. Therefore, this solution is not only applicable to scenarios where the relative positions of the on-site devices are fixed, but also to complex scenarios where the relative positions of the casting machine and the inoculant nozzle change due to production factors such as vibration.

[0084] like Figure 3 The positioning module is used for:

[0085] 1. Receive the location information of the pouring area in the casting production line screen;

[0086] 2. Use template matching to locate and mark the positions of the casting machine and the inoculant nozzle in the monitoring image;

[0087] 3. When in the "in the pouring area" state, use the target tracking method to track and mark the positions of the pouring machine and inoculant nozzle in the pouring area. When either of the targets leaves the pouring area, stop tracking and output the "not in the pouring area" signal.

[0088] 4. When in the "in the pouring zone" state, determine and mark the pouring position of molten steel and the inoculant inoculation position based on the relative invariance of the positions of the pouring machine and molten steel, and the inoculant nozzle;

[0089] 5. To ensure the accuracy of the positioning results, a multi-frame positioning result comparison method was used in step 2 of the above positioning module. The optimal value after multi-frame comparison was used in the output result. In step 3, the target tracking and marking results were optimized. When the target position does not change, the target tracking result will have a very small directional fluctuation. Therefore, a target tracking result optimization method was designed. When the multi-frame result has only a very small directional (pixel level) fluctuation, the target is determined to be stationary and only one target tracking result is output. When the multi-frame result changes significantly, the target is determined to be moving and the tracking result is output normally.

[0090] Compared with patent WO2007 / 088285, this invention can identify not only the state of the inoculant but also the state of the molten steel through video images. This not only improves the accuracy of inoculant identification but also provides users with more on-site information.

[0091] The recognition module is used for:

[0092] 1. Receive information on the molten steel pouring position and the inoculant inoculation position from the positioning module;

[0093] 2. Molten steel identification within the pouring location;

[0094] 3. Identification of progestin within the progestin's location;

[0095] 4. Calculate the pixel area of ​​the molten steel identification result and the inoculant identification result respectively, and establish a coordinate system with real time. The horizontal axis represents real time and the vertical axis represents the pixel area of ​​molten steel or inoculant. Establish coordinate systems to obtain the real-time curve of "molten steel area-time" and the real-time curve of "inoculant area-time" respectively.

[0096] The identification module creates a dynamic mask based on the detailed locations of molten steel and inoculant. When identifying molten steel, the inoculant area is masked; when identifying inoculant, the molten steel area is masked. This prevents interference between molten steel and inoculant, thus achieving good identification results for both molten steel and inoculant.

[0097] The analysis module is used for:

[0098] 1. Steel condition analysis, including two states: "in pouring" and "not poured";

[0099] 2. Pregnancy status analysis, including two states: "in pregnancy" and "not in pregnancy";

[0100] 3. Based on the real-time curves of "molten steel area-time" and "inoculant area-time", the molten steel pouring cycle, the inoculant incubation cycle, the peak and average amounts of molten steel and inoculant within the cycle, the cumulative number of molten steel pouring cycles, and the cumulative number of inoculant incubation cycles are analyzed.

[0101] 4. Based on the number of incubation cycles, peak incubation value, and average incubation value of the incubator, establish a coordinate system with the horizontal axis representing the number of incubation cycles and the vertical axis representing the peak incubation value and average incubation value of the incubator, to obtain a real-time curve of "incubator cycle - incubation value usage".

[0102] like Figure 4 The alarm module is used for:

[0103] 1. When the molten steel is in the "pouring" state and the inoculant is in the "not inoculated" state, it is an abnormal state of "inoculant nozzle blockage";

[0104] When the probiotic is in a "probiotic insufficiency" state, it is identified as an "abnormal state of probiotic insufficiency".

[0105] When either of the two abnormal states mentioned above occurs, the audible and visual alarm will be invoked to execute an alarm event.

[0106] 2. Calculate the dosage of progesterone in real time.

[0107] The alarm module also uses a delayed judgment method. It will only determine an abnormal state and output an alarm signal when it receives abnormal state signals from the analysis module for 25 consecutive frames (one second). If the accumulated abnormal state signals do not reach 25 frames (one second), it is determined to be interference from an unexpected situation, and no alarm signal is output. The number of consecutive frames can be freely set according to actual needs.

[0108] The data management module is used for:

[0109] 1. Store screenshots, time, production line, type, and remarks when an alarm event occurs;

[0110] 2. Access to database alarm information;

[0111] 3. Playback of alarm data. Example

[0112] This embodiment discloses a real-time video monitoring and intelligent analysis alarm method for in-flow progesterone, implemented using the real-time video monitoring and intelligent analysis alarm system for in-flow progesterone disclosed in Embodiments 1 and 2, and includes the following steps:

[0113] Step 1: Data Collection

[0114] Real-time raw stream data in H264 encoded format was collected using image acquisition equipment from the casting production line.

[0115] Step 2: Process the collected real-time data through the monitoring module.

[0116] The RTSP protocol is used to read data from the image acquisition device to achieve real-time preview of the monitoring screen; the FFMPEG decoding library is used to decode the H264 raw stream data to achieve recording of the monitoring screen.

[0117] Step 3: Marking the pouring area

[0118] The pouring area is marked by drawing a rectangle on the monitoring screen using the interactive method of this system.

[0119] Step 4: Determine whether the pouring machine and inoculant nozzle are located within the pouring area.

[0120] The template matching method of the positioning module locates and marks the positions of the casting machine and the inoculant nozzle in the monitoring image. Template matching always outputs one casting machine position and one inoculant nozzle position. When neither the casting machine nor the inoculant nozzle is detected in the image, the output position information consists of two standard rectangles located in the upper left corner of the image. When the casting machine and nozzle are detected, their positions are compared with the received casting area position. If both are within the casting area, a "in casting area" signal is output; otherwise, a "not in casting area" signal is output. When neither the casting machine nor the nozzle is detected, or both are in the "not in casting area" state, the template matching method is intermittently used to find the casting machine and inoculant nozzle positions until both are in the "in casting area" state, at which point the template matching method stops.

[0121] To ensure the accuracy of the localization results, a multi-frame localization result comparison method was used for the template matching method described above. The optimal value after multi-frame comparison was used in the output result to reduce the false positive rate.

[0122] Step 5: Target tracking of the casting machine and inoculant nozzle

[0123] When in the "in the pouring area" state, the target tracking method is used to track and mark the positions of the pouring machine and inoculant nozzle in the pouring area. When either of the targets leaves the pouring area, the tracking stops and the "not in the pouring area" signal is output.

[0124] To ensure the accuracy of the positioning results, the target tracking and marking results were optimized. A target tracking result optimization method was designed to solve the problem of extremely small directional fluctuations in the target tracking results when the target position remains unchanged, which affects the image viewing effect. In this optimization method, when there are only extremely small directional fluctuations in the results across multiple frames, the target is determined to be stationary, and only one target tracking result is output. When the changes in the results across multiple frames are large, the target is determined to be moving, and the tracking result is output normally.

[0125] Step 6: Locating the molten steel and inoculant

[0126] When in the "in the pouring zone" state, the pouring position of molten steel and the inoculant position are determined and marked based on the relative invariance of the positions of the pouring machine and the molten steel, and the inoculant and the inoculant nozzle.

[0127] Step 7: Identify the molten steel and inoculant in the location area.

[0128] Based on the positioning results obtained in step 5 above, the developed molten steel identification algorithm and inoculant identification algorithm are used to complete the identification of molten steel and inoculant.

[0129] To address the interference caused by the close proximity of molten steel and inoculant to the steel and inoculant identification algorithm, this embodiment employs an image masking method to shield the inoculant region during molten steel identification. Similarly, the molten steel region is shielded during inoculant identification.

[0130] In addition, to address the issue of flames appearing in the molten steel area during inoculant inoculation, a flame removal algorithm was used for molten steel identification. The flame removal algorithm is implemented by post-processing the molten steel detection results. It collects the molten steel detection results of the current frame and the previous frame, performs an image AND operation on the two results, and extracts the common molten steel area in the two frames. This effectively solves the problem of the flame in the molten steel area affecting molten steel identification and ensures the stability of the molten steel identification results.

[0131] Step 8: Data statistics and visualization of recognition results

[0132] Based on the identification results of molten steel and inoculant, the pixel areas of the identification results of molten steel and inoculant are counted respectively, and a coordinate system is established with real time. The horizontal axis represents real time, and the vertical axis represents the pixel area of ​​molten steel and inoculant, respectively, to obtain the real-time curve of "time-molten steel area" and the real-time curve of "time-inoculant area".

[0133] Step 9: Results Data Analysis

[0134] Based on the real-time curves of "time-molten steel area" and "time-inoculant area" from step seven, data analysis was performed to determine the following states: when the molten steel area is less than 20, it is in the "not being poured" state; when the molten steel area is greater than 20, it is in the "pouring" state; when the inoculant area is less than 10, it is in the "not inoculated" state; when the inoculant area is greater than 10, it is in the "inoculated" state; and when the inoculant area is between 0 and 30 within a single pouring cycle, it is in the "insufficient inoculant" state.

[0135] Based on the state results of the molten steel and the inoculant, the moment when the molten steel changes from the "not being poured" state to the "being poured" state is determined as the pouring start time, and the moment when the molten steel changes from the "being poured" state to the "not being poured" state again from the pouring start time is determined as the pouring end time. The absolute value of the difference between the pouring start time and the pouring end time is determined as one pouring cycle.

[0136] The inoculum is defined as the moment when it changes from a "non-inoculum" state to a "inoculum" state. The inoculum is defined as the moment when it changes from a "inoculum" state to a "non-inoculum" state again after the inoculum is poured. The absolute value of the difference between the inoculum start time and the inoculum end time is defined as one inoculum cycle.

[0137] The maximum area of ​​molten steel in one pouring cycle is defined as the peak value of molten steel pouring, and the ratio of the total area of ​​molten steel in one pouring cycle to the total cycle time is defined as the average value of molten steel pouring.

[0138] The maximum area of ​​the progesterone in one incubation cycle is defined as the peak incubation value of the progesterone, and the ratio of the total area of ​​the progesterone to the total cycle time in one incubation cycle is defined as the average incubation value of the progesterone.

[0139] Each time the incubation cycle is updated, the difference between the incubation agent dosage in the current incubation cycle and the dosage in the previous incubation cycle is determined based on the real-time curve of "incubation agent cycle - incubation agent dosage".

[0140] Step 10: Abnormal Data Filtering and Alarms

[0141] Based on the data analysis results of step nine, it is determined that when the molten steel is in the "pouring" state and the inoculant is in the "not inoculated" state, it is an abnormal state of "inoculant nozzle blockage".

[0142] When the probiotic is in a "probiotic insufficiency" state, it is identified as an "abnormal state of probiotic insufficiency".

[0143] An alarm event is triggered when either of the two abnormal states mentioned above occurs.

[0144] When an alarm event is triggered, the industrial controller's industrial computer sends an alarm command to the audible and visual alarm via the RS485 communication protocol. After receiving the alarm signal, the audible and visual alarm continuously plays the alarm audio and flashes the alarm light, while simultaneously changing the system interface to alarm mode.

[0145] There are two ways to clear an alarm event: manual clearing and automatic clearing. Manual clearing means that the system user closes the alarm event after completing the relevant interaction with the system alarm interface. Automatic clearing means that the alarm mode is automatically closed when the data analysis result changes from an abnormal state to a normal state while in alarm mode.

[0146] Step 11: Abnormal data management and abnormal data review

[0147] Whenever an alarm event is triggered, the system calls the data management module to take a screenshot of the alarm event, record the alarm event time, production line, type, and alarm information notes manually added by the system operator, and integrate them into a single alarm data entry into the system database.

[0148] System operators can briefly view the on-site status when an alarm occurred by accessing screenshots of each alarm data entry in the database, and can also view on-site video footage of the alarm occurrence based on that alarm data entry. Example

[0149] The following describes the application of a real-time video monitoring, intelligent analysis, and alarm system for in-flow inoculants in actual production:

[0150] like Figure 1 As shown, the image acquisition device is installed on one side of the production line and connected to the industrial controller in the control room. After the image acquisition device is installed, the connection with the industrial controller should be completed. Then, the image acquisition device should be initialized within the industrial controller, initializing its IP address, communication port number, account and password information, and completing the corresponding settings such as image acquisition resolution, frame rate, and bitstream transmission type. After completing the corresponding settings, the preview should be started using the real-time video monitoring intelligent analysis and alarm system. Once the real-time video interface transmitted from the camera can be seen in the system interface, the installation and debugging of the image acquisition device and controller are complete.

[0151] See process Figure 5 After receiving a production order, employees complete the necessary preparations and begin actual production. Upon starting production, the operator first turns on the system's monitoring switch. The system automatically starts real-time recording. When the recorded data exceeds 1024MB, the system automatically segments and saves it, naming the video file with "start time - end time". When the operator turns off the monitoring switch, the system simultaneously disables the recording function. After monitoring begins, the operator can see the real-time monitoring screen. The operator needs to mark the approximate working areas of the casting machine, inoculant nozzle, molten steel, and inoculant by drawing rectangles on the monitoring screen. Then, the operator turns on the identification switch and alarm switch. Figure 3The positioning module logic view shown indicates that the system will detect the casting machine and inoculant nozzle within the selected rectangular area. If both the casting machine and inoculant nozzle can be detected simultaneously, the steel pouring position and inoculant inoculation position will be determined based on their locations. If neither can be detected simultaneously, detection will continue until both are detected. If the system calibrates the steel pouring position and inoculant inoculation position and there is a corresponding deviation from the actual position, the operator can manually adjust it.

[0152] After the locations of the molten steel and inoculant are calibrated, the system will call the identification module to identify the molten steel and inoculant. In this embodiment, a quenching algorithm is used in the molten steel identification. When producing certain special types of castings, the inoculant used in the process may react violently with the molten steel, resulting in a large number of flames at the inoculation production location 3. Without the quenching algorithm, some flames may be mistakenly detected as molten steel, affecting subsequent processes. The quenching algorithm achieves the quenching effect through post-processing of the molten steel detection results.

[0153] Figure 8(a) shows the steel detection results without the addition of the annealing algorithm, and Figure 8(b) shows the steel detection results after the addition of the annealing algorithm. It can be seen that the annealing algorithm effectively solves the influence of flame on steel detection. After obtaining the steel and inoculant detection results, a coordinate system is established by the pixel area of ​​the steel and inoculant identification results and the real-time time. At the same time, the steel identification results, inoculant identification results, the real-time curve of "steel area-time" and the real-time curve of "inoculant area-time" are displayed on the interactive interface.

[0154] In obtaining such Figure 6 , Figure 7 After displaying the real-time curves of "molten steel area - time" and "inoculant area - time," the system will call the analysis module to perform real-time analysis of the molten steel and inoculant states. Based on preset thresholds, the system determines the state as follows: when the molten steel area is less than 20, it is in the "not being poured" state; when the molten steel area is greater than 20, it is in the "pouring" state; when the inoculant area is less than 10, it is in the "not inoculated" state; when the inoculant area is greater than 10, it is in the "inoculating" state; and when the inoculant area is between 0 and 30 within a single pouring cycle, it is in the "insufficient inoculant" state.

[0155] Subsequently, the following call is made: Figure 4The alarm module shown in this embodiment, when the molten steel is in the "pouring" state and the inoculant is in the "not inoculated" state, determines it as an "inoculant nozzle blockage" abnormal state, issues an alarm signal, and calls the audible and visual alarm to execute the alarm event. When the alarm event occurs, the alarm's LED flashes to notify the operator to handle the abnormal event immediately. At the same time, the system will record the alarm event. This alarm information includes a monitoring screenshot of the alarm time, the alarm event, the alarm casting production line, and the abnormal alarm type determined by the system. After the operator handles the abnormal event, they can manually add the corresponding alarm remarks, and the system will record this information in the database. System operators can briefly view the on-site status when the alarm occurred by accessing the screenshot of this alarm data, and can also view the on-site video when the alarm occurred based on this alarm data.

[0156] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A real-time video monitoring intelligent analysis alarm system for stream inoculant, characterized in that, The application relates to a real-time video monitoring intelligent analysis alarm system software for stream inoculant embedded in an industrial controller and an image acquisition device. The image acquisition device is used for monitoring and collecting the states of a pouring machine, liquid steel, stream inoculant and a stream inoculant nozzle. The industrial controller comprises an audible and visual alarm (8) and an industrial computer (9), wherein the industrial computer (9) is used for running the real-time video monitoring intelligent analysis alarm system software for stream inoculant; when the real-time video monitoring intelligent analysis alarm system software for stream inoculant transmits an alarm signal to the audible and visual alarm (8), the audible and visual alarm (8) executes an alarm event. The real-time video monitoring intelligent analysis alarm system software for stream inoculant comprises a monitoring module, a positioning module, an identification module, an analysis module and an alarm module. The monitoring module is used for monitoring a casting production flow line in real time. The positioning module is used for receiving pouring area position information in a casting production flow line picture; the positioning module finds and marks the positions of a pouring machine and an inoculant nozzle in a monitoring image. The identification module is used for receiving the liquid steel pouring position and the inoculant inoculation position information of the positioning module, and then performing liquid steel identification in the liquid steel pouring position and inoculant identification in the inoculant inoculation position. The analysis module is used for performing liquid steel state analysis and inoculant state analysis; the liquid steel state comprises two states of "pouring" and "not pouring"; the inoculant state comprises two states of "inoculating" and "not inoculating". When the liquid steel state is in the "pouring" state and the inoculant state is in the "not inoculating" state, an "inoculant nozzle blockage" abnormal state is determined. When the inoculant is in the "inoculant deficiency" state, an "inoculant deficiency" abnormal state is determined. The alarm module is used for alarming after determining an abnormal state; when any one of the abnormal states occurs, an alarm event is triggered, an alarm signal is sent out, and the audible and visual alarm (8) is called to execute an audible and visual alarm event.

2. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to claim 1, characterized in that, In the positioning module, the pouring area position information comprises "in a pouring area" and "not in a pouring area". When in the "in a pouring area" state, a target tracking method is used to track and mark the positions of the pouring machine and the inoculant nozzle in the pouring area; when any one of the pouring machine and the inoculant nozzle leaves the pouring area, tracking is stopped and the "not in a pouring area" signal is outputted. When in the "in a pouring area" state, the liquid steel pouring position and the inoculant inoculation position are determined and marked.

3. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to claim 1, characterized in that, The identification module is also used for respectively counting the pixel areas of liquid steel identification results and inoculant identification results, and establishing a coordinate system with real-time time, so as to respectively obtain a "liquid steel area-time" real-time curve and an "inoculant area-time" real-time curve.

4. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to claim 3, characterized in that, The analysis module is also used for analyzing the "liquid steel area-time" real-time curve and the "inoculant area-time" real-time curve to obtain a "stream inoculant cycle-inoculant amount" real-time curve.

5. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to claim 1, characterized in that, In the alarm module, when 25 consecutive frames receive the abnormal state signal sent by the analysis module, an alarm signal is output; when the abnormal state signal does not accumulate for 25 consecutive frames, no alarm signal is output.

6. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to any one of claims 1-5, characterized in that, The real-time video monitoring intelligent analysis alarm system software of the stream inoculant further comprises a data management module. The data management module is used for storing screenshot, time, production line, type and note information when an alarm event occurs.

7. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to any one of claims 1-5, characterized in that, The image acquisition device comprises a camera (5) and a sealed cover, the camera (5) is arranged in parallel to the casting production line, the camera (5) is installed in the sealed cover, a glass lens is installed at the front end opening of the sealed cover, and the sealed cover and the glass lens are in sealing connection.

8. The real-time video monitoring intelligent analysis alarm system of stream inoculant according to claim 7, characterized in that, A forced air cooler (7) is fixed at the rear end of the sealed cover, and the camera (5) is fixed at the front end of the forced air cooler (7).

9. A real-time video monitoring intelligent analysis alarm method for stream inoculant, characterized in that, The method comprises the following steps: S1. Data acquisition; using an image acquisition device to acquire real-time bare stream data of the casting production line; S2. Processing the acquired real-time data through a monitoring module; S3. Calibrating the pouring area; calibrating the pouring area in the monitoring picture; S4. Determining whether the pouring machine and the inoculant nozzle position are in the pouring area; finding and marking the pouring machine position and the inoculant nozzle position in the monitoring picture and outputting a position state signal; S5. Target tracking of the pouring machine and the inoculant nozzle; Outputting the position state signal; S6. Positioning of the molten steel and the inoculant; marking the molten steel pouring position and the inoculant inoculation position according to the position state signal; S7. Identifying the molten steel and the inoculant in the positioning area; using the image mask method in the identification, shielding the inoculant area when identifying the molten steel, and shielding the molten steel area when identifying the inoculant; S8. Statistical analysis and visualization of the identification result data; drawing the real-time curves of "time-molten steel area" and "time-inoculant area"; S9. Result data analysis; judging the molten steel state and the inoculant state according to the threshold value; S10. Abnormal state data screening and alarm; triggering an alarm event when the molten steel and the inoculant are abnormal; S11. Abnormal state data management and abnormal data backtracking.

10. The real-time video monitoring intelligent analysis alarm method of stream inoculant according to claim 9, characterized in that, In the step S7, the fire removal algorithm is used in the molten steel identification, the fire removal algorithm collects the current frame molten steel detection result and the last frame molten steel detection result, and performs image operation on the two results, takes the common molten steel area of the two frames, and realizes the fire removal effect.

Images