A monitoring and early warning method for sintering machine trolley wheel swing

Abstract

The application provides a sintering machine trolley wheel swing monitoring and early warning method, relates to the technical field of image early warning, and comprises the following steps: acquiring an image set of trolley wheels in a production scene, training the image set through a semantic segmentation model to obtain a trained model; using the trained model to process monitoring images, performing segmentation mask processing on the wheel area in the monitoring images of the on-site trolley, and acquiring a wheel deflection angle; determining whether the wheel deflection angle is in a normal operating state based on a set rule and processing according to a risk level classification; and making corresponding alarms according to the risk level of the wheel deflection angle. The sintering machine trolley wheel swing monitoring and early warning method can analyze the real-time collected sintering machine trolley wheel images, determine whether the sintering machine trolley wheels have deviated, improve the detection efficiency, and avoid safety hazards caused by manual detection negligence.

Description

TECHNICAL FIELD

[0001] The present application relates to the image early warning technical field, and in particular to a sintering machine trolley wheel swing monitoring and early warning method. BACKGROUND

[0002] In the pellet production process of the iron-containing kiln, the main equipment for producing pellets is composed of a sintering machine, a rotary kiln and a ring cooler, wherein the sintering machine is a key equipment. In industrial production, the surface temperature of the sintering machine is as high as 1000 DEG C, and due to the heat conduction of the trolley body, the temperature of the trolley body can reach about 45 DEG C. During the circulation and reciprocating operation of the trolley, the high and low temperature heating is uneven, and the working environment is harsh due to the dust erosion in the production site. During the sintering process, the weight of the sintering machine trolley wheel operation includes the weight of the body and the weight of the material, and the sintering machine operates continuously. The trolley wheel will appear torque, and the axle is prone to faults such as grease saponification, cage rupture and bearing ball damage. When the trolley continues to run with any of the above faults, the wheel will swing, and in severe cases, it may fall off. If the trolley is not replaced in time, the wheel is easy to fall off when entering the tail curve, which will greatly affect the safe operation of the entire sintering system and the normal production and economic benefits of the enterprise.

[0003] At present, the swing monitoring of the wheel mainly relies on manual inspection and judgment. Manual monitoring is time-consuming and laborious, and the human eye cannot continuously and stably complete these highly repetitive work, which is easy to cause missed detection and has low intelligence.

[0004] With the development of image intelligent identification technology, the use of machine vision for wheel swing deviation monitoring can not only further improve the intelligent level of monitoring in the production process, but also reduce repetitive labor and improve economic benefits. SUMMARY

[0005] In order to solve the problem of monitoring and early warning of the swing deviation of the sintering machine trolley wheel, the present application provides a sintering machine trolley wheel swing monitoring and early warning method, which can analyze the real-time collected sintering machine trolley wheel image, determine whether the sintering machine trolley wheel has deviation fault, improve the detection efficiency, and avoid the safety hidden danger caused by manual detection negligence.

[0006] The technical solution adopted by this application to solve its technical problem is: a method for monitoring and early warning of wheel swaying of a sintering machine trolley, comprising: acquiring an image set of trolley wheels in a production scene; training the image set through a semantic segmentation model to obtain a trained model; using the trained model to process the monitoring images, performing segmentation and masking processing on the wheel area in the monitoring images of the trolley on site to obtain the wheel deflection angle; determining whether the wheel deflection angle is in a normal operating state based on set rules, and classifying and processing it according to risk level; and issuing a corresponding alarm according to the risk level of the wheel deflection angle.

[0007] In one specific implementation, the method for obtaining the wheel deflection angle is as follows:

[0008] The trained model is used to segment and mask the wheels in the monitoring image, and the wheel mask area is separated from the background to obtain a binary image and eliminate image noise.

[0009] Construct the circumscribed rectangle of the wheel in the binary image, obtain the geometric relationship between the long side of the circumscribed rectangle of the trolley wheel and the baseline, and obtain the deflection angle of the wheel.

[0010] In one specific implementation, the baseline is determined by using a straight line detection algorithm to detect a fixed linear object in the monitoring image that is parallel to the side of the wheel under the designed operating state, including:

[0011] The linear monitoring algorithm detects the set of line segments {l1,…,l} of a fixed linear object parallel to the side of the wheel during the design and operation state. n Based on the actual operating conditions on site, one of the l lines was selected. m The straight line segment is used as the baseline, and the coordinates of the two endpoints of the baseline are (x, y, y) and (x, y, y). m ,y m ), (x m+1 ,y m+1 ).

[0012] In one specific implementation, the step of constructing the circumscribed rectangle of the wheel in the binary image, obtaining the geometric relationship between the long side of the circumscribed rectangle of the trolley wheel and the baseline, and deriving the wheel's deflection angle includes:

[0013] From baseline l m The corresponding coordinates of the two endpoints (x) m ,y m ), (x m+1 ,y m+1 ), to obtain l m Compared to the slope k of the horizontal line m :

[0014]

[0015] By l m Compared to the slope k of the horizontal line m , get l m Compared to the curvature value of the horizontal line And obtain the baseline l m The angle θ relative to the horizontal line m :

[0016]

[0017] Where θ m The value range is [-90°, 0°] and [0°, 90°], where negative numbers represent counterclockwise rotation and positive numbers represent clockwise rotation;

[0018] The angle θ between the baseline and the horizontal line m Given the angle α between the longer side of the circumscribed rectangle and the horizontal line, where α ranges from [-90°, 0°] and [0°, 90°], calculate the relative angle β between the longer side of the circumscribed rectangle and the baseline: β = |α - θ| m |。 .

[0019] In one specific implementation, the normal operating state is a state in which the deflection angle between the long side of the outer rectangular frame of the wheel and the baseline is no greater than n°, where n° is the maximum allowable swing of the wheel under the on-site wheel operating state, and n° is set according to the actual on-site operating conditions.

[0020] In one specific implementation, the risk level is determined by comparing the wheel deflection angle with a preset warning rule.

[0021] In one specific implementation scheme, the preset early warning rule is as follows:

[0022] When 0° ≤ wheel deflection angle ≤ n°, it is considered to be within the normal swing range.

[0023] When n° < wheel deflection angle ≤ 2n°, it is judged as a low-risk level, but should be monitored in subsequent operations.

[0024] When the wheel deflection angle is greater than 2n°, it is considered a high-risk level and personnel must be dispatched immediately to handle the situation.

[0025] In one specific implementation, the monitoring and early warning method for the sintering machine trolley wheel sway is equipped with an image recorder to obtain images of the running status of the trolley wheels and reference position images of the trolley's running position. A supplementary lighting device is also provided to provide illumination for the image recorder's shooting. A switch, an image processing server, and a display terminal are also provided. The switch is communicatively connected to the image recorder and the image processing server, and the display terminal is communicatively connected to the image processing server.

[0026] In a specific embodiment, the image processing server processes the monitoring image through the trained model, identifies the wheel, judges the wheel deflection angle, and makes an alarm prompt according to the risk level, and the display terminal displays the alarm prompt, the running state information of the wheel of the trolley at the time of triggering the alarm, and the reference position information of the trolley running.

[0027] In a specific embodiment, in the corresponding alarm according to the risk level of the wheel deflection angle, the middle region position in the camera field of view is selected through multiple determinations, the collection time threshold t is set according to the wheel running speed v v , and the wheel is detected once every t v time to collect an image of the wheel deflection angle, and make a corresponding alarm by comparing with the preset early warning rule.

[0028] The advantages of the present application are:

[0029] 1. The sintering machine trolley wheel swing monitoring and early warning method uses an improved semantic segmentation model algorithm to segment the wheel region of the real-time collected trolley wheel image to determine whether the deflection angle of the long side of the minimum rectangular frame of the target wheel in the monitoring image and the set reference line is greater than the risk threshold, and timely safety warning is made to the trolley running state, which improves the monitoring efficiency and intelligent level and reduces the risk of accidents.

[0030] 2. The sintering machine trolley wheel swing monitoring and early warning method uses the relative angle between the minimum rectangular frame of the trolley wheel and the reference line under actual running state as the wheel deflection angle, which clearly shows the wheel boundary relationship and relative position relationship, improves the analysis and calculation speed and the accuracy of the monitoring data.

[0031] 3. The sintering machine trolley wheel swing monitoring and early warning method determines the wheel deflection angle according to different risk levels, further saves the detection effect diagram of the wheel image and the reference position image of the corresponding trolley when detecting that the wheel has swing failure, displays the monitoring effect diagram and the reference position image on the display terminal, and makes corresponding alarm information, thereby improving the visualization and functional level of the monitoring.

[0032] 4. The sintering machine trolley wheel swing monitoring and early warning method trains the image set of the trolley wheel under different working scenes with a semantic segmentation model, and then processes the monitoring image by calling the trained model to identify the trolley wheel and judge the wheel deflection angle, thereby further improving the intelligent level of the monitoring mode and the recognition rate of the monitoring target in the image. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1A work flow diagram of a sintering machine trolley wheel swing monitoring and early warning method of the present application;

[0034] Figure 2 A platform building diagram of a sintering machine trolley wheel swing monitoring and early warning method of the present application;

[0035] Figure 3 A trolley wheel overhead view diagram of a sintering machine trolley wheel swing monitoring and early warning method of the present application;

[0036] Figure 4 A straight line detection diagram of a sintering machine trolley wheel swing monitoring and early warning method of the present application;

[0037] Figure 5 A wheel region image processing algorithm flow diagram of a sintering machine trolley wheel swing monitoring and early warning method of the present application;

[0038] Figure 6 A wheel deflection angle diagram constructed by the longest side of the minimum circumscribed rectangle and the set reference line of a sintering machine trolley wheel swing monitoring and early warning method of the present application. DETAILED DESCRIPTION

[0039] The embodiment of the present application provides a sintering machine trolley wheel swing monitoring and early warning method, solves the problems of difficult monitoring, inaccurate results, and failure to realize corresponding early warning for different risk levels of the wheel swing deviation of the existing sintering machine trolley in the field work process, and the overall idea is as follows:

[0040] Please refer to Figure 1 The present application provides a sintering machine trolley wheel swing monitoring and early warning method, which comprises: acquiring an image set of the trolley wheel in a production scene, training the image set through a semantic segmentation model to obtain a trained model; using the trained model to process a monitoring image, performing segmentation mask processing on the wheel region in the monitoring image of the field trolley to acquire a wheel deflection angle; determining whether the wheel deflection angle is in a normal operating state based on a set rule and processing according to a risk level classification; and making a corresponding alarm according to the risk level of the wheel deflection angle. A large number of overhead images of the trolley wheel in the production scene are acquired and labeled, the image set is trained through a semantic segmentation model to obtain a trained model. The improved trained model algorithm is used to segment the wheel region of the real-time collected trolley wheel image, the real-time collected sintering machine trolley wheel image is analyzed, and it is determined whether the sintering machine trolley wheel has a deviation fault, so that the detection efficiency is improved and the safety hidden danger caused by manual detection negligence is avoided.

[0041] In this embodiment, the method for identifying the wheel and determining its deflection angle is as follows: A trained model is used to segment and mask the wheel in the monitoring image, separating the masked area from the background to obtain a binary image and eliminate image noise. A bounding box of the wheel is constructed in the binary image, and the geometric relationship between the long side of the bounding box and the baseline is obtained to determine the wheel's deflection angle. Specifically, the wheel deflection angle is the change in angle between the geometric positional relationship between the minimum bounding box of the sintering machine wheel and the baseline under actual operating conditions and under designed operating conditions. Specifically, the on-site acquired video stream is input into the semantic segmentation model. After segmenting and masking the wheel area, image processing methods are used to construct the minimum bounding box. Based on the positional relationship between the minimum bounding box of the sintering machine trolley wheel and the set baseline, the deflection angle of the sintering machine trolley wheel is determined, and corresponding countermeasures are taken according to different levels of deflection angle.

[0042] In this example, the baseline is determined by using a line detection algorithm to detect fixed linear objects parallel to the wheel side in the designed operating state in the monitoring image. The setting method is as follows (taking the direction of the trolley track line as the baseline): With the video recorder, such as a camera, fixed, the line detection algorithm detects the set of line segments {l1,…,l...} of the fixed linear objects parallel to the wheel side in the designed operating state. n Based on the actual operating conditions on site, one of the l lines was selected. m The straight line segment is used as the baseline, and the coordinates of the two endpoints of the baseline are (x, y, y) and (x, y, y). m ,y m ), (x m+1 ,y m+1 ).

[0043] Construct the bounding rectangle of the wheel in the binary image, obtain the geometric relationship between the long side of the bounding rectangle of the trolley wheel and the baseline, and derive the wheel's deflection angle, including:

[0044] From baseline l m The corresponding coordinates of the two endpoints (x) m ,y m ), (x m+1 ,y m+1 ), to obtain l m Compared to the slope k of the horizontal line m :

[0045]

[0046] By l m Compared to the slope k of the horizontal line m , get l mThe angle θ of the baseline l compared to the horizontal line and the baseline l is obtained m The angle θ of the baseline l compared to the horizontal line m :

[0047]

[0048] The angle θ of the baseline l compared to the horizontal line m , where negative numbers represent counterclockwise rotation and positive numbers represent clockwise rotation;

[0049] The angle θ of the baseline l compared to the horizontal line m and the angle α of the long side of the circumscribed rectangle compared to the horizontal line, where α is in the range [-90°, 0°] and [0°, 90°], to calculate the relative angle β = |α-θ m | between the long side of the circumscribed rectangle of the wheel and the baseline l.

[0050] In the data labeling and training phase, video data of the trolley wheel of the sintering machine in different working scenarios can be obtained, and then the video data is converted into an image set, and the wheel area is labeled with the help of the image labeling tool Labelme.

[0051] In the image segmentation and processing phase, the improved semantic segmentation model, such as the improved DeepLabv3+ model, is used to process each frame of image in the video data to obtain the target contour. The segmented wheel image is marked with a mask of the same color, and the target contour of the wheel image mask is obtained. The wheel image mask area and the background are separated using the HSV color space segmentation method, and a binary image containing only the wheel area and the background area is reorganized, and then the sintering machine trolley wheel binary image is obtained. The discontinuous area in the binary image can be further negated and filled. After obtaining the binary image containing the wheel area, a straight line detection algorithm is used to detect the fixed linear object in parallel with the side of the wheel in the design state. From the detected straight line set, a baseline is set, and the angle of the baseline compared to the horizontal direction is obtained. After setting the baseline, the minimum circumscribed rectangle box function of OpenCV is used to form a minimum circumscribed rectangle surrounding box for the semantic segmented target image, and the deflection angle of the trolley wheel compared to the horizontal direction is obtained. The relative angle relationship between the long side of the minimum circumscribed rectangle surrounding box of the sintering machine trolley wheel in the image and the baseline can be used to obtain the wheel deflection angle. The wheel swing fault in the field is divided into different levels by comparing with the preset risk threshold, and corresponding measures are taken for different levels.

[0052] Further, the improved semantic segmentation model specifically includes: replacing the feature extraction network of the segmentation model with a lightweight MobileNetv2 network to significantly reduce model parameters, accelerate convergence and inference time according to the needs of the production scene; introducing an attention module in the channel to make up for the loss of segmentation accuracy caused by model lightweight, ensuring the segmentation accuracy of the wheel edge in the image; using Dice Loss instead of cross-entropy loss function (CE Loss) to improve the smoothness of the network when performing edge semantic segmentation, guiding the network to converge in the correct direction. Input the sintering machine trolley wheel image set under different working scenes into the improved network model for training to obtain the trained wheel semantic segmentation model, and then use the trained model to perform semantic segmentation on the wheels in the monitoring image to obtain the wheel mask image.

[0053] In this example, the normal operating state is a state in which the deflection angle between the long side of the wheel circumscribed rectangle frame and the reference line is not greater than n°, where n° is the maximum value of the wheel swing allowed in the field wheel operating state, and n° is set according to the actual operating conditions in the field.

[0054] The risk level is the level determined by comparing the wheel deflection angle with the preset warning rule. The preset warning rule is:

[0055] When 0°≤wheel deflection angle≤n°, it means that the wheel swing is within the allowed range, and it is determined to be in the normal swing range.

[0056] When n°<wheel deflection angle≤2n°, it is determined to be in a low risk level, and attention should be paid in the next operation.

[0057] When the wheel deflection angle is greater than 2n°, it is determined to be in a high risk level, and personnel should immediately go to handle it.

[0058] Please refer to Figure 2In this embodiment, the method for monitoring and early warning of sintering machine trolley wheel sway includes an image recorder to acquire images of the trolley wheel's running status and reference position. Specifically, it includes a wheel sway monitoring camera and a side plate number recording camera. The wheel sway monitoring camera is installed above the sintering machine trolley wheel to capture video frame images of the trolley wheel; the side plate number recording camera captures video images of the trolley side plates during operation and records the side plate numbers. A supplementary lighting device is provided to illuminate the image recorder's recording. This example also includes a switch, an image processing server, and a display terminal. The switch is communicatively connected to the image recorder and the image processing server, and the display terminal is communicatively connected to the image processing server. The switch is used for transmitting video signal data between the video recorder and the image processing server. The image processing server has a video image processing system that processes the acquired video images using algorithms. It processes the monitored images through a trained model, identifies the wheels, determines their deflection angle, and issues corresponding alarm prompts based on the risk level. The display terminal shows the alarm prompts, the running status information of the trolley wheels at the time of the alarm (e.g., the corresponding monitoring effect diagram), and the reference position information of the trolley's movement (e.g., side plate number). Furthermore, considering the randomness and speed of wheel failures, a single detection may miss faulty wheels. Therefore, a method of multiple detections and verifications within the video recorder's field of view is used to determine the sintering machine trolley wheel sway. Specifically, the middle area within the camera's field of view is selected, and the acquisition time threshold t is set according to the wheel's running speed v. v Every t v The system acquires an image of the wheel once per second to detect the wheel's deflection angle. This angle is then compared with a preset warning rule to generate an alarm.

[0059] In practical applications, during platform setup, the video recorder includes two wheel sway monitoring cameras and two side panel numbering cameras. The wheel sway monitoring cameras are industrial cameras, while the side panel numbering cameras are standard high-definition cameras. The wheel sway monitoring cameras are mounted on brackets directly above the trolley wheels, and the video footage they acquire is as follows: Figure 3 As shown, the camera captures real-time footage of the trolley wheels' movement. A side-plate number recording camera is installed opposite the side plate of the support frame, recording the side plate number corresponding to each wheel in real time. A supplementary lighting device is installed next to the wheel sway monitoring camera on the support frame; in poor lighting conditions, the supplementary lighting device operates to provide a stable light source.

[0060] Please continue reading. Figure 3 , Figure 4The wheel swing monitoring camera shoots the running state of the wheel from directly above. The wheel swing monitoring camera is used to monitor the swing of the wheel in real time. A straight line detection algorithm is used to detect a fixed linear object in parallel with the side of the wheel in the design state, a reference line is constructed, and the angle between the reference line and the horizontal line is determined. The angle between the wheel and the horizontal line is also determined. The relative angle between the wheel and the reference line is obtained. Specifically, by setting a reference line, and using the OpenCV minimum enclosing rectangle box function to form a minimum enclosing rectangle box around the target mask image after semantic segmentation of the model, the angle between the long side of the minimum enclosing rectangle box and the reference line is obtained, the wheel deflection angle is obtained, compared with the preset risk threshold, and corresponding measures are taken according to different levels. Please refer to Figure 2 、 Figure 3 、 Figure 4 , Figure 5 After the wheel image is shot by the wheel swing monitoring camera above and sent back to the image processing server, the image processing server processes the obtained image, the main process including segmenting the wheel region of the image, and performing mask processing, grayscale, binarization, and constructing a minimum enclosing rectangle box on the wheel region. The relative angle between the long side of the minimum enclosing rectangle box and the reference line is used to obtain the wheel deflection angle. If a wheel swing fault is detected, a warning process is performed according to the corresponding risk level, the side plate number recording camera records the corresponding side plate number, and the fault picture and the side plate number corresponding to the fault point are sent to the command center for manual review.

[0061] Please refer to Figure 6 In the geometric positional relationship between the minimum enclosing rectangle box of the wheel region and the reference line established by the extension direction of the reference line, the deflection angle of the wheel can be calculated specifically by the change amount of the angle between the long side of the minimum enclosing rectangle box of the wheel and the reference line, thereby forming a clear wheel boundary relationship. According to the actual running state of the site, n° can be set as the maximum value of the wheel swing allowed in the running state of the wheel on site. When the deflection angle is greater than n°, the risk level of the wheel working condition is determined by the size of the angle between the minimum enclosing rectangle box and the reference line. Generally, the greater the change amount of the angle between the minimum enclosing rectangle box of the wheel region and the reference line, the higher the fault risk level, and timely treatment is needed.

[0062] For example, in the sintering process of iron smelting, when the sintering machine trolley runs on the guide rail track, the wheel swing monitoring camera above the trolley wheel obtains the real-time video image of the on-site wheel running, and the side plate number recording camera beside the wheel records the corresponding side plate number in real time. The wheel video image is transmitted to the video image processing server through the switch, processed by the semantic segmentation model, and the segmentation effect diagram of the wheel area of each image is obtained. The wheel area is masked, and image processing algorithms are used for grayscale and binary processing. The processed wheel picture is used to construct the reference line and the minimum enclosing rectangle surrounding frame of the wheel area, and the relative angle between the long side of the minimum enclosing rectangle frame and the reference line is used to measure the wheel deflection angle. When the wheel deflection angle is greater than the preset risk threshold, an alarm prompt is given. When a fault is detected, the image processing server transmits the detection effect diagram, the alarm prompt and the side plate number recorded by the side plate number recording camera to the display terminal, and guides personnel to process, so as to guide the industrial production.

[0063] In summary, the sintering machine trolley wheel swing monitoring and early warning method provided by the present application can analyze the real-time collected sintering machine trolley wheel image, determine whether the sintering machine trolley wheel has deviated, improve the detection efficiency, and avoid safety hazards caused by manual detection negligence.

[0064] Finally, it should be noted that: obviously, the above embodiments are only examples for clearly illustrating the present application, and are not limitations of the embodiments. For ordinary skilled persons in the art, other different forms of changes or variations can be made on the basis of the above description. Here, it is not necessary and impossible to enumerate all the embodiments. The obvious changes or variations derived therefrom are still within the protection scope of the present application.

Claims

1. A method for monitoring and early warning of sintering machine trolley wheel oscillation, characterized in that, The application relates to a sintering machine trolley wheel swing monitoring and early warning method. An image set of a trolley wheel under a production scene is acquired, a semantic segmentation model is trained based on the image set, and a trained model is obtained; A trained model is used to process a monitoring image, a wheel region in the monitoring image of the on-site trolley is subjected to segmentation mask processing, and a wheel deflection angle is acquired; Whether the wheel deflection angle is in a normal running state is judged based on a set rule, and the wheel deflection angle is classified and processed according to a risk level; According to the risk level of the wheel deflection angle, corresponding alarm is made; The method for acquiring the wheel deflection angle comprises the following steps: A trained model is used to segment and mask process the wheel in the monitoring image, the wheel mask region is separated from the background, a binary image is obtained, and image noise is eliminated; An external rectangular frame of the wheel is constructed in the binary image, the geometric relationship of the long side of the external rectangular frame of the trolley wheel relative to a reference line is acquired, and the deflection angle of the wheel is obtained; The reference line is determined by using a straight line detection algorithm to detect a fixed linear object in parallel with the side of the wheel under the design running state of the monitoring image, and the method comprises the following steps: The linear monitoring algorithm detects the set of line segments {l1,…,l} of a fixed linear object parallel to the side of the wheel during the design and operation state. n Based on the actual operating conditions on site, one of the l lines was selected. m The straight line segment is used as the baseline, and the coordinates of the two endpoints of the baseline are (x, y, y) and (x, y, y). m ,y m ), (x m+1 ,y m+1 ); The external rectangular frame of the wheel is constructed in the binary image, the geometric relationship of the long side of the external rectangular frame of the trolley wheel relative to the reference line is acquired, and the deflection angle of the wheel is obtained, and the method comprises the following steps: From the reference line l m The corresponding two end point coordinates (x m ,y m ), (x m+1 ,y m+1 ) are obtained l m The slope k m relative to the horizontal line: By l m The slope k of the horizontal line m , get l m The arc value of the horizontal line And get the reference line l m The angle θ of the horizontal line m : where θ m ranges from [-90°, 0°] and [0°, 90°], where negative numbers indicate counterclockwise rotation and positive numbers indicate clockwise rotation; through the angle θ of the reference line and the horizontal line m and the angle a of the long side of the circumscribed rectangular frame and the horizontal line, a is in the range [-90°, 0°] and [0°, 90°], the relative angle β = |a - θ| between the long side of the circumscribed rectangular frame and the reference line is calculated m | 2. The method of monitoring and warning of sintering machine trolley wheel oscillation as claimed in claim 1, wherein, The normal running state is a state in which the deflection angle between the long side of the external rectangular frame of the wheel and the reference line is not greater than n°, wherein n° is the maximum value of the allowable swing of the wheel under the on-site wheel running state, and n° is set according to the actual running state of the scene.

3. The method of monitoring and warning of sintering machine trolley wheel oscillation as claimed in claim 2, wherein, The risk level is a level determined by comparing the wheel deflection angle with a preset early warning rule.

4. The method of monitoring and warning of sintering machine trolley wheel oscillation as claimed in claim 3, wherein, The preset early warning rule is as follows: When 0°<= the wheel deflection angle <= n°, the wheel is determined to be in a normal swing range, When n°< the wheel deflection angle <= 2n°, the wheel is determined to be in a low risk level, and attention should be paid to the wheel in the next running, When the wheel deflection angle > 2n°, the wheel is determined to be in a high risk level, and personnel should immediately go to the scene to handle the wheel.

5. The method of monitoring and warning of sintering machine trolley wheel oscillation as claimed in claim 4, wherein, In the sintering machine trolley wheel swing monitoring and early warning method, an image recorder is arranged to acquire the running state image of the on-site trolley wheel and the reference position image of the trolley running, a light supplementing device is arranged to provide illumination for the image recording of the image recorder, an exchange, an image processing server and a display terminal are arranged, the exchange is in communication connection with the image recorder and the image processing server, and the display terminal is in communication connection with the image processing server.

6. The method of monitoring and warning of sintering machine trolley wheel oscillation as claimed in claim 5, wherein, The image processing server processes the monitoring image through the trained model, identifies the wheel, judges the wheel deflection angle, and makes an alarm prompt according to the risk level, and the display terminal displays the alarm prompt, the running state information of the on-site trolley wheel when the alarm is triggered and the reference position information of the trolley running.

7. The method of monitoring and warning of sintering machine trolley wheel oscillation as claimed in claim 6, wherein, In the corresponding alarm according to the risk level of the wheel deflection angle, through multiple determinations, the middle region position in the camera field of view is selected, the collection time threshold t is set according to the wheel running speed v v , and the wheel is detected once every t v time to collect an image, obtain the wheel deflection angle in the collected image, and make a corresponding alarm by comparing with the preset early warning rule.

Images