Material loading control method and system based on image recognition and storage medium

Abstract

The invention relates to the technical field of Internet of Things, and discloses a material loading control method and system based on image recognition and a storage medium, and the method comprises the steps: recognizing a first target of a loading region through real-time image data, extracting a local gray cloud region around the first target, and calculating the area; if the area exceeds the standard, further identifying a global grey cloud region, and recording the initial area and the outline of the global grey cloud region; and updating the grey cloud data after a preset time interval, and calculating the diffusion velocity and displacement. And respectively generating first diffusion data and second diffusion data by combining the reference values of the speed and the displacement, and comprehensively obtaining an ash cloud diffusion value. If the value is lower than a threshold value, triggering ash emitting early warning; and if the standard exceeds the standard, an ash emitting state is determined and loading is stopped. The method can timely and accurately detect the ash emitting state of the superfine slag powder, and ensures the safety and high efficiency of loading.

Description

Technical Field

[0001] The present application relates to the technical field of the Internet of Things, and in particular, to a material loading control method, system, and storage medium based on image recognition. Background Art

[0002] Slag powder, also known as granulated blast furnace slag powder, is abbreviated as slag powder or water slag powder. It is a fine powdery substance obtained by drying and grinding the slag generated during the smelting of pig iron in steel mills. With the enhancement of environmental awareness and the increasing demand for high-performance materials in the building materials industry, the market prospect of slag powder is very broad.

[0003] Slag powder plays many important roles in fields such as construction. Adding it to concrete can increase fluidity, improve workability, enhance mechanical properties, and improve durability. Using bulk tank trucks is the most common way to load slag powder in bulk. Through special bulk cement tank trucks or powder and granular material transport trucks, the slag powder is transported from the storage tank to the tank truck using a pneumatic conveying system. During the loading process, it is necessary to ensure that the conveying pipeline is tightly connected to the inlet of the tank truck to prevent leakage. At the same time, the conveying speed and pressure should be controlled to avoid the overflow of slag powder due to excessive pressure or too fast conveying speed.

[0004] However, during the loading process, the staff not only needs to observe the state of the inlet to prevent the ash of slag powder from escaping, but also needs to operate the feeding speed and feeding amount in front of the equipment. Once the ash escape state cannot be observed in time, and when the staff sees the slag powder in front of the equipment and quickly stops the equipment, a large amount of slag powder has already overflowed, and these overflowed slag powders will fill the entire working space, making the cleaning work very troublesome. Summary of the Invention

[0005] In order to be able to automatically detect the ash escape state of slag powder and trigger the equipment to stop urgently earlier, the present application provides a material loading control method, system, and storage medium based on image recognition.

[0006] In a first aspect, the present application provides a material loading control method, system, and storage medium based on image recognition, adopting the following technical solutions:

[0007] A material loading control method based on image recognition includes the following steps:

[0008] Obtain image data in real time and identify a first target from the image data;

[0009] Extract a local ash cloud area around the first target within a preset first detection range;

[0010] Calculate the first area of the local ash cloud area;

[0011] If the area of the first region is greater than a preset reference area value, a global grey cloud region is identified from the image data;

[0012] Extract the first area and the corresponding first contour of the global grey cloud region;

[0013] After a preset time period T, update the global grey cloud region, and extract the second area and the corresponding second contour of the global grey cloud region;

[0014] Calculate the diffusion speed of the grey cloud according to the first area and the second area;

[0015] Calculate the diffusion displacement of the grey cloud according to the first contour and the second contour;

[0016] Calculate the first diffusion data according to the diffusion speed and a preset speed reference value; calculate the second diffusion data according to the diffusion displacement and a preset displacement reference value, and calculate the grey cloud diffusion value according to the first diffusion data and the second diffusion data;

[0017] If the grey cloud diffusion value is less than a preset diffusion reference value, a grey cloud warning is prompted; otherwise, a grey cloud state is prompted and a stop loading instruction is output.

[0018] By adopting the above technical solution, through real-time image analysis technology, the local grey cloud region around the feed inlet is automatically identified, and the diffusion speed and displacement of the grey cloud are dynamically tracked, realizing the accurate monitoring of the grey cloud state of slag powder. Compared with manual observation, the system can detect the initial grey cloud sign earlier and avoid large-scale leakage caused by reaction delay. The diffusion value is calculated by integrating the dual parameters of the area change (diffusion speed) and contour displacement (diffusion direction) of the grey cloud, and the warning or emergency stop instruction is triggered by preset thresholds. This judgment logic of multi-index fusion significantly improves the detection accuracy and reduces the risk of misjudgment.

[0019] Optionally, in the step of calculating the diffusion speed of the grey cloud according to the first area and the second area, and calculating the diffusion displacement of the grey cloud according to the first contour and the second contour, the following sub-steps are included:

[0020] Diffusion speed = (second area - first area) / preset time period T;

[0021] Establish a coordinate system with the center of the second target as the starting point;

[0022] According to the first contour and the second contour in the coordinate system, calculate the increment of the second contour relative to the first contour on each coordinate axis, and the diffusion displacement is the sum of all increments.

[0023] By adopting the above technical solution, by precisely calculating the growth rate (diffusion speed) of the ash cloud area and the spatial displacement amount (diffusion displacement) of the contour within a unit time, the originally blurred visual ash-emitting phenomenon is transformed into quantifiable dynamic parameters. A reference system is established with the center of the second target (such as the tanker's feed inlet) as the origin of the coordinate system, so that the contour displacement calculation is always based on the key components of the loading process, eliminating measurement errors caused by vehicle position offset or camera angle change. Using the summation of contour coordinate increments instead of pixel-level comparison reduces the calculation amount while ensuring accuracy, meeting the requirements for real-time response during the loading process.

[0024] Optionally, in the step of calculating the first diffusion data according to the diffusion speed and a preset speed reference value, calculating the second diffusion data according to the diffusion displacement and a preset displacement reference value, and calculating the ash cloud diffusion value according to the first diffusion data and the second diffusion data, the following sub-steps are included:

[0025] First diffusion data = diffusion speed / speed reference value;

[0026] Second diffusion data = diffusion displacement / displacement reference value;

[0027] Ash cloud diffusion value = a1 × first diffusion data + a2 × second diffusion data, where a1 and a2 are weight coefficients.

[0028] By adopting the above technical solution, using a calculation method of normalization and weighted fusion, multi-dimensional diffusion characteristics are transformed into a single decision-making index, while ensuring the detection accuracy, taking into account flexibility, anti-interference, and real-time performance, providing a reliable intelligent anti-leakage guarantee for the slag powder loading process.

[0029] Optionally, in the step of calculating the area of the ash cloud region, the following sub-steps are further included:

[0030] Identifying the second target from the image data;

[0031] After emitting multiple uniformly parallel distributed laser beams towards the second target, obtaining a diffuse reflection image;

[0032] Identifying the region of the diffuse reflection beams from the diffuse reflection image and counting the total length of the region of the diffuse reflection beams;

[0033] Estimating the area of the ash cloud region according to the total length of the region and the diameter of the laser beam.

[0034] By adopting the above technical solution, the accuracy and anti-interference ability of dust detection are enhanced through the characteristics of the laser beam. A clear spot profile can still be maintained in a dusty environment, and it is less affected by environmental light changes (such as shadows and backlighting) compared with traditional RGB image recognition. By detecting the scattering degree of the laser beam, the density of the ash cloud can be more accurately quantified, avoiding misjudgment caused by relying solely on visual gray-scale changes.

[0035] Optionally, in the step of calculating the area of the ash cloud region, the following sub-steps are further included:

[0036] Adjust the diameter of the laser beam according to the historical data of the area; the smaller the area of the ash cloud region in the historical data, the larger the diameter of the laser beam; the larger the area of the ash cloud region in the historical data, the smaller the diameter of the laser beam.

[0037] By adopting the above technical solution, when the historical data shows that the ash cloud region is small, such as in the initial ash-emitting stage, the diameter of the laser beam is automatically increased to enhance the diffuse reflection signal of weak dust and improve the detection sensitivity for small-scale leaks. When a large-area ash cloud is detected, the diameter of the laser beam is automatically decreased to avoid signal saturation caused by excessive scattering and ensure the measurement accuracy under high-concentration dust.

[0038] Optionally, in the step of emitting the laser beam towards the second target, the following sub-steps are further included:

[0039] There are N laser beams, and the N laser beams are scanned back and forth crosswise at random angles;

[0040] After emitting N laser beams towards the second target, a diffuse reflection image is obtained;

[0041] Identify the region of the diffuse reflection beam from the diffuse reflection image, and overlap the regions of the diffuse reflection beam within at least half a scanning period to fit the contour of the comprehensive region; calculate the comprehensive area of the comprehensive region to correct the ash cloud area.

[0042] By adopting the above technical solution, through multi-angle random cross-scanning and dynamic contour fitting technology, the measurement error caused by the dynamic diffusion of dust is effectively compensated. The accuracy of monitoring the dust diffusion during the loading process is improved.

[0043] Optionally, the method further includes the following steps:

[0044] Adjust the length of the preset time period T in an inverse correlation with the frequency of the ash-emitting warning; the higher the frequency of the ash-emitting warning, the shorter the length of the preset time period T; the lower the frequency of the ash-emitting warning, the longer the length of the preset time period T;

[0045] Adjust the diffusion reference value in positive correlation with the area of the region. The larger the area of the region, the larger the diffusion reference value; the smaller the area of the region, the smaller the diffusion reference value.

[0046] By adopting the above technical solution, through high-frequency early warning, T is automatically shortened, the system response speed is increased, and sudden leaks are quickly captured; through low-frequency early warning, T is extended, the calculation load is reduced, and false alarms are avoided. Through small-area gray clouds, the reference value is reduced, enhancing the sensitivity to initial ash emission; through large-area gray clouds, the reference value is increased, avoiding misjudgment of normal operation dust.

[0047] In a second aspect, the present application provides a material loading control system based on image recognition, adopting the following technical solution:

[0048] A material loading control system based on image recognition includes a processor, and the processor executes the steps of the material loading control method based on image recognition as described in any one of the above.

[0049] In a third aspect, the present application provides a storage medium, adopting the following technical solution:

[0050] A storage medium stores a program, and when the program is executed by a processor, it implements the steps of the material loading control method based on image recognition as described in any one of the above.

[0051] In summary, the present application includes at least one of the following beneficial technical effects:

[0052] Precise monitoring: Through real-time image analysis technology, automatically identify the local gray cloud area around the feed inlet, and dynamically track the diffusion speed and displacement of the gray cloud, realizing precise monitoring of the ash emission state of slag powder. Compared with manual observation, it can detect the initial ash emission signs earlier and avoid large-scale leaks caused by reaction delay.

[0053] Improve detection accuracy: Calculate the diffusion value by comprehensively considering the dual parameters of the area change (diffusion speed) and contour displacement (diffusion direction) of the gray cloud, and trigger early warning or emergency stop instructions through preset threshold grading. This multi-index fusion judgment logic significantly improves the detection accuracy and reduces the risk of misjudgment.

[0054] Quantify dynamic parameters: Accurately calculate the growth rate of the gray cloud area (diffusion speed) and the spatial displacement of the contour (diffusion displacement) per unit time, convert the originally fuzzy visual ash emission phenomenon into quantifiable dynamic parameters, establish a reference system with the second target center as the coordinate origin, eliminate measurement errors caused by vehicle position offset or camera angle change, and use contour coordinate increment summation to reduce the calculation amount, meeting the real-time response requirements during the loading process.

[0055] Reliable intelligent anti-leakage guarantee: By adopting the calculation method of normalization and weighted fusion, the multi-dimensional diffusion characteristics are transformed into a single decision index, which not only ensures the detection accuracy, but also takes into account flexibility, anti-interference ability and real-time performance, providing a reliable intelligent anti-leakage guarantee for the slag powder loading process.

[0056] Enhanced accuracy and anti-interference ability: Utilizing the characteristics of the laser beam, the accuracy and anti-interference ability of dust detection are enhanced. In a dust environment, a clear spot profile can still be maintained. Compared with traditional RGB image recognition, it is less affected by environmental light changes (such as shadows and backlighting). By detecting the scattering degree of the laser beam, the density of the ash cloud can be more accurately quantified, avoiding misjudgments caused by relying solely on visual gray-scale changes. Description of the Drawings

[0057] Figure 1 It is a step diagram of a material loading control method based on image recognition.

[0058] Figure 2 It is a schematic diagram for calculating the diffusion displacement of the ash cloud area in a coordinate system starting from the center of the second target.

[0059] Figure 3 It is a schematic diagram of multiple uniformly parallel distributed laser beams in the ash cloud area.

[0060] Figure 4 It is a schematic diagram of N laser beams in the ash cloud area scanning back and forth at random angles. Detailed Implementation Manner

[0061] The following details the implementation manners of the present application, and the examples of the implementation manners are shown in the drawings.

[0062] In the description of this specification, the description of reference terms "certain implementation manners", "one implementation manner", "some implementation manners", "schematic implementation manners", "examples", "specific examples" or "some examples" means that the specific features, structures, materials or characteristics described in connection with the implementation manner or example are included in at least one implementation manner or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same implementation manner or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more implementation manners or examples.

[0063] The embodiments of the present application disclose a material loading control method based on image recognition, as Figure 1 shown, including the following steps:

[0064] Obtain image data in real time, identify the first target from the image data; extract the local gray cloud area around the first target within a preset first detection range. The system collects the image data of the loading area in real time through an industrial camera and uses computer vision algorithms (such as YOLO, MaskR-CNN, etc.) to identify the first target (such as the feeding port, the contour of the material pile). Subsequently, a preset first detection range (such as a circular area with a radius of 2 meters) is delimited around the first target, and the local gray cloud area within this range is extracted.

[0065] Calculate the first area of the local gray cloud area; if the first area is greater than the preset reference area value, then identify the global gray cloud area from the image data. Perform binary processing and contour extraction on the local gray cloud area, and calculate its first area (S1). If S1 exceeds the preset reference area value (such as 5000 pixels, corresponding to an actual 0.1m 2 ), it indicates that there may be an initial gray dust emission phenomenon, and the system triggers the global monitoring mode.

[0066] Extract the first area and the corresponding first contour of the global gray cloud area; after a preset time period T, update the global gray cloud area, and extract the second area and the corresponding second contour of the global gray cloud area. After entering the global monitoring, the system records the first area (A1) and the first contour (C1) of the current global gray cloud area. After a preset time interval T (such as 5 seconds), update the image data again, and extract the second area (A2) and the second contour (C2); calculate the diffusion speed of the gray cloud according to the first area and the second area; calculate based on the area change rate, and the formula is: V = |A2 - A1| / T.

[0067] Calculate the diffusion displacement of the gray cloud according to the first contour and the second contour; calculate the overall moving distance (D) of the gray cloud through contour matching (such as optical flow method or contour centroid offset).

[0068] Calculate the first diffusion data according to the diffusion speed and the preset speed reference value; calculate the second diffusion data according to the diffusion displacement and the preset displacement reference value, and calculate the gray cloud diffusion value according to the first diffusion data and the second diffusion data. Compare the diffusion speed (V) and the displacement (D) with the speed reference value (V_ref) and the displacement reference value (D_ref) respectively, and calculate the gray cloud diffusion value (K) after normalization and weighting. K = a1 × V / V_ref + a2 × D / D_ref, where a1 and a2 are weight coefficients, such as a1 = 0.6 and a2 = 0.4. Adopt the calculation method of normalization and weighted fusion to convert multi-dimensional diffusion features into a single decision index, which not only ensures the detection accuracy, but also takes into account flexibility, anti-interference and real-time performance, providing a reliable intelligent anti-leakage guarantee for the slag powder loading process.

[0069] If the ash cloud diffusion value is less than the preset diffusion reference value, an ash leakage warning is prompted; otherwise, an ash leakage state is prompted and a stop loading instruction is output. K < diffusion reference value K_ref: triggers an ash leakage warning (such as an audible and visual alarm) to prompt the operator to check; K ≥ K_ref: it is judged as serious ash leakage, and an immediate stop loading instruction is output, and the feed port is linked to close.

[0070] Through the rapid identification of local ash clouds, intervention can be made at the initial stage of ash emission to avoid the lag of manual observation. Comprehensive area (diffusion intensity) and contour displacement (diffusion trend) can be used to avoid single parameter misjudgment (such as area fluctuation caused by light interference). The time interval T can be adjusted according to the material characteristics (such as shortening T if slag powder diffuses quickly) to improve the flexibility of the system. Direct linkage loading equipment emergency stop forms a closed-loop management to reduce the risk of environmental pollution.

[0071] like Figure 2 As shown in Figure 1, the diffusion rate is used to quantify the volume expansion trend of the ash cloud per unit time and reflect the severity of the ash emission. The specific sub-steps are as follows:

[0072] Diffusion speed V = (second area A2 - first area A1) / preset time period T; if V is positive and large, it indicates that the ash cloud spreads rapidly; if V is close to zero or a negative value (such as the dust removal equipment is started), the risk is low.

[0073] Establish a coordinate system with the center of the second target as the starting point; establish a local rectangular coordinate system with the second target (such as the center of the tanker feed port flange) as the origin O(0,0). Track the origin position in real time through target detection to ensure that the coordinate system is dynamically updated as the vehicle moves, avoiding benchmark drift caused by tanker displacement during loading.

[0074] According to the first contour and the second contour in the coordinate system, the first contour is the area shown by the circular dotted line; the second contour is the area shown by the circular solid line. The increment of the second contour on each coordinate axis relative to the first contour is calculated, and the coordinate increment in each quadrant is displayed by the dotted straight line; the diffusion displacement is the sum of all increments. The absolute value of the increments of all points is summed to eliminate the direction cancellation effect and reduce the complexity of the calculation.

[0075] By accurately calculating the growth rate of the ash cloud area per unit time (diffusion rate) and the spatial displacement of the contour (diffusion displacement), the originally fuzzy visual ash phenomenon is converted into quantifiable dynamic parameters. These parameters can more accurately and intuitively reflect the diffusion of the ash cloud. Establishing a reference system with the center of the second target as the origin of the coordinate system and using the method of summing the contour coordinate increments not only improves the accuracy of the measurement, but also reduces the amount of calculation, making the entire monitoring system more stable and efficient in practical applications, and can better ensure the safety and smooth progress of the material loading process.

[0076] In the steps of calculating the area of the ash cloud region, the following sub-steps are further included:

[0077] Identify the second target from the image data; the second target is the inlet of the tanker. Determine the geometric center position of the inlet of the tanker through high-precision vision recognition (such as Hough transform or template matching), which serves as the reference point for laser emission. Synchronously obtain the attitude data of the inlet (such as pitch angle, yaw angle) for calibrating the projection direction of the laser beam.

[0078] After emitting multiple uniformly parallel distributed laser beams towards the second target, obtain the diffuse reflection image; emit multiple groups of parallel laser beams (such as 16 beams, 905nm infrared laser with a wavelength) from the center point of the inlet to form a detection grid covering the diffusion path of the ash cloud. As Figure 3 shown, for example, two relatively close straight lines are the presentation diagrams of the diffuse reflection of the laser beams. The distance between the laser beams is dynamically adjusted according to the size of the tanker (default 10cm) to ensure coverage of the effective monitoring area.

[0079] Identify the region of the diffuse reflection beam from the diffuse reflection image and count the total length of the region of the diffuse reflection beam; use a narrow-band filter camera to only receive the laser wavelength (such as 905±5nm) to suppress the interference of ambient light (sunlight, lamp light). The exposure time is adaptively adjusted to avoid overexposure or underexposure of the laser spots.

[0080] Estimate the area of the ash cloud region based on the total length of the region and the diameter d of the laser beam. Perform skeleton extraction on the diffuse reflection region of each laser beam to obtain the total scattering length L of a series of length values. Based on the laser beam diameter d (known hardware parameter) and the scattering length, estimate the equivalent area of the ash cloud: S = k×d×L, where k is the calibration coefficient (determined through experiments, usually 0.8 - 1.2) to compensate for the scattering projection error.

[0081] Laser beams have high directivity and monochromaticity, and can maintain a clear spot contour in a dust environment, which enables the system to more accurately identify the region of the diffuse reflection beam, thereby improving the detection accuracy. Compared with the traditional RGB image recognition method, laser beams are much less affected by changes in ambient light. In traditional RGB image recognition, changes in ambient light conditions such as shadows and backlighting will cause changes in the color and brightness of objects in the image, thus affecting the recognition and judgment of the ash cloud region and prone to misjudgment.

[0082] In the steps of calculating the area of the ash cloud region, the following sub-steps are further included:

[0083] Adjust the diameter of the laser beam according to the historical data of the area of the region; the smaller the area of the ash cloud region in the historical data, the larger the diameter of the laser beam; the larger the area of the ash cloud region in the historical data, the smaller the diameter of the laser beam.

[0084] In small-area ash cloud (initial ash emission): when historical data shows that the area is below the threshold, such as 0.05m 2 When the laser is controlled to increase the beam diameter from the default value of 5mm to 10mm, the beam cross-sectional area is increased, the diffuse reflection signal intensity of low-concentration dust is enhanced, and the sensitivity is improved.

[0085] In large-scale ash cloud (diffusion period): when the area exceeds a threshold such as 1m 2 When the laser beam is scattering, the diameter is reduced to 2 mm. Function: To avoid signal saturation caused by excessive laser scattering under high dust concentration, such as the fusion of light spots into pieces, and ensure that the contour can be resolved.

[0086] By dynamically adjusting the diameter of the laser beam according to the historical data of the area, the entire material loading control method has stronger adaptability and accuracy. At different stages of ash cloud diffusion, the system can automatically adjust the parameters of the laser beam to achieve the best detection effect. This adaptive adjustment mechanism not only improves the detection capability of small-scale leakage such as initial ash emission, but also ensures the measurement accuracy in large-scale ash emission. It provides reliable protection for ash emission monitoring during material loading, effectively avoids problems such as material leakage caused by inaccurate detection, and improves the safety and efficiency of the entire loading operation.

[0087] like Figure 4 As shown, the step of emitting a laser beam toward the second target further includes the following sub-steps:

[0088] There are N laser beams, and the N laser beams perform cross-reciprocating scanning at random angles; N laser emitters (N≥8, 16 beams are recommended) are set up and distributed in a ring around the second target (tank truck feed port), and the initial angle of each laser beam is randomly generated (to avoid periodic scanning blind spots). Each laser beam scans in a sinusoidal reciprocating mode, with a horizontal swing angle range of ±30° and a vertical swing of ±15° to form cross coverage. The scanning frequency is set to 10Hz to ensure that the area coverage is completed within half a cycle (50ms). In the figure, the two adjacent parallel lines are the diffuse reflection presentation of the laser beam, and there are multiple laser beams staggered in the figure.

[0089] After emitting N laser beams toward the second target, a diffuse reflection image is obtained; the area of ​​the diffuse reflection beam is identified from the diffuse reflection image, and the area of ​​the diffuse reflection beam within at least half a scanning cycle is overlapped to fit the outline of the comprehensive area; the comprehensive area of ​​the comprehensive area is calculated to correct the area of ​​the ash cloud. Within half a scanning cycle (50ms), multiple frames of diffuse reflection images (such as 5 frames) are continuously collected. Compared with the traditional single-time, single-angle measurement method, it can more accurately reflect the actual area and diffusion range of the ash cloud during the loading process, and effectively compensate for the measurement error caused by the dynamic diffusion of dust.

[0090] Through multi-angle random cross-scanning and dynamic contour fitting technology, the system can comprehensively and accurately capture the diffusion of dust during the loading process, greatly improving the accuracy of monitoring dust diffusion during the loading process, and providing strong data support and technical guarantee for timely detection of dust risks and ensuring the safety of material loading.

[0091] The method further comprises the steps of:

[0092] According to the frequency of dust warning, the length of the preset time period T is adjusted inversely; the higher the frequency of dust warning, the shorter the length of the preset time period T; the lower the frequency of dust warning, the longer the length of the preset time period T. The exponential decay algorithm is used to dynamically adjust the monitoring interval T to T1, T1 = T×e -kM , where T is the reference interval, such as 5 seconds; k is the attenuation coefficient, which is calibrated by historical data; and M is the number of dust warnings per unit time, such as the number of warnings per minute.

[0093] The diffusion reference value K_ref is adjusted according to the positive correlation of the regional area. The larger the regional area S is, the larger the diffusion reference value is; the smaller the regional area S is, the smaller the diffusion reference value is.

[0094] The piecewise function relationship between the ash cloud area S and the diffusion reference value K_ref is established:

[0095]

[0096] Example: S1 = 0.1 m 2 ,S2=1m 2 ,Kmin=0.3,Kmax=0.8

[0097] When a small area of ​​ash cloud (S<0.1m 2 ):

[0098] Set K_ref to the lowest value (0.3) to enhance the sensitivity to initial dust emission. Experiments show that 0.05m 2 Minor leakage.

[0099] Large area of ​​ash cloud (S>1m 2 ):

[0100] Increase K_ref to 0.8 to avoid false stops caused by normal dust during loading (such as dust from material drop), and reduce the false alarm rate by 60%.

[0101] By dynamically adjusting the preset time period T based on the ash emission warning frequency and dynamically adjusting the diffusion reference value based on the area of the region, this material loading control method can flexibly optimize system parameters under different loading conditions, achieve precise monitoring and efficient processing of ash emission situations, and provide a solid guarantee for the safe and stable operation of the material loading process.

[0102] In other embodiments, the step of calculating the diffusion displacement of the ash cloud includes the following sub-steps:

[0103] Based on two cameras, first image data and second image data are respectively obtained. Two cameras with fixed positions are used and installed at different positions at the loading site, such as on both sides symmetric to the feeding pipe, to capture the ash cloud from different angles. In this way, the two cameras respectively obtain the first image data and the second image data, and these data contain the characteristic information of the ash cloud from different perspectives. The same effect can be achieved by using a displaceable camera installed on a movable device.

[0104] First targets are respectively identified from the first image data and the second image data, and then the two images are combined to obtain an image with a more comprehensive angle, so as to calculate the diffusion displacement. After obtaining the first image data and the second image data, these two sets of data need to be processed: using an image recognition algorithm, first targets are respectively identified from the first image data and the second image data. The two images after identifying the first targets are combined to obtain an image with a more comprehensive angle. During the splicing process, common feature points in the two images need to be found, such as certain fixed equipment, buildings, etc. Through these feature points, the relative positions and directions of the two images are determined, and then precise splicing is carried out. The comprehensive angle image obtained in this way contains the information of the ash cloud from multiple perspectives. For the combined comprehensive angle image, the diffusion displacement of the ash cloud can be calculated more accurately. Determine the first contour of the ash cloud in the initial image and record its position information in the image coordinate system. After a preset time period, extract the second contour of the ash cloud from the combined image again and record its position information. By comparing the position changes of the first contour and the second contour in the image coordinate system, the displacement amounts of the ash cloud in each direction can be calculated. In order to obtain a more accurate actual diffusion displacement, the image data also needs to be calibrated and converted. Considering factors such as the installation position of the camera, the shooting angle, and the image scale, the displacement amount in the image is converted into the actual physical displacement amount.

[0105] The embodiment of the present application also discloses a material loading control system based on image recognition, including a processor, and the processor executes the steps of the material loading control method based on image recognition as described in any one of the above.

[0106] The embodiment of the present application also discloses a storage medium, in which a program is stored, and when the program is executed by a processor, the steps of the material loading control method based on image recognition described in any one of the above are implemented.

[0107] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A material loading control method based on image recognition, characterized in that It includes the following steps: Obtain image data in real time and identify the first target from the image data; Extract the local gray cloud area within a preset first detection range around the first target; Calculate the first area of the local gray cloud area; If the first area is greater than the preset reference area value, identify the global gray cloud area from the image data; Extract the first area and the corresponding first contour of the global gray cloud area; After a preset time period T, update the global gray cloud area, and extract the second area and the corresponding second contour of the global gray cloud area; Calculate the diffusion speed of the gray cloud according to the first area and the second area; Calculate the diffusion displacement of the gray cloud according to the first contour and the second contour; Calculate the first diffusion data according to the diffusion speed and the preset speed reference value; Calculate the second diffusion data according to the diffusion displacement and the preset displacement reference value, and calculate the gray cloud diffusion value according to the first diffusion data and the second diffusion data; If the gray cloud diffusion value is less than the preset diffusion reference value, prompt a gray cloud warning; Otherwise, prompt the gray cloud state and output a stop loading instruction.

2. The material loading control method based on image recognition according to claim 1, wherein In the step of calculating the diffusion speed of the gray cloud according to the first area and the second area; calculating the diffusion displacement of the gray cloud according to the first contour and the second contour, the following sub-steps are included: Diffusion speed = (Second area - First area) / Preset time period T; Establish a coordinate system with the center of the second target as the starting point; According to the first contour and the second contour in the coordinate system, calculate the increment of the second contour relative to the first contour on each coordinate axis, and the diffusion displacement is the sum of all increments.

3. The material loading control method based on image recognition according to claim 1, wherein In the step of calculating the first diffusion data according to the diffusion speed and the preset speed reference value; calculating the second diffusion data according to the diffusion displacement and the preset displacement reference value, and calculating the gray cloud diffusion value according to the first diffusion data and the second diffusion data, the following sub-steps are included: First diffusion data = Diffusion speed / Speed reference value; Second diffusion data = Diffusion displacement / Displacement reference value; Gray cloud diffusion value = a1 × First diffusion data + a2 × Second diffusion data, where a1 and a2 are weight coefficients.

4. The material loading control method based on image recognition according to claim 1, wherein In the step of calculating the area of the gray cloud area, the following sub-steps are further included: Identify the second target from the image data; After emitting multiple uniformly parallel distributed laser beams towards the second target, obtain a diffuse reflection image; Identify the area of the diffuse reflection beam from the diffuse reflection image and count the total length of the area of the diffuse reflection beam; Estimate the area of the gray cloud area according to the total length of the area and the diameter of the laser beam.

5. The material loading control method based on image recognition according to claim 1, wherein In the step of calculating the area of the gray cloud area, the following sub-steps are further included: Adjust the diameter of the laser beam according to the historical data of the area of the gray cloud area; the smaller the area of the gray cloud area in the historical data, the larger the diameter of the laser beam; the larger the area of the gray cloud area in the historical data, the smaller the diameter of the laser beam.

6. The material loading control method based on image recognition according to claim 1, wherein In the step of emitting the laser beam towards the second target, the following sub-steps are further included: The laser beams are provided with N numbers, and the N laser beams perform cross-reciprocating scanning at random angles; After emitting N laser beams toward the second target, a diffuse reflection image is obtained; The area of ​​the diffuse reflection light beam is identified from the diffuse reflection image, and the area of ​​the diffuse reflection light beam within at least half a scanning period is overlapped to fit the outline of the comprehensive area; the comprehensive area of ​​the comprehensive area is calculated to correct the gray cloud area.

7. The material loading control method based on image recognition according to claim 1, characterized in that The method further comprises the steps of: According to the frequency of the dust warning, the length of the preset time period T is adjusted inversely; the higher the frequency of the dust warning, the shorter the length of the preset time period T; The lower the frequency of the dust warning, the longer the length of the preset time period T; The diffusion reference value is adjusted in a positive correlation with the area of ​​the region. The larger the area of ​​the region, the larger the diffusion reference value; and the smaller the area of ​​the region, the smaller the diffusion reference value.

8. A material loading control system based on image recognition, characterized in that, It includes a processor, in which the steps of the material loading control method based on image recognition as described in any one of claims 1 to 7 are executed.

9. A storage medium, characterized in that, The medium stores a program, and when the program is executed by the processor, the steps of the material loading control method based on image recognition described in any one of claims 1 to 7 are implemented.

Images