A method, device, equipment and medium for identifying abnormal grain surface in a silo
By combining radar and video monitoring modules, a 3D point cloud map is generated and panoramic images are captured. Multi-channel visual algorithms are used to identify grain surface anomalies, solving the problems of blurry imaging and insufficient multi-dimensional detection in existing technologies, and realizing efficient and automated grain surface monitoring.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- MUYUAN FOODS CO LTD
- Filing Date
- 2026-03-17
- Publication Date
- 2026-06-19
AI Technical Summary
Existing technologies cannot dynamically adapt to the imaging blur caused by changes in grain surface height, lack the ability to simultaneously detect multi-dimensional anomalies such as physical deformation and biological deterioration, rely on manual intervention for equipment maintenance, and have insufficient monitoring continuity.
A radar module generates a 3D point cloud map, which is then combined with the optimal focus parameter adjustment and panoramic shooting of the video monitoring module. Multi-channel visual algorithms are used to identify abnormalities on the grain surface, including mold, clumping, pests, and leakage.
It achieves high-precision, real-time identification of grain surface anomalies, reduces the false positive and false negative rates, reduces the need for manual maintenance, and improves the automation and continuity of monitoring.
Smart Images
Figure CN122244789A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of grain surface detection technology, and more specifically, to a method, device, equipment, and medium for identifying abnormal grain surface conditions inside silos. Background Technology
[0002] Currently, grain condition monitoring in silos mainly relies on three technical approaches: manual inspection, fixed monitoring equipment, and single sensors, all of which have significant limitations.
[0003] Manual inspection is the traditional mainstream method, relying on personnel to periodically open the warehouse for visual inspection. This method has the problems of low inspection frequency and high subjectivity. It not only cannot quantitatively assess the temperature, humidity and degree of mold on the grain surface, but also makes it difficult to detect early mold, insect infestation and other microscopic anomalies inside the grain pile. Often, the problem is only discovered after it has worsened.
[0004] To replace manual labor, some grain depots have introduced fixed monitoring equipment for remote observation. However, this solution is not adaptable to dynamic grain storage scenarios: the grain level changes with operations, and fixed-focus equipment cannot maintain a clear image, resulting in the loss of crucial details. Furthermore, dust inside the silos easily contaminates the lenses, and existing equipment lacks an effective automatic cleaning mechanism, still requiring manual maintenance and making it difficult to guarantee long-term, continuous, and stable observation.
[0005] Another approach involves indirectly inferring grain conditions using temperature, humidity, or weight sensors. While these methods can collect environmental parameters, they are indirect inferences and cannot directly identify physical anomalies such as leakage or clumping. Furthermore, the correlation between environmental parameters and grain quality deterioration is delayed and prone to misjudgment, and the information is limited in scope, making it difficult to meet the needs of precise monitoring.
[0006] In summary, existing technologies generally have the following limitations: they cannot dynamically adapt to imaging blur caused by changes in grain surface height; they lack the ability to simultaneously detect multi-dimensional anomalies such as physical deformation and biological deterioration; and equipment maintenance relies on manual intervention, resulting in insufficient monitoring continuity. Summary of the Invention
[0007] In view of this, the purpose of this application is to provide a method, device, equipment and medium for identifying abnormal grain surface conditions in silos, which effectively solves the problem that the existing technology has many limitations that have a variety of negative impacts on grain condition monitoring in silos.
[0008] In a first aspect, embodiments of this application provide a method for identifying abnormal grain levels in silos, applied to an identification system. The identification system includes a radar module, a video monitoring module, and a main control module. The method includes: The radar module is controlled to rotate and emit a high-frequency beam to detect the grain surface inside the silo, obtain a three-dimensional point cloud map of the grain surface inside the silo, and calculate radar height measurement data based on the three-dimensional point cloud map. The main control module matches the radar altimetry data, grain variety, and a preset focal length matching database to determine the optimal focusing parameters for the video monitoring module, and then controls the video monitoring module to adjust according to these optimal focusing parameters; the video monitoring module has been pre-cleaned. The path planning algorithm processes the geometric dimensions of the silo and the current imaging distance of the adjusted video monitoring module to generate a panoramic shooting coordinate sequence. The adjusted video monitoring module is then controlled to capture high-definition images of each sub-area of the silo according to the panoramic shooting coordinate sequence. The high-definition images are quality-screened, and anomaly recognition is performed on the high-definition images selected by the quality screening using a multi-channel vision algorithm to identify abnormalities in the grain surface inside the silo.
[0009] In conjunction with the first aspect, embodiments of this application provide a first possible implementation of the first aspect, wherein the anomaly identification of the target high-definition image selected through quality screening using a multi-channel visual algorithm includes: The corresponding channels are pre-set based on the type of grain surface anomaly; different channels correspond to different visual sub-algorithms; The visual sub-algorithm is run to identify anomalies in the corresponding channels of the target high-definition image and generate anomaly identification results.
[0010] In conjunction with the first aspect, embodiments of this application provide a second possible implementation of the first aspect, wherein running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel includes: The color space of the target high-definition image is converted, and the chromaticity feature components of the converted target high-definition image are extracted. By comparing the colorimetric feature components with the preset standard colorimetric threshold for grains, mold candidate regions are obtained to identify mold abnormalities.
[0011] In conjunction with the first aspect, this application provides a third possible implementation of the first aspect, wherein running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel further includes: Extract the image texture features of the target high-definition image, and match the image texture features with the standard texture features pre-stored in the grain variety feature database to obtain the matching result; Based on the matching results, the corresponding region is determined as a candidate region for clumping, so as to identify clumping anomalies.
[0012] In conjunction with the first aspect, this application provides a fourth possible implementation of the first aspect, wherein running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel also includes: The dynamic frame difference method is applied to the image sequence composed of high-resolution images of the target to extract the contours of the moving target between adjacent frames; By combining a pre-set pest morphology feature model with matching and filtering, dynamic trajectory areas that match the pest activity characteristics are identified to detect pest anomalies.
[0013] In conjunction with the first aspect, this application provides a fifth possible implementation of the first aspect, wherein running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel further includes: Edge detection is performed on the target high-definition image, and the edge lines of the grain surface area are extracted after detection; The detected abnormal edges are spatially registered with the pre-built grain surface settling model to identify leakage candidate areas and thus identify leakage anomalies.
[0014] In conjunction with the first aspect, this application provides a sixth possible implementation of the first aspect, wherein a panoramic shooting coordinate sequence is generated by processing the geometric dimensions of the silo and the current imaging distance of the video monitoring module through a path planning algorithm, including: Based on the geometric dimensions of the silo and the current imaging distance of the video monitoring module, calculate the reference value of the step spacing for a single shot; A Cartesian coordinate system is established within the silo. The coordinates of shooting points covering the entire silo area are calculated point by point at intervals based on the step spacing reference value to generate a panoramic shooting coordinate sequence with a dynamic optimization mechanism.
[0015] Secondly, embodiments of this application provide a grain surface anomaly identification device for silos, applied to an identification system. The identification system includes a radar module, a video monitoring module, and a main control module. The device includes: The detection module is used to control the radar module to rotate and emit high-frequency beams to detect the grain surface inside the silo, obtain a three-dimensional point cloud map of the grain surface inside the silo, and calculate radar height measurement data based on the three-dimensional point cloud map. The matching module is used to match the radar altimetry data, grain variety, and a preset focal length matching database through the main control module to find the optimal focusing parameters for the video monitoring module, and control the video monitoring module to adjust according to the optimal focusing parameters; the video monitoring module has been pre-cleaned. The snapshot module is used to process the geometric dimensions of the silo and the current imaging distance of the adjusted video monitoring module through a path planning algorithm, generate a panoramic shooting coordinate sequence, and control the adjusted video monitoring module to capture high-definition images of each sub-area of the silo according to the panoramic shooting coordinate sequence. The identification module is used to perform quality screening on the high-definition images and to perform anomaly identification on the target high-definition images selected by quality screening through a multi-channel vision algorithm, so as to identify abnormalities in the grain surface inside the silo.
[0016] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of any of the methods for identifying abnormal grain levels in silos are performed.
[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of any of the methods for identifying abnormal grain levels in a silo.
[0018] This application provides a method for identifying abnormal grain surface conditions inside a silo, applied to an identification system. The identification system includes a radar module, a video monitoring module, and a main control module. The method first controls the radar module to rotate and emit a high-frequency beam to detect the grain surface inside the silo, obtaining a three-dimensional point cloud map of the grain surface. Based on the three-dimensional point cloud map, radar height measurement data is calculated. Next, the main control module matches the radar height measurement data, grain type, and a preset focal length matching database to determine the optimal focusing parameters for the video monitoring module. The video monitoring module is then controlled to adjust according to these optimal focusing parameters. The video monitoring module is pre-cleaned. Then, a path planning algorithm processes the silo's geometric dimensions and the adjusted current imaging distance of the video monitoring module to generate a panoramic shooting coordinate sequence. The adjusted video monitoring module is then controlled to capture high-definition images of each sub-region of the silo according to the panoramic shooting coordinate sequence. Finally, the high-definition images undergo quality screening, and a multi-channel vision algorithm is used to identify anomalies in the selected high-definition images to identify abnormal grain surface conditions inside the silo. This application addresses the core pain points of existing grain monitoring technologies in silos through the synergistic effect of hardware reconstruction and algorithm optimization. It significantly improves monitoring accuracy, timeliness, reliability, and automation, providing reliable technical support for the safe and efficient management of grain storage in silos. It also avoids the lagging management model that relies on manual inspections, reduces the rate of misjudgment and missed judgment, promptly detects risks such as grain deterioration and leakage, and reduces grain storage losses. Attached Figure Description
[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0020] Figure 1 The illustration shows a flowchart of a method for identifying abnormal grain levels in a silo, as provided in an embodiment of this application. Figure 2 This illustration shows a photographic anomaly caused by dust obscuring the image, as provided in an embodiment of this application. Figure 3 This illustration shows a snapshot recording provided in an embodiment of this application; Figure 4 This illustration shows a schematic diagram of the snapshot record and corresponding high-definition image provided in an embodiment of this application; Figure 5 A schematic diagram of the target high-definition image provided in an embodiment of this application is shown; Figure 6 This paper presents a structural block diagram of a grain surface anomaly identification device in a silo, according to an embodiment of this application. Figure 7 A structural block diagram of an electronic device provided in an embodiment of this application is shown. Detailed Implementation
[0021] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. It should be understood that the accompanying drawings in this application are for illustrative and descriptive purposes only and are not intended to limit the scope of protection of this application. Furthermore, it should be understood that the schematic drawings are not drawn to scale. The flowcharts used in this application illustrate operations implemented according to some embodiments of this application. It should be understood that the operations in the flowcharts may not be implemented in sequence, and steps without logical contextual relationships may be reversed or implemented simultaneously. In addition, those skilled in the art, guided by the content of this application, may add one or more other operations to the flowcharts, or remove one or more operations from the flowcharts.
[0022] Furthermore, the described embodiments are merely some, not all, of the embodiments of this application. The components of the embodiments of this application described and illustrated herein can typically be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0023] It should be noted that the term "comprising" will be used in the embodiments of this application to indicate the presence of the features declared thereafter, but does not exclude the addition of other features.
[0024] Existing grain monitoring in silos relies heavily on manual sampling, resulting in low accuracy and large errors. Visual monitoring is based on fixed points, and images are easily distorted by dust. Anomaly identification relies on manual methods or single-dimensional algorithms, which are slow and have a high false positive rate. Equipment is prone to failure in high-dust environments, has high maintenance costs, low automation, and is difficult to achieve efficient unattended management.
[0025] Based on this, this application provides a method, apparatus, equipment, and medium for identifying abnormal grain levels in silos, which will be described below through embodiments.
[0026] Example 1 To facilitate understanding of this embodiment, a method for identifying abnormal grain levels in silos, as disclosed in this application, will first be described in detail. For example... Figure 1 The diagram shows a flowchart of a method for identifying abnormal grain surface conditions inside a silo. This application provides a method for identifying abnormal grain surface conditions inside a silo, applied to an identification system. The identification system includes a radar module, a video monitoring module, and a main control module. The method includes: S101. Control the radar module to rotate and emit a high-frequency beam to detect the grain surface inside the silo, obtain a three-dimensional point cloud map of the grain surface inside the silo, and calculate the radar height measurement data based on the three-dimensional point cloud map. S102. The main control module matches the radar altimetry data, grain variety, and a preset focal length matching database to determine the optimal focusing parameters for the video monitoring module, and controls the video monitoring module to adjust according to the optimal focusing parameters; the video monitoring module has been pre-cleaned. S103. Process the geometric dimensions of the silo and the current imaging distance of the adjusted video monitoring module through a path planning algorithm to generate a panoramic shooting coordinate sequence, and control the adjusted video monitoring module to capture high-definition images of each sub-area of the silo according to the panoramic shooting coordinate sequence. S104. The high-definition image is quality-screened, and anomaly recognition is performed on the target high-definition image selected by the quality screening through a multi-channel vision algorithm to identify abnormalities in the grain surface inside the silo.
[0027] In step S101, the identification system installed in the silo of this application is activated daily at a preset time. After activation, the radar module rotates and emits a high-frequency beam covering the entire silo area to comprehensively detect the grain surface inside the silo. It receives the reflected signals to obtain a three-dimensional point cloud map of the grain surface. Based on the three-dimensional point cloud map, spatial gridding is performed. By fitting the curvature changes of the grain surface, radar height measurement data, including the current average height of the grain surface and local concave areas, is calculated. The radar height measurement data, combined with the silo's geometric parameters, is used to deduce the estimated volume and weight of the grain inside the silo. This also assists in gimbal positioning calibration, preventing point offset. Compared to traditional manual sampling measurement methods, this application completely solves the error problem caused by sparse sampling, significantly improves monitoring accuracy, makes grain inventory data more valuable, and provides accurate data support for warehouse scheduling and inventory counting.
[0028] In step S102, after the task of detecting grain surface anomalies is initiated, the obtained radar height measurement data is input into a preset focal length matching database. The optimal focusing parameters are automatically calculated based on the characteristics of the grain variety. These optimal focusing parameters include the target focal length, aperture, and fine-tuning offset. Specifically, based on the radar height measurement data and the camera's installation height, the actual imaging distance is calculated. Using the grain variety ID associated with the current silo number as an index, the basic focal length-object distance mapping curve, variety correction coefficient, and depth-of-field constraint threshold corresponding to that variety are retrieved from the focal length matching database. Different grain varieties correspond to different variety correction coefficients. The focal length-object distance mapping curve, variety correction coefficient, and depth-of-field constraint threshold are then fused to obtain the optimal focusing parameters, including the target focal length, aperture, and fine-tuning offset. The target focal length is then converted into a PWM control signal, which drives the lens motor in the video monitoring module to adjust the camera lens to the target focal length. This ensures the camera can clearly capture grain surface details, providing a high-quality image data source for subsequent image preprocessing and anomaly recognition. Otherwise, abnormal shooting phenomena may occur. Figure 2 As shown. Simultaneously, this focusing process forms a closed loop with the cleaning action of the ring-shaped wiper mechanism. Once cleaning is complete, the input of radar altimetry data and the focusing process are immediately initiated, preventing dust from re-adhering and affecting focusing accuracy, thus further ensuring image acquisition quality.
[0029] The video monitoring module described in this application is pre-cleaned. To avoid the influence of dust and other contaminants on the video monitoring module, a ring-shaped wiper mechanism is set in front of the camera lens to reciprocate and clean the camera lens surface, thereby ensuring the clarity of the captured image. In addition, a centrifugal dust removal mechanism continuously removes fine dust through airflow vortex. This dual mechanism ensures the long-term cleanliness of the lens in the high-dust environment of the silo, avoiding image quality degradation caused by dust accumulation in traditional equipment.
[0030] The focal length matching database pre-stores optimal focusing parameter samples for different grain varieties (such as wheat, corn, rice, etc.) at different grain surface heights, covering key data such as focal length range, lens motor speed, and focusing accuracy threshold. It also presets corresponding parameter matching rules for the differences in particle size and surface texture of different grain varieties.
[0031] The depth-of-field constraint threshold is set according to the type of grain and the sharpness of the image. For large grains such as corn and soybeans, the aperture needs to be reduced (the F-number increased) to increase the depth of field and ensure that the front and rear edges of the grains are sharp. At the same time, the resolution requirement should be slightly reduced to avoid over-sharpening and artifacts. Small grains such as millet and rapeseed require higher resolution to identify tiny mold spots, so the aperture should be opened as wide as possible (the F-number decreased) and the focus should be precisely on the surface of the grain. For highly reflective grains such as soybeans, the exposure parameters need to be adjusted in conjunction with the focus to avoid overexposure of reflective areas, which would lead to focus failure. For low-reflective grains such as rice, auxiliary lighting may need to be added or the exposure time extended. The focus algorithm needs to take into account the signal-to-noise ratio.
[0032] In step S103, after determining the target focal length value, this application keeps the recognition system fixed in the vertical direction (Z-axis). Based on the collected silo radius and the current imaging distance of the adjusted video monitoring module, a panoramic shooting coordinate sequence on the XY plane is generated using a path planning algorithm. The pan-tilt mechanism in the adjusted video monitoring module is controlled to successively deflect the camera angle according to each point in the calculated panoramic shooting coordinate sequence, thereby capturing high-definition images of each sub-region of the pre-divided silo according to the panoramic shooting coordinate sequence. Figure 3-4 As shown, this ensures that each sub-region can complete high-definition image acquisition at the optimal focal length, eliminating monitoring blind spots caused by edge distortion. The gimbal mechanism is equipped with closed-loop control and self-testing algorithms, combined with feedback calibration of radar altitude data, effectively eliminating the risk of monitoring point offset caused by mechanical wear, ensuring the long-term stability of the equipment in high-dust, high-frequency operating environments, reducing equipment failure rate and manual maintenance costs, and forming a sustainable identification system. During the shooting process, the identification system calls the grain type feature database corresponding to the current warehouse number in real time, loading the standard texture, color threshold, and morphological feature parameters of that type of grain.
[0033] In a specific implementation of step S103, one embodiment is as follows: The geometric dimensions of the silo and the current imaging distance of the video monitoring module are processed using a path planning algorithm to generate a panoramic shooting coordinate sequence, including: S1031. Calculate the reference value of the step spacing for a single shot based on the geometric dimensions of the silo and the current imaging distance of the video monitoring module. S1032. Establish a Cartesian coordinate system within the silo, and calculate the coordinates of shooting points covering the entire silo area at intervals based on the step spacing reference value to generate a panoramic shooting coordinate sequence with a dynamic optimization mechanism.
[0034] In steps S1031-S1032, this application calls the pre-set silo geometric parameter database in the identification system to extract the precise radius dimension of the currently monitored silo. This dimension is the distance from the center of the silo's cross-section to the silo wall, which is the core benchmark for determining the shooting coverage area. Simultaneously, the current imaging distance of the adjusted camera is collected in real time. This distance is derived from the determined optimal focus parameters and is positively correlated with the focal length, directly determining the coverage area, clarity, and edge distortion of a single captured image. Subsequently, the path planning algorithm establishes an XY plane rectangular coordinate system within the silo with the center of the silo's cross-section as the origin. The silo radius dimension is used as the maximum boundary of the shooting coverage. Combined with the current imaging distance, the algorithm calculates the sector angle that a single image can effectively cover and the interval distance between adjacent shooting points, ensuring that there is a reasonable overlap area between adjacent images (avoiding monitoring blind spots) while avoiding excessively large overlap areas that lead to data redundancy. The algorithm iteratively calculates and determines the X-axis and Y-axis coordinates of each shooting point in sequence, forming a complete panoramic shooting coordinate sequence. Each point in the panoramic shooting coordinate sequence corresponds to an optimal shooting angle, ensuring that when the camera shoots according to this sequence, it can completely cover the entire grain surface of the silo with the adjusted optimal focal length, without blind spots or distortion, providing complete and high-quality image data support for subsequent image stitching and anomaly recognition.
[0035] In step S104, since invalid frames exist in the acquired high-definition images, this application performs quality screening on the acquired high-definition images. Specifically, a blur detection algorithm is used to remove invalid frames that are out of focus or obscured by dust, thereby preserving the existence of valid target high-definition images. Figure 5 As shown, the target high-definition image also needs to enter the preprocessing process, including barrel wall reflection suppression, shadow compensation and grayscale equalization processing. The target high-definition image selected by quality screening is then subjected to anomaly identification through a multi-channel vision algorithm to identify whether the grain surface inside the silo is abnormal or normal. After identifying the grain surface anomaly, an early warning information containing the coordinates of the abnormal location, type code and suggested handling measures is pushed to a designated group through the communication interface bound to the monitoring platform.
[0036] This application automates the entire monitoring process. The identification system automatically starts radar scanning, mirror cleaning, focus adjustment, panoramic shooting, data processing, and anomaly warning at preset times every day, without the need for manual intervention. At the same time, it is bound to the monitoring platform through the communication module and automatically pushes early warning information, realizing unattended intelligent monitoring, which significantly reduces the labor costs of warehouse management and improves management efficiency.
[0037] In a specific implementation of step S104, one embodiment is as follows: the anomaly identification of the target high-definition image selected through quality screening using a multi-channel vision algorithm includes: S1041. Pre-set corresponding channels based on the type of grain surface anomaly; different channels correspond to different visual sub-algorithms; S1042. Run the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel and generate anomaly identification results.
[0038] In steps S1041-S1042, this application pre-summarizes and statistically analyzes the types of anomalies appearing on the grain surface in the silo, including pests, mold, leakage, and clumping. Corresponding channels are then set up for each channel, with different visual sub-algorithms for each channel. Each channel operates independently without interference. At the same time, a decision tree model is used to establish the correlation weights of each channel to ensure the comprehensiveness and accuracy of anomaly identification. The visual sub-algorithms are run to identify anomalies in the target high-definition image in the corresponding channel and generate anomaly identification results. Through the multi-channel parallel operation architecture, the limitations of traditional single-dimensional detection are broken through. At the same time, automatic early warning is performed based on the anomaly identification results, realizing full automation of the early warning process. The response time is shortened to the real-time level, completely changing the lagging management mode that relies on manual inspection, reducing the false judgment and missed judgment rates, timely detecting risks such as grain deterioration and leakage, and reducing grain storage losses.
[0039] In a specific implementation of step S1042, one embodiment involves running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel, including: A1. Convert the color space of the target high-definition image and extract the chromaticity feature components of the converted target high-definition image; A2. Compare the color feature components with the preset standard color threshold for grains to obtain the mold candidate area to identify mold abnormalities.
[0040] In steps A1-A2, this application sets up an HSV color space transformation sub-algorithm corresponding to the mold detection channel to convert the color space of the target high-definition image to the HSV color space. Compared with the RGB color space, the HSV color space is closer to the human eye's perception of color and can effectively separate the three feature components of chroma, saturation, and brightness, avoiding interference from brightness changes on color recognition. It is more suitable for capturing subtle color changes on the grain surface in the early stages of mold. After the color space conversion is completed, the HSV color space transformation sub-algorithm corresponding to the mold detection channel is used to specifically extract the converted target color space. The high-resolution image contains the chromaticity (H) feature component, which directly reflects the color attribute of the grain surface and is the core feature for distinguishing normal grain color from moldy grain color. At the same time, the main control module calls the grain feature database corresponding to the current warehouse number in real time to extract the preset standard chromaticity threshold for the grain. This threshold is based on the statistical analysis of chromaticity data from a large number of normal and moldy grain samples, covering the chromaticity range under normal conditions and the critical values of chromaticity deviation in the early and middle stages of mold. In addition, corresponding threshold correction parameters are preset for different storage environments (such as temperature and humidity) to ensure the accuracy of the comparison. Subsequently, the extracted chromaticity feature components are compared pixel by pixel with the preset standard chromaticity threshold for grains. All pixel areas whose chromaticity feature components exceed the standard threshold range are screened out and marked as mold candidate areas. To avoid misjudgment, the mold candidate areas are also verified a second time to remove false candidate areas caused by dust residue or uneven lighting. Finally, based on the area, distribution density and chromaticity deviation of the candidate areas, the abnormal mold areas on the grain surface are accurately identified, achieving early and accurate identification of mold abnormalities.
[0041] In a specific implementation of step S1042, another embodiment includes: running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel, further comprising: B1. Extract the image texture features of the target high-definition image, and match the image texture features with the standard texture features pre-stored in the grain variety feature database to obtain the matching result; B2. Based on the matching results, the corresponding region is determined as a candidate region for clumping, so as to identify clumping anomalies.
[0042] In steps B1-B2, this application, for cases including agglomeration anomalies, runs a texture analysis sub-algorithm. First, for the pre-processed target high-resolution image, a texture feature extraction process is initiated. The texture analysis sub-algorithm extracts the core texture features of the image, specifically including four core parameters: texture roughness, texture contrast, texture direction, and texture uniformity. Normal grain surfaces exhibit a uniform and regular texture distribution, while agglomeration areas, due to grain particle aggregation and adhesion, show significant differences in increased roughness, enhanced contrast, disordered texture direction, and decreased uniformity. These parameters accurately characterize whether agglomeration exists on the grain surface. During the extraction process, the algorithm divides the target high-resolution image into blocks, extracting texture feature parameters block by block to ensure no local texture anomalies are missed. Simultaneously, it filters out image noise interference with texture extraction, improving the accuracy of the feature parameters. Subsequently, the processing module calls the grain type feature database corresponding to the current warehouse number in real time to extract the pre-stored standard texture feature parameters for that grain type. These standard parameters are based on the textures of a large number of normal grain samples. Data statistics reveal a standard range covering the texture roughness, contrast, direction, and uniformity of the grain under normal storage conditions. Corresponding texture feature correction coefficients are preset for different storage periods to ensure matching compatibility. Next, the extracted texture feature parameters of the target high-definition image are precisely matched parameter-by-parameter and region-by-region with the pre-stored standard texture feature parameters in the grain feature database. The similarity value is calculated to obtain the specific matching result. When the similarity value is higher than a preset threshold, the texture of the region is determined to conform to normal grain surface characteristics; when the similarity value is lower than the preset threshold, the texture of the region is determined to be abnormal, and the abnormal texture region is marked as a clumping candidate region. To further improve recognition accuracy and avoid misjudging impurities, dust accumulation, etc., as clumping, a secondary verification is performed on the clumping candidate region. Combining the shape, area, and surrounding texture distribution characteristics of the region, false candidate regions are eliminated. Finally, based on the number, distribution range, and degree of texture abnormality of the candidate regions, the clumping abnormal region is accurately determined, achieving efficient and accurate identification of clumping abnormalities.
[0043] The impurity contamination identification channel and the clumping identification channel share some texture analysis logic. By extracting the texture features of the grain surface image and comparing them with the standard texture, the channel focuses on identifying areas with abrupt texture changes and significant differences from the standard texture of the grain variety. Combined with color features to assist in differentiation, the channel accurately identifies impurities and contaminants.
[0044] In the specific implementation of step S1042, there is also an embodiment in which: running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel also includes: C1. Apply the dynamic frame difference method to the image sequence composed of high-definition images of the target to extract the contours of the moving target between adjacent frames; C2. By combining the preset pest morphology feature model for matching and filtering, dynamic trajectory areas that match the pest activity characteristics are identified in order to identify pest anomalies.
[0045] In steps C1-C2, the specific visual sub-algorithm used in this application for identifying insect pest anomalies is a dynamic frame difference sub-algorithm and a shape feature matching sub-algorithm. First, a continuous image sequence composed of high-definition target images in the order of capture time is acquired. This image sequence is generated point-by-point by the pan-tilt mechanism according to the panoramic shooting coordinate sequence, with a fixed time interval between frames to ensure complete capture of dynamic changes in the grain surface area. Then, the dynamic frame difference method is applied to this image sequence. The algorithm performs pixel-level difference calculations on adjacent frames one by one. First, adjacent frames are converted to grayscale images to reduce color interference. Then, the grayscale difference between corresponding pixels in the two frames is calculated. A reasonable grayscale difference threshold is set, and pixels with a grayscale difference greater than the threshold are marked as moving pixels, while pixels with a grayscale difference less than or equal to the threshold are determined as static background pixels. This process extracts the contours of moving targets between adjacent frames, initially screening out all areas with dynamic activity within the grain surface area and eliminating interference from static backgrounds. To avoid misidentifying non-pest dynamic targets such as floating dust and slight grain surface settling as pests, the extracted moving target contours need to be matched and filtered. The processing module calls a pre-set pest morphology feature model in real time. This model is generated through machine learning algorithms based on data on the size, contour features, and movement posture of various common storage pests (such as rice weevils, grain borers, and wheat moths). It includes contour templates, morphological parameters (such as body length, body width, and contour complexity), and movement characteristics (such as movement speed and trajectory patterns) for different pests. Furthermore, corresponding morphological matching correction parameters are preset for different grain types and particle sizes to ensure matching accuracy. Subsequently, each extracted moving target contour is aligned and matched one by one with the contour template and morphological parameters in the pest morphology feature model. The morphological similarity and movement feature fit between the two are calculated to obtain the matching result. The matching results are used for filtering: if the matching degree between the contour of the moving target and the pest morphology feature model is higher than the preset matching threshold, and its movement trajectory conforms to the activity pattern of pests (such as irregular crawling, short-distance rapid movement, etc.), then the dynamic area is determined to be a dynamic trajectory area that conforms to the pest activity characteristics and is marked as a pest candidate area; if the matching degree is lower than the preset threshold, or the movement trajectory does not conform to the pest activity characteristics (such as irregular floating of dust, uniform movement of overall grain surface settling, etc.), then it is determined to be a non-pest dynamic target and is removed. At the same time, combined with the density statistics sub-algorithm, the number of pest candidate areas is counted. When the density of dynamic targets that conform to the pest characteristics in the candidate area reaches the preset threshold, it can be finally determined as an abnormal pest area, realizing accurate identification of pest anomalies, effectively avoiding misjudgment and missed judgment, ensuring that signs of pest activity on the grain surface can be captured in a timely manner, and providing accurate data support for subsequent pest control.
[0046] In the specific implementation of step S1042, another embodiment exists: running the visual sub-algorithm to identify anomalies in the target high-definition image in the corresponding channel, further including: D1. Perform edge detection on the target high-definition image, and extract the edge lines of the grain surface area after detection; D2. Spatial registration is performed between the detected abnormal edges and the pre-built grain surface settling model to identify leakage candidate areas and thus identify leakage anomalies.
[0047] In steps D1-D2, the specific visual sub-algorithm used in this application to identify material leakage anomalies is an edge detection algorithm. First, an edge detection process is initiated for the target high-definition image. The Canny edge detection algorithm is used to extract edges from the target high-definition image. This algorithm has advantages such as strong noise resistance and accurate edge positioning, effectively filtering interference factors such as dust and uneven lighting inside the silo, and accurately capturing the boundary lines between the grain surface area and the background, and between the normal and abnormal areas of the grain surface. Specifically, the target high-definition image is first subjected to Gaussian filtering to smooth image noise and reduce false edge interference. Then, the gradient magnitude and gradient direction of each pixel in the image are calculated, and pixels with gradient magnitudes greater than a preset threshold are selected to initially determine candidate edge pixels. Finally, a non-maximum suppression algorithm is used to remove redundant pixels in the edge contour, retaining clear and continuous edge lines, ultimately extracting the complete edge lines of the grain surface area, including the boundary edge between the grain surface and the silo wall, the undulating edges of the grain surface itself, and any possible leakage anomaly edges. To accurately distinguish between normal grain surface settling and abnormal leakage, the detected abnormal edges need to be spatially registered with a pre-constructed grain surface settling model. First, a pre-constructed grain surface settling model is built. This model is based on the physical characteristics of grain storage in silos, combined with a large amount of normal grain surface settling data (such as settling patterns for different storage periods and grain types). It is generated using 3D modeling technology, encompassing key parameters such as edge morphology, settling gradient, and spatial distribution characteristics of normal grain surface settling. It clarifies the curvature, extension direction, and relative position of the normal settling edges to the silo wall, while also pre-setting characteristic thresholds for abnormal leakage edges (such as edge abrupt change angle, fracture degree, and extension length). Subsequently, the processing module calls the 3D point cloud data generated by the radar module, spatially correlates the extracted grain surface edge lines with the 3D point cloud data, obtains the corresponding 3D spatial coordinates of the edge lines, and then performs spatial registration calculations with the pre-constructed grain surface settling model. Through coordinate alignment and morphological comparison, the degree of fit between the detected abnormal edges and the normal settling edges in the grain surface settling model is calculated. Based on spatial registration results, leakage anomaly identification is performed: if the detected abnormal edge has a higher degree of fit with the grain surface settling model than the preset registration threshold, and the edge morphology conforms to the normal grain surface settling pattern, then the edge is determined to be a normal settling edge and is excluded; if the degree of fit is lower than the preset registration threshold, and the edge exhibits leakage characteristics such as abrupt changes, breaks, and irregular extensions (such as extending downwards towards the silo wall, with disordered edge contours), and the grayscale changes and texture differences of the edge area are combined, the area corresponding to the abnormal edge is marked as a leakage candidate area.To further improve recognition accuracy and avoid misjudging edge anomalies caused by localized depressions in the grain surface or accumulation of impurities as material leakage, a secondary verification of the leakage candidate area will be performed. By combining the height change characteristics of the area in the radar point cloud data (the leakage area will show a significant drop in height), false candidate areas will be eliminated. Finally, based on the area, extension range, and degree of edge anomaly of the leakage candidate area, the leakage anomaly area will be accurately identified, achieving timely and accurate identification of leakage anomalies and avoiding losses caused by continuous grain leakage.
[0048] Example 2 This application also provides a device for identifying abnormal grain levels inside a silo, such as... Figure 6 The diagram shows a block diagram of a grain surface anomaly detection device inside a silo. This device performs functions corresponding to the steps of the aforementioned method for detecting grain surface anomalies inside a silo on a terminal device. This device can be understood as a server component including a processor. The grain surface anomaly detection device described in this application is applied to an identification system, which includes a radar module, a video monitoring module, and a main control module. The device includes: The detection module 601 is used to control the radar module to rotate and emit a high-frequency beam to detect the grain surface inside the silo, obtain a three-dimensional point cloud map of the grain surface inside the silo, and calculate radar height measurement data based on the three-dimensional point cloud map. The matching module 602 is used to match the radar altimetry data, grain variety, and a preset focal length matching database through the main control module to find the optimal focusing parameters for the video monitoring module, and to control the video monitoring module to adjust according to the optimal focusing parameters; the video monitoring module has been pre-cleaned. The snapshot module 603 is used to process the geometric dimensions of the silo and the current imaging distance of the adjusted video monitoring module through a path planning algorithm, generate a panoramic shooting coordinate sequence, and control the adjusted video monitoring module to capture high-definition images of each sub-area of the silo according to the panoramic shooting coordinate sequence. The identification module 604 is used to perform quality screening on the high-definition image and to perform anomaly identification on the target high-definition image selected by the quality screening through a multi-channel vision algorithm, so as to identify the abnormal grain surface in the silo.
[0049] In one feasible implementation, the identification module includes: The corresponding channels are pre-set based on the type of grain surface anomaly; different channels correspond to different visual sub-algorithms; The visual sub-algorithm is run to identify anomalies in the corresponding channels of the target high-definition image and generate anomaly identification results.
[0050] In one feasible implementation, the identification module further includes: The color space of the target high-definition image is converted, and the chromaticity feature components of the converted target high-definition image are extracted. By comparing the colorimetric feature components with the preset standard colorimetric threshold for grains, mold candidate regions are obtained to identify mold abnormalities.
[0051] In one feasible implementation, the identification module also includes: Extract the image texture features of the target high-definition image, and match the image texture features with the standard texture features pre-stored in the grain variety feature database to obtain the matching result; Based on the matching results, the corresponding region is determined as a candidate region for clumping, so as to identify clumping anomalies.
[0052] In one feasible implementation, the identification module further includes: The dynamic frame difference method is applied to the image sequence composed of high-resolution images of the target to extract the contours of the moving target between adjacent frames; By combining a pre-set pest morphology feature model with matching and filtering, dynamic trajectory areas that match the pest activity characteristics are identified to detect pest anomalies.
[0053] In one feasible implementation, the identification module further includes: Edge detection is performed on the target high-definition image, and the edge lines of the grain surface area are extracted after detection; The detected abnormal edges are spatially registered with the pre-built grain surface settling model to identify leakage candidate areas and thus identify leakage anomalies.
[0054] In one feasible implementation, the image capture module includes: Based on the geometric dimensions of the silo and the current imaging distance of the video monitoring module, calculate the reference value of the step spacing for a single shot; A Cartesian coordinate system is established within the silo. The coordinates of shooting points covering the entire silo area are calculated point by point at intervals based on the step spacing reference value to generate a panoramic shooting coordinate sequence with a dynamic optimization mechanism.
[0055] Example 3 This application also provides an electronic device, such as Figure 7 As shown, it includes: a processor 701, a memory 702, and a bus 703. The memory 702 stores machine-readable instructions that can be executed by the processor 701. When the electronic device is running, the processor 701 and the memory 702 communicate through the bus 703. When the machine-readable instructions are executed by the processor 701, the steps of any one of the methods for identifying abnormal grain surface in a silo are performed.
[0056] Example 4 This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of any one of the methods for identifying abnormal grain levels in a silo.
[0057] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems and devices described above can be referred to the corresponding processes in the method embodiments, and will not be repeated here. In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling or direct coupling or communication connection can be through some communication interfaces; the indirect coupling or communication connection of devices or modules can be electrical, mechanical, or other forms.
[0058] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.
[0059] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0060] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a platform server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.
[0061] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for identifying abnormal grain surface conditions inside a silo, characterized in that, The method is applied to an identification system, which includes a radar module, a video surveillance module, and a main control module, and includes: The radar module is controlled to rotate and emit a high-frequency beam to detect the grain surface inside the silo, obtain a three-dimensional point cloud map of the grain surface inside the silo, and calculate radar height measurement data based on the three-dimensional point cloud map. The main control module matches the radar altimetry data, grain variety, and a preset focal length matching database to determine the optimal focusing parameters for the video monitoring module, and then controls the video monitoring module to adjust according to these optimal focusing parameters; the video monitoring module has been pre-cleaned. The path planning algorithm processes the geometric dimensions of the silo and the current imaging distance of the adjusted video monitoring module to generate a panoramic shooting coordinate sequence. The adjusted video monitoring module is then controlled to capture high-definition images of each sub-area of the silo according to the panoramic shooting coordinate sequence. The high-definition images are quality-screened, and anomaly recognition is performed on the high-definition images selected by the quality screening using a multi-channel vision algorithm to identify abnormalities in the grain surface inside the silo.
2. The method according to claim 1, characterized in that, The anomaly identification of the target high-definition image selected through quality screening using a multi-channel visual algorithm includes: The corresponding channels are pre-set based on the type of grain surface anomaly; different channels correspond to different visual sub-algorithms; The visual sub-algorithm is run to identify anomalies in the corresponding channels of the target high-definition image and generate anomaly identification results.
3. The method according to claim 2, characterized in that, Running the visual sub-algorithm to identify anomalies in the target high-resolution image in the corresponding channel includes: The color space of the target high-definition image is converted, and the chromaticity feature components of the converted target high-definition image are extracted. By comparing the color feature components with the preset standard color threshold for grains, mold candidate regions are obtained to identify mold abnormalities.
4. The method according to claim 2, characterized in that, Running the visual sub-algorithm to identify anomalies in the corresponding channels of the target high-resolution image further includes: Extract the image texture features of the target high-definition image, and match the image texture features with the standard texture features pre-stored in the grain variety feature database to obtain the matching result; Based on the matching results, the corresponding region is determined as a candidate region for clumping, so as to identify clumping anomalies.
5. The method according to claim 2, characterized in that, Running the visual sub-algorithm to identify anomalies in the target high-resolution image in the corresponding channel also includes: The dynamic frame difference method is applied to the image sequence composed of high-resolution images of the target to extract the contours of the moving target between adjacent frames; By combining a pre-set pest morphology feature model with matching and filtering, dynamic trajectory areas that match the pest activity characteristics are identified to detect pest anomalies.
6. The method according to claim 2, characterized in that, Running the visual sub-algorithm to identify anomalies in the target high-resolution image in the corresponding channel further includes: Edge detection is performed on the target high-definition image, and the edge lines of the grain surface area are extracted after detection; The detected abnormal edges are spatially registered with the pre-built grain surface settling model to identify leakage candidate areas and thus identify leakage anomalies.
7. The method according to claim 1, characterized in that, The path planning algorithm processes the silo's geometric dimensions and the current imaging distance of the video surveillance module to generate a panoramic shooting coordinate sequence, including: Based on the geometric dimensions of the silo and the current imaging distance of the video monitoring module, calculate the reference value of the step spacing for a single shot; A Cartesian coordinate system is established within the silo. The coordinates of shooting points covering the entire silo area are calculated point by point at intervals based on the step spacing reference value to generate a panoramic shooting coordinate sequence with a dynamic optimization mechanism.
8. A device for identifying abnormal grain levels in a silo, characterized in that, The device is applied to an identification system, which includes a radar module, a video surveillance module, and a main control module. The detection module is used to control the radar module to rotate and emit high-frequency beams to detect the grain surface inside the silo, obtain a three-dimensional point cloud map of the grain surface inside the silo, and calculate radar height measurement data based on the three-dimensional point cloud map. The matching module is used to match the radar altimetry data, grain variety, and a preset focal length matching database through the main control module to find the optimal focusing parameters for the video monitoring module, and control the video monitoring module to adjust according to the optimal focusing parameters; the video monitoring module has been pre-cleaned. The snapshot module is used to process the geometric dimensions of the silo and the current imaging distance of the adjusted video monitoring module through a path planning algorithm, generate a panoramic shooting coordinate sequence, and control the adjusted video monitoring module to capture high-definition images of each sub-area of the silo according to the panoramic shooting coordinate sequence. The identification module is used to perform quality screening on the high-definition images and to perform anomaly identification on the target high-definition images selected by quality screening through a multi-channel vision algorithm, so as to identify abnormalities in the grain surface inside the silo.
9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of a method for identifying abnormal grain levels in a silo as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of a method for identifying abnormal grain levels in a silo as described in any one of claims 1 to 7.