Method and device for detecting cooked grain clumps, electronic equipment and storage medium

Abstract

The application relates to the technical field of solid-state fermentation wine brewing, and provides a cooked grain clumping detection method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring geometric shape features and impact response features of cooked grains to be detected; confidence degree interaction verification is performed to obtain geometric shape features carrying a first confidence degree and impact response features carrying a second confidence degree; a first clumping size distribution of the cooked grains to be detected is determined according to the geometric shape features, and a second clumping size distribution of the cooked grains to be detected is determined according to the impact response features; a third clumping size distribution is determined according to the geometric shape features carrying the first confidence degree and the impact response features carrying the second confidence degree; the first clumping size distribution, the second clumping size distribution and the third clumping size distribution are fused, and a clumping detection result of the cooked grains to be detected is determined according to a fusion result. The application can improve the reliability and precision of cooked grain clumping detection under complex industrial working conditions.

Description

Technical Field

[0001] This application belongs to the field of solid-state fermentation brewing technology, and in particular relates to a method, device, electronic equipment and storage medium for detecting grain clumping. Background Technology

[0002] In solid-state fermentation brewing, the grain spreading process is a crucial physical pretreatment step. Freshly steamed grains have extremely high temperature and humidity, requiring rapid cooling and moisture dissipation via a grain spreading machine on a conveyor belt to meet the requirements of subsequent processes such as adding koji, saccharification, and fermentation in the pit. However, due to the high starch content and surface stickiness of the grains, they easily stick together during cooling and transport, forming clumps of varying sizes—a phenomenon known as grain agglomeration. If agglomeration is severe and not detected and addressed promptly, it will reduce the permeability of the grain layer, severely affecting the uniform distribution of saccharifying enzymes and the normal metabolism of fermenting microorganisms, ultimately leading to a decrease in alcohol yield and the quality of the raw liquor. Related technologies often employ visual inspection to identify grain agglomeration. While visual inspection can locate the spatial position of agglomerates and measure their geometric dimensions, image quality deteriorates in high-dust and high-humidity environments. Therefore, agglomeration detection results obtained through visual inspection are not accurate. Summary of the Invention

[0003] In view of the shortcomings of the prior art, this application provides a method, apparatus, electronic device and storage medium for detecting clumping of cooked grains, which at least solves the problem of inaccurate clumping detection results obtained by visual inspection in related technologies.

[0004] Firstly, this application provides a method for detecting clumping of cooked grains, including: Acquire the initial geometric morphological features and initial impact response features of the cooked grain to be tested; The initial geometric morphological features and initial impact response features are cross-validated with confidence levels to obtain geometric morphological features carrying the first confidence level and impact response features carrying the second confidence level. The first agglomeration size distribution of the cooked grain to be tested is determined based on the initial geometric morphology characteristics, and the second agglomeration size distribution of the cooked grain to be tested is determined based on the initial impact response characteristics. The third cluster size distribution is determined based on the geometric morphological characteristics carrying the first confidence level and the impact response characteristics carrying the second confidence level. The size distributions of the first, second, and third agglomerates are fused, and the agglomeration detection results of the cooked grain to be tested are determined based on the fusion results.

[0005] In one embodiment of this application, confidence cross-validation is performed on the initial geometric morphological features and the initial impact response features, including: If a first preset event exists in the initial geometric morphology features, then the target impact response features are determined from the initial impact response features according to the pre-determined feature correspondence to perform confidence-interactive verification of the first preset event, and the first confidence level of the initial geometric morphology features is determined based on the verification result; if a second preset event exists in the initial impact response features, then the target geometric morphology features are determined from the initial geometric morphology features according to the feature correspondence to perform confidence-interactive verification of the second preset event, and the second confidence level of the initial impact response features is determined based on the verification result.

[0006] In one embodiment of this application, before performing confidence cross-validation on the initial geometric features and the initial impact response features, the method further includes: Obtain the sample geometric morphology features and sample impact response features of the cooked grain sample to be tested within a preset time period; determine the time boundary based on the falling height of the cooked grain sample to be tested, and determine the sample event pair between the first preset event in the sample geometric morphology features and the second preset event in the sample impact response features within the time boundary; determine the feature correspondence based on the sample event pair.

[0007] In one embodiment of this application, the geometric features include the area and dispersion of the connected domains, and the impact response features include the impact force amplitude and the frequency domain energy. The third cluster size distribution is determined based on the geometric morphological characteristics carrying the first confidence level and the impact response characteristics carrying the second confidence level, including: The comprehensive clumping index is determined based on the area and dispersion of the connected region carrying the first confidence level, and the impact force amplitude and frequency domain energy carrying the second confidence level. Based on the comprehensive agglomeration index and the preset agglomeration grading threshold, the initial agglomeration size matching probability of the cooked grain to be tested is determined. The initial matching probability is normalized to obtain the third cluster size distribution.

[0008] In one embodiment of this application, the fusion of the first cluster size distribution, the second cluster size distribution, and the third cluster size distribution includes: Based on the first confidence level, determine the first sub-distribution confidence level of the first cluster size distribution; based on the second confidence level, determine the second sub-distribution confidence level of the second cluster size distribution. Based on the first and second confidence levels, determine the reliability index of the third cluster size distribution characteristic; Based on the reliability index and the characteristic consistency index, the third distribution reliability of the third cluster size distribution characteristics is determined, where the characteristic consistency index characterizes the degree of correspondence between geometric morphological characteristics and impact response characteristics. The reliability of the first, second, and third sub-distributions is normalized, and the first, second, and third cluster size distributions are weighted and fused based on the normalization results.

[0009] In one embodiment of this application, before normalizing the confidence levels of the first, second, and third sub-distributions, the method further includes: If any of the confidence levels of the first sub-distribution confidence, the second sub-distribution confidence, and the third sub-distribution confidence is lower than the first preset threshold, then the confidence level of the sub-distribution that is lower than the first preset threshold is set to zero. If any of the confidence levels of the first sub-distribution confidence, the second sub-distribution confidence, and the third sub-distribution confidence is greater than or equal to the first preset threshold and less than the second preset threshold, then the sub-distribution confidence that is greater than or equal to the first preset threshold and less than the second preset threshold is updated by a preset attenuation coefficient.

[0010] In one embodiment of this application, determining the clumping detection result of the cooked grain to be tested based on the fusion result includes: Obtain the fusion results within the target time window; Based on the current working condition characteristics of the cooked grain to be tested, the size distribution of the fused agglomerates is standardized to obtain a standardized agglomerate size distribution. The current working condition characteristics include the grain variety, the thickness of the grain layer on the plate chain, and the plate chain speed. Based on the standardized clumping size distribution and the preset clumping judgment rules, the clumping test results of the cooked grain to be tested are determined.

[0011] Secondly, this application also provides a device for detecting grain clumping, comprising: The feature acquisition module is configured to acquire the initial geometric morphological features and initial impact response features of the cooked grain to be detected; The feature verification module is configured to perform confidence-based interactive verification on the initial geometric morphology features and the initial impact response features, respectively, to obtain geometric morphology features carrying a first confidence level and impact response features carrying a second confidence level. The first determining module is configured to determine the first agglomeration size distribution of the cooked grain to be tested based on the initial geometric morphological features, and to determine the second agglomeration size distribution of the cooked grain to be tested based on the initial impact response features. The second determining module is configured to determine the third agglomeration size distribution based on the geometric morphological features carrying a first confidence level and the impact response features carrying a second confidence level. The result detection module is configured to fuse the first cluster size distribution, the second cluster size distribution, and the third cluster size distribution, and determine the cluster detection result of the cooked grain to be tested based on the fusion result.

[0012] Thirdly, this application also provides an electronic device, which includes: one or more processors; and a storage device for storing one or more programs, which, when executed by one or more processors, cause the electronic device to perform the steps of the above-described method for detecting grain clumping.

[0013] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a computer's processor, enables the computer to implement the steps of the above-described method for detecting grain clumping.

[0014] The beneficial effects of this technical solution are as follows: First, by acquiring the initial geometric morphological features and initial impact response features of the cooked grain to be tested, and establishing a confidence-based interactive verification mechanism for the initial geometric morphological features and initial impact response features, the confidence of the initial geometric morphological features and initial impact response features can be improved, reducing the impact of low-confidence features on agglomeration detection. Second, by reconstructing the third agglomeration size distribution using geometric morphological features carrying the first confidence level and impact response features carrying the second confidence level, the transmission path of single-mode deviation at the distribution level can be cut off, providing an independent and highly reliable third data source for subsequent detection. Finally, by fusing the first, second, and third agglomeration size distributions, and determining the agglomeration detection result of the cooked grain to be tested based on the fusion result, the high reliability of the geometric morphological features, impact response features, and reconstructed third agglomeration size distribution can be combined to form a more comprehensive and objective fusion result through multi-source fusion, reducing the one-sided influence of a single data source and improving the reliability and accuracy of cooked grain agglomeration detection under complex industrial conditions.

[0015] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings: Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of the present application of a method for detecting clumping of cooked grains; Figure 2 This is a schematic diagram of the structure of a cooked grain clumping detection device, as shown in an exemplary embodiment of this application; Figure 3 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation

[0017] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0018] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. The drawings only show the components related to this application and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the shape, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0019] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the present application. However, it will be apparent to those skilled in the art that embodiments of the present application may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the present application.

[0020] Please see Figure 1 , Figure 1 This is a flowchart illustrating an exemplary embodiment of the method for detecting clumping of cooked grains, as presented in this application. Figure 1 As shown, in an exemplary embodiment, the method for detecting clumping of cooked grain includes steps S110 to S140, and each step is described in detail below.

[0021] S110, acquire the initial geometric morphological features and initial impact response features of the cooked grain to be tested; Geometric morphological features refer to the feature parameters extracted by the visual detection unit after processing the image of the cooked grain to be detected, which are used to characterize the morphological attributes of cooked grain particles or clumps in the image plane.

[0022] Impact response characteristics refer to the characteristic parameters extracted by the impact detection unit after processing the mechanical signals generated when cooked grain falls and hits the load-bearing plate, which are used to characterize the physical impact properties of the material.

[0023] It is understandable that in the actual grain spreading process, visual inspection units (such as high-definition cameras and industrial cameras) can be deployed in advance on the grain spreading and material dropping side to collect images of the cooked grain to be inspected in real time, and the initial geometric features of the cooked grain to be inspected can be extracted through image processing technology. Among them, geometric features include, but are not limited to, the area of ​​connected regions, aspect ratio, dispersion, etc. Simultaneously, an impact detection unit (e.g., an array of accelerometers arranged on the load-bearing plate) can be deployed on the load-bearing plate to collect the initial impact response characteristics generated by the falling cooked grain hitting the load-bearing plate in real time. The impact response characteristics include, but are not limited to, the impact force amplitude, impact duration, and frequency domain energy distribution.

[0024] S120, perform confidence cross-validation on the initial geometric morphology features and the initial impact response features respectively to obtain the geometric morphology features carrying the first confidence level and the impact response features carrying the second confidence level. Confidence cross-validation refers to the ability to verify the initial shock response features using the initial geometric morphology features, or the ability to verify the initial geometric morphology features using the initial shock response features. If a single-modal feature (i.e., any one of the initial geometric morphology feature and the initial impact response feature) is misjudged, the detection result will be incorrect. This step uses confidence level cross-validation to allow the initial geometric morphology feature and the initial impact response feature to query and verify each other, dynamically correct the confidence level of each feature, and obtain the geometric morphology feature carrying the first confidence level and the impact response feature carrying the second confidence level.

[0025] S130, determine the first agglomeration size distribution of the cooked grain to be tested based on the initial geometric morphology characteristics, and determine the second agglomeration size distribution of the cooked grain to be tested based on the initial impact response characteristics; Within a single detection cycle, the initial geometric morphological features acquired by the visual detection unit can be used to independently estimate and output the first agglomeration size distribution, while the initial impact response features acquired by the impact detection unit can be used to independently estimate and output the second agglomeration size distribution. The first and second agglomeration size distributions represent the preliminary estimation results of the cooked grain agglomeration size state by the visual detection unit and the impact detection unit, respectively.

[0026] In some examples, the size distribution of the first cluster and the size distribution of the second cluster can both be represented as the percentage composition of the first cluster, the second cluster, and the third cluster (e.g., the first cluster accounts for 20%, the second cluster accounts for 30%, and the third cluster accounts for 50%), where the first cluster can be characterized as a large cluster, the second cluster can be characterized as a medium cluster, and the third cluster can be characterized as a small cluster.

[0027] S140, Based on the geometric morphological characteristics carrying the first confidence level and the impact response characteristics carrying the second confidence level, determine the third cluster size distribution; It is understandable that the third cluster size distribution does not directly depend on the original outputs of the first and second cluster size distributions. Instead, it is reconstructed based on the geometric morphological features carrying the first confidence level and the impact response features carrying the second confidence level after cross-validation of confidence levels. Since the features involved in the reconstruction are all verified features carrying confidence level labels, even if there is a deviation in the cluster size distribution of a certain mode, the deviation will not be transmitted to the third cluster size distribution. This provides an independent and highly reliable third cluster size distribution for subsequent decision-making.

[0028] S150, the first agglomeration size distribution, the second agglomeration size distribution and the third agglomeration size distribution are fused, and the agglomeration detection result of the cooked grain to be tested is determined based on the fusion result.

[0029] According to the technical solution provided in this application, firstly, by acquiring the initial geometric morphological features and initial impact response features of the cooked grain to be tested, and establishing a confidence-based interactive verification mechanism for the initial geometric morphological features and initial impact response features, the confidence of the geometric morphological features and impact response features can be improved, reducing the impact of low-confidence features on agglomeration detection. Secondly, by reconstructing the third agglomeration size distribution using geometric morphological features carrying the first confidence level and impact response features carrying the second confidence level, the transmission path of single-mode deviation at the distribution level can be cut off, providing an independent and highly reliable third data source for subsequent detection. Finally, the first agglomeration size distribution, the second agglomeration size distribution, and the third agglomeration size distribution are fused, and the agglomeration detection result of the cooked grain to be tested is determined based on the fusion result. By combining the high reliability of geometric morphological features, impact response features, and the reconstructed third agglomeration size distribution, a more comprehensive and objective fusion result can be formed through multi-source fusion, reducing the one-sided influence of a single data source and improving the reliability and accuracy of cooked grain agglomeration detection under complex industrial conditions.

[0030] In some embodiments, confidence cross-validation is performed on the initial geometric features and the initial impact response features, respectively, including: If a first preset event exists in the initial geometric morphology features, the target impact response features are determined from the initial impact response features according to the pre-determined feature correspondence relationship to perform confidence interactive verification of the first preset event, and the first confidence level of the initial geometric morphology features is determined based on the verification result. If a second preset event exists in the initial impact response features, the target geometric morphology features are determined from the initial geometric morphology features according to the feature correspondence to perform confidence interaction verification on the second preset event, and the second confidence level of the initial impact response features is determined based on the verification results.

[0031] Among them, the first preset event can characterize the event where the area of ​​the connected domain of the cooked grain to be detected is greater than a preset area threshold, and the second preset event can characterize the event where the impact amplitude of the cooked grain to be detected is greater than a preset amplitude threshold.

[0032] The predetermined feature correspondence refers to the matching and mapping relationship between the initial geometric features acquired by the visual detection unit and the initial impact response features acquired by the impact detection unit in the time and spatial dimensions.

[0033] It is understandable that in actual production, high humidity causes moisture to adhere to the surface of cooked grain, which can easily form large areas of false connected regions during image processing. At the same time, the loose accumulation of cooked grain or the accidental vibration of mechanical equipment may also generate a strong impact force response on the sensor array of the impact detection unit.

[0034] Therefore, when the initial geometric features acquired by the visual detection unit contain a first preset event (i.e., the area of ​​the connected domain of the cooked grain to be detected is greater than the preset area threshold), the target impact response features can be determined from the initial impact response features according to the pre-determined feature correspondence, and the first preset event can be verified by confidence interaction using the target impact response features.

[0035] For example, the impact force amplitude in the initial target impact response feature can be used for confidence interaction verification. If the impact force amplitude is lower than the first preset impact threshold, it indicates that the first preset event lacks a corresponding physical impact response. It can be determined that the suspected clumping area is an adhesion illusion under high humidity, the confidence interaction verification fails, the geometric feature is a visual pseudo-feature, and the first confidence level is assigned to the geometric feature.

[0036] If the impact force amplitude is higher than or equal to the first preset impact threshold, it indicates that the first preset event has real physical impact support, the confidence level interactive verification is passed, the initial geometric feature is a true visual feature, and the initial geometric feature is assigned a first confidence level.

[0037] It should be noted that feature confidence levels can be pre-classified into three levels: high, medium, and low. Features that have not undergone cross-validation of confidence can be marked as "medium" level; features that pass cross-validation of confidence can be upgraded one level to "high" level; and features that fail cross-validation of confidence can be downgraded one level to "low" level. To facilitate subsequent continuous numerical calculations, a confidence mapping table can be pre-defined, mapping the three levels of high, medium, and low to the values ​​1.0, 0.6, and 0.2, respectively.

[0038] In addition, when the initial impact response features acquired by the impact detection unit contain a second preset event (i.e., the impact force amplitude of the cooked grain to be detected is greater than the preset amplitude threshold), the target geometric features can be determined from the initial geometric features according to the feature correspondence, and the second preset event can be verified by confidence interaction using the target geometric features.

[0039] For example, the confidence level can be interactively verified using the area of ​​connected components in the initial target geometric features. If the area of ​​connected components is less than the first preset area threshold, it indicates that the second preset event lacks corresponding visual morphological support, and the abnormal impact force amplitude can be determined to be an illusion caused by accidental vibration or loose accumulation. The confidence level interactive verification fails, and the initial impact response feature is an impact pseudo-feature. At the same time, a second confidence level is assigned to the initial impact response feature. If the area of ​​connected components is greater than or equal to the first preset area threshold, it indicates that the second preset event has real morphological support, and the confidence level interactive verification passes. The initial impact response feature is an impact true feature, and a second confidence level is assigned to the initial impact response feature. The specific implementation method can refer to the implementation method of assigning the first confidence level to the geometric features, which will not be elaborated here.

[0040] According to the technical solution provided in this embodiment, by performing bidirectional confidence-based interactive verification of the initial geometric morphological features obtained by the visual detection unit and the initial impact response features obtained by the impact detection unit, visual false features caused by high humidity environment, as well as impact false features caused by equipment vibration and loose accumulation, can be effectively eliminated, thereby improving the accuracy of the detection results of cooked grain clumping, reducing the probability of misjudgment caused by a single detection method, and ensuring the reliability of the detection results.

[0041] In some embodiments, before performing confidence cross-validation on the initial geometric features and the initial impact response features respectively, the method further includes: Obtain the geometric morphological features and impact response features of the cooked grain samples to be tested within a preset time period; The time boundary is determined based on the falling height of the cooked grain sample to be tested. Within the time boundary, sample event pairs are determined between the first preset event in the geometric morphology characteristics of the sample and the second preset event in the impact response characteristics of the sample. Determine the feature correspondence based on the sample event pairs.

[0042] It is understandable that visual inspection units are typically deployed in the space above the material drop side, while impact inspection units are located on the load-bearing plate. Therefore, there is a time delay in the material's fall and transmission between the time the cooked grain is captured by the visual inspection unit and the time it is impacted by the load-bearing plate.

[0043] Therefore, during system initialization or process parameter changes, the geometric morphology and impact response characteristics of the sample material to be tested can be collected within a preset time period (e.g., M minutes, where the value of M can be set according to material flow rate and statistical stability requirements). The time and spatial location of each first preset event, as well as the time and sensor location of each second preset event (e.g., the generation of abnormal impact force), are recorded. Then, based on the physical fall height of the sample material to be tested, combined with gravitational acceleration, a reasonable time range required for the material to fall from the imaging position to the load-bearing plate is estimated, and this reasonable time range is determined as the time boundary. Within this time boundary, for each first preset event, a matching second preset event is searched within the subsequent time range to determine the sample event pair between the first preset event in the sample geometric morphology characteristics and the second preset event in the sample impact response characteristics. The feature correspondence is then determined based on the sample event pair.

[0044] Furthermore, to ensure the validity of event pairs, the equivalent diameter deviation between sample event pairs can be calculated, and valid sample event pairs with deviations less than a preset screening threshold can be selected, thereby eliminating accidental random coincidences. Then, based on the selected valid sample event pairs, feature correspondences are determined; for example, the average time difference of all valid sample event pairs can also be calculated and used as a fixed time correspondence value; simultaneously, based on the spatial distribution pattern of valid sample event pairs, a spatial correspondence table between visual positions and impact sensor numbers is established. Thus, the combination of time correspondence values ​​and the spatial correspondence table constitutes the feature correspondence.

[0045] In addition, in some embodiments, after determining the feature correspondence, a registration quality assessment can be performed every preset period (e.g., every P minutes, where the value of P can be set according to the stability requirements of the operating conditions): the pass rate of feature interaction verification in the most recent period is calculated, and the registration quality score is reduced when the pass rate is lower than a preset pass threshold; the degree of match between the size distribution of the first cluster and the size distribution of the second cluster in the most recent period is calculated, and the registration quality score is reduced when the degree of match is lower than a preset match threshold; the comprehensive score forms a registration quality index. When the registration quality index is lower than the re-registration threshold, a re-registration process is triggered, the existing correspondence is cleared, and the registration process during system initialization or process parameter changes is re-executed; when the registration quality index is in the fine-tuning range, a gradual fine-tuning can be performed to make a small correction to the fixed time correspondence value, and the correction amount can be determined based on the average time deviation observed in the most recent period. In this way, through closed-loop quality assessment and automatic correction mechanism, the system can dynamically self-correct the drift of the impact delay of cooked grain falling, further ensuring the accuracy of the feature correspondence in long-term operation.

[0046] According to the technical solution provided in this embodiment, the time boundary is determined by the falling height of the cooked grain sample to be detected, and a spatiotemporal matching relationship is established for cross-modal queries. This enables each confidence interaction verification to locate the corresponding signal of the same sample to be detected in another modality, reducing the probability of query failure due to spatiotemporal matching misalignment and ensuring the effectiveness of cross-modal feature verification.

[0047] In some embodiments, geometric features include the area and dispersion of connected domains, and impact response features include impact force amplitude and frequency domain energy. The third cluster size distribution is determined based on the geometric morphological characteristics carrying the first confidence level and the impact response characteristics carrying the second confidence level, including: The comprehensive clumping index is determined based on the area and dispersion of the connected region carrying the first confidence level, and the impact force amplitude and frequency domain energy carrying the second confidence level. Based on the comprehensive clumping index and the preset clumping grading threshold, the initial clumping size matching probability of the cooked grain to be tested is determined. The initial matching probability is normalized to obtain the third cluster size distribution.

[0048] The comprehensive clumping index is a scalar value that characterizes the current clumping state of the cooked grain to be tested, obtained by weighted integration of features carrying confidence.

[0049] It is understandable that the area of ​​the connected domain can characterize the size of the agglomeration in two-dimensional space; the dispersion can characterize the degree of aggregation of the agglomeration in terms of morphological compactness; the impact force amplitude can characterize the mass and density of the agglomeration in terms of physical impact force; and the frequency domain energy can characterize the structural characteristics of the agglomeration in terms of internal density. Therefore, these four types of features can characterize the essential properties of cooked grain agglomeration from four different physical dimensions.

[0050] By using preset feature fusion mapping rules, the area and dispersion of the connected domain carrying the first confidence level, as well as the impact force amplitude and frequency domain energy carrying the second confidence level, can be weighted and integrated to obtain the comprehensive clumping index.

[0051] For example, a comprehensive clumping index can be generated by fusing the connected region area, dispersion and corresponding first confidence level, impact force amplitude, frequency domain energy and corresponding second confidence level through a pre-trained feature fusion model.

[0052] Introducing first and second confidence levels into the weighted integration process allows features that have undergone two-way confidence verification to dominate the fusion process, while features that have not undergone two-way confidence verification are suppressed.

[0053] Then, based on the comprehensive clustering index and preset clustering classification thresholds (e.g., pre-configured comprehensive index classification thresholds for large, medium, and small clusters), the membership degree matching algorithm is used to perform interval adaptation calculations on the comprehensive clustering index, obtaining the initial cluster size matching probabilities for the three types of clusters (large, medium, and small) corresponding to the current clustering state. The membership degree matching algorithm refers to a matching calculation method that uses a membership function to calculate the degree of belonging of the comprehensive clustering index to each size level interval. This method can provide a gradual, rather than abrupt, belonging probability for index values ​​near the classification threshold boundaries, thereby improving the smoothness of the classification determination.

[0054] The initial matching probability is normalized to ensure that the sum of the probability proportions of large, medium and small clusters is 1, thus obtaining the final third cluster size distribution.

[0055] According to the technical solution provided in this embodiment, the reconstructed third cluster size distribution does not depend on the original first cluster size distribution and second cluster size distribution output by the visual detection unit and the impact detection unit. Instead, it is directly reconstructed from the geometric morphological features carrying the first confidence level and the impact response features carrying the second confidence level after verification. This can cut off the transmission path of the original distribution deviation of the single mode and improve the anti-interference ability of size estimation under complex working conditions.

[0056] In some embodiments, fusing the first cluster size distribution, the second cluster size distribution, and the third cluster size distribution includes: Based on the first confidence level, determine the first sub-distribution confidence level of the first cluster size distribution; based on the second confidence level, determine the second sub-distribution confidence level of the second cluster size distribution. Based on the first and second confidence levels, determine the reliability index of the third cluster size distribution characteristic; Based on the reliability index and the characteristic consistency index, the third distribution reliability of the third cluster size distribution characteristics is determined, where the characteristic consistency index characterizes the degree of correspondence between geometric morphological characteristics and impact response characteristics. The reliability of the first, second, and third sub-distributions is normalized, and the first, second, and third cluster size distributions are weighted and fused based on the normalization results.

[0057] The first sub-distribution reliability of the first cluster size distribution can be obtained by averaging the first confidence level and the second sub-distribution reliability of the second cluster size distribution by averaging the second confidence levels of the impact response features. For example, if the first confidence levels of multiple geometric features are 1.0, 0.6, 0.6, 0.2, and 1.0, the average value is 0.68, which is taken as the first sub-distribution reliability of the first cluster size distribution. If the first confidence levels of multiple geometric features are 0.2, 0.6, 0.6, 0.2, and 1.0, the average value is 0.52, which is taken as the second sub-distribution reliability of the second cluster size distribution. The first and second sub-distribution reliability reflect the overall reliability of the visual detection unit and the impact detection unit within the current detection cycle, respectively, and serve as the evaluation basis for the reliability of the corresponding original distribution results.

[0058] For the third cluster size distribution, the evaluation of its third sub-distribution reliability takes into account both the validity of the input features and the consistency of the bimodal determination.

[0059] First, based on the first and second confidence levels, a reliability index for the third cluster size distribution feature is determined. In some embodiments, this reliability index can be obtained by taking the arithmetic mean of the independent confidence mapping values ​​of the four types of features (i.e., connected region area, dispersion, impact force amplitude, and frequency domain energy) involved in the reconstruction model, and is used to evaluate the overall effectiveness of the underlying features input into the reconstruction model. Then, a feature consistency index is introduced. This index quantifies the matching deviation between visual morphological features and impact response features, that is, it characterizes the consistency of the bimodal judgments on the same cluster state. The feature consistency index ranges from [0,1], with a larger value indicating higher consistency between the two types of features and a smaller value indicating lower consistency. Then, a weighted coupling calculation is performed based on the reliability index and the feature consistency index, and the calculation results are normalized and truncated to obtain the third distribution confidence constrained within the [0,1] interval.

[0060] Then, the reliability scores of the first, second, and third sub-distributions are normalized proportionally. Based on the normalization results, normalized weights for the three distributions are obtained, denoted as w_vis for the first cluster size distribution, w_imp for the second cluster size distribution, and w_est for the third cluster size distribution, respectively, satisfying w_vis + w_imp + w_est = 1. Next, the first, second, and third cluster size distributions are weighted and fused. The fusion calculation method is as follows: Fusion result = w_vis × first cluster size distribution + w_imp × second cluster size distribution + w_est × third size distribution; For example, if the first cluster size distribution indicates that the proportions of large, medium, and small clusters are 20%, 30%, and 50%, respectively; the second cluster size distribution indicates that the proportions of large, medium, and small clusters are 24%, 36%, and 40%, respectively; and the third cluster size distribution indicates that the proportions of large, medium, and small clusters are 20%, 36%, and 44%, respectively; w_vis is 0.2; w_imp is 0.3; and w_est is 0. If 0.5, then the final fusion result is: large clusters = 0.2 × 20% + 0.3 × 24% + 0.5 × 20% = 21.2%; medium clusters = 0.2 × 30% + 0.3 × 36% + 0.5 × 36% = 34.8%; small clusters = 0.2 × 50% + 0.3 × 40% + 0.5 × 44% = 44%; that is, the proportions of large clusters, medium clusters, and small clusters are 21.2%, 34.8%, and 44%, respectively.

[0061] According to the technical solution provided in this embodiment, through the above-mentioned confidence-based weight allocation and linear weighted fusion, the fusion process always tilts towards the direction of better data quality. Moreover, due to the use of confidence-driven dynamic weight allocation, the high-confidence distribution contributes more to the fusion result, while the contribution of the low-confidence distribution is correspondingly reduced, thereby improving the accuracy of the fusion cluster size distribution.

[0062] In some embodiments, before normalizing the confidence scores of the first sub-distribution, the second sub-distribution, and the third sub-distribution, the method further includes: If any of the confidence levels of the first sub-distribution confidence, the second sub-distribution confidence, and the third sub-distribution confidence is lower than the first preset threshold, then the confidence level of the sub-distribution that is lower than the first preset threshold is set to zero. If any of the confidence levels of the first sub-distribution confidence, the second sub-distribution confidence, and the third sub-distribution confidence is greater than or equal to the first preset threshold and less than the second preset threshold, then the sub-distribution confidence that is greater than or equal to the first preset threshold and less than the second preset threshold is updated by a preset attenuation coefficient.

[0063] For distributions with confidence levels below the first preset threshold, they can be set directly to zero, achieving hard isolation at the algorithm level. This means cutting off the weight allocation path of the distribution during normalized weight calculation, ensuring that distorted data is shielded from fusion decisions, while also maintaining continuous operation capability in the event of a single data source failure.

[0064] For distribution confidence scores greater than or equal to the first preset threshold and less than the second preset threshold, they are corrected and updated by multiplying them by a preset attenuation coefficient less than 1. This preserves some of the remaining valid information in the distribution while appropriately reducing its contribution to the fusion, thus smoothly transitioning to the volatility risk brought about by medium-quality distribution confidence scores.

[0065] For distribution confidence levels greater than or equal to the third preset threshold, no attenuation is applied, and the distribution confidence level can be fully accepted.

[0066] According to the technical solution provided in this embodiment, dynamic allocation of multi-source data streams is realized under complex working conditions. Since low-quality data is shielded or attenuated and high-quality data is fully accepted, the reliability of the fusion cluster size distribution can be guaranteed, and the probability of low-quality or unreliable single-modal distribution data affecting the final fusion result can be reduced.

[0067] In some embodiments, determining the clumping detection result of the cooked grain to be tested based on the fusion result includes: Acquire the fusion results within the target time window; based on the current operating conditions of the cooked grain to be tested, standardize the fusion agglomeration size distribution to obtain a standardized agglomeration size distribution, where the current operating conditions include grain variety, grain layer thickness on the chain conveyor, and chain conveyor speed. Determine the agglomeration detection results of the cooked grain to be tested based on the standardized agglomeration size distribution and preset agglomeration judgment rules.

[0068] It is understandable that the operating state of the grain spreading conveyor belt is dynamically changing on the actual production line of solid-state fermentation brewing. For example, even when agglomerates of a certain area are detected, the severity of agglomeration varies depending on the thickness of the grain layer. Similarly, differences in density and particle size among different grain varieties will cause clumps of the same geometric size to correspond to different physical qualities. If the detection signal is not considered within the specific physical conditions, the output size distribution value will lack guiding significance.

[0069] The target time window is established based on sliding time series statistics, that is, the sliding time window is established by receiving the fusion results of the most recent N moments, where N can be dynamically adjusted according to the speed of the board chain: N takes a smaller value under high-speed conditions to ensure the detection response speed, and N takes a larger value under low-speed conditions to improve statistical stability.

[0070] Then, based on the current working conditions of the cooked grain to be tested, the size distribution of the fused agglomerates is standardized in multiple dimensions.

[0071] In one optional implementation, the standardization process may sequentially include the following conversion steps: First, based on the set value of the grain layer thickness, the fusion result is divided by the set value of the grain layer thickness on the plate chain to obtain the relative size distribution, thereby eliminating the geometric spatial reference difference caused by uneven material thickness; Second, based on the grain variety code, the density and particle size parameters corresponding to the variety are called to correct the size-mass conversion relationship, and the relative size distribution is converted according to the standard density of the grain variety to obtain the standard mass distribution, thereby eliminating the physical reference difference caused by different grain materials and achieving cross-variety comparability; Finally, based on the plate chain speed feedback value, the standard mass distribution is time-normalized according to the plate chain speed to eliminate the time response difference caused by the transmission cycle, and finally output the standardized agglomeration size distribution.

[0072] Then, the proportion of large cluster size range can be statistically analyzed in the standardized cluster size distribution: when the proportion exceeds the preset cluster judgment threshold, the existence of clusters is determined, the cluster existence flag is output as true, and the proportion of large size can be converted into the cluster existence probability through the preset monotonic mapping as the judgment confidence output; when the proportion does not exceed the cluster judgment threshold, no clusters are determined, and the cluster existence flag is output as false.

[0073] According to the technical solution provided in this embodiment, by standardizing the grain layer thickness, grain variety and plate chain speed, the interference of external working condition fluctuations on the detection data can be eliminated. Since the benchmark differences in the three dimensions are eliminated one by one, the final output standardized agglomeration size distribution has comparability across batches and working conditions.

[0074] In some embodiments, the method for detecting clumping of cooked grain provided in this application further includes: real-time monitoring of the operating status of the visual detection unit and the impact detection unit: for the visual detection unit, real-time monitoring of frame rate stability, marking an anomaly when frame rate fluctuation exceeds a preset fluctuation threshold; real-time monitoring of image clarity, marking an anomaly when clarity is lower than a preset clarity threshold; real-time monitoring of the degree of environmental dust interference, marking an anomaly when interference exceeds a preset interference threshold; for the impact detection unit, real-time monitoring of signal-to-noise ratio, marking an anomaly when signal-to-noise ratio is lower than a preset signal-to-noise threshold; real-time monitoring of baseline stability, marking an anomaly when baseline drift exceeds a preset drift threshold; and real-time monitoring of sensor response consistency, marking an anomaly when the difference in response between sensors exceeds a preset difference threshold.

[0075] When any detection unit is determined to be abnormal, a single-modal degradation process can be performed: During the feature interaction verification stage, the confidence scores of all features of the abnormal modality are set to zero; during the distribution fusion stage, the normalization weights corresponding to the abnormal modality are reset to zero; and the system is downgraded to only accepting the fusion result of the distribution of the remaining valid modalities and the third cluster size distribution, with a single-modal degradation flag appended to the output. For example, when a visual detection unit is marked as abnormal, the system sets the first confidence score to zero, sets the normalization weight w_vis to zero, and only accepts the fusion result of the second and third cluster size distributions; the same degradation process is applied when an impact detection unit is marked as abnormal.

[0076] It should be noted that when any detection unit is judged to be abnormal, the feature reliability index of the third cluster size distribution can be recalculated according to the preset missing mode processing rules; when the remaining effective features cannot meet the minimum input conditions of the feature size mapping model, the confidence of the third cluster size distribution can be set to zero.

[0077] When both the visual inspection unit and the impact inspection unit are determined to be abnormal, all automatic inspection outputs can be paused, a fault signal can be sent to the control system, and a prompt can be made to switch to manual intervention mode.

[0078] In this way, through the real-time health status assessment mechanism and the double reset zero processing of confidence and weight, it is ensured that the three-source fusion can be automatically downgraded to two-source fusion when a single mode fails, reducing the impact of fault data on fusion decision-making, maintaining the uninterrupted operation of the system, and improving the availability and anti-interference capability of the system in harsh environments such as high dust, high humidity, and strong mechanical vibration in the brewing workshop.

[0079] In a specific example, during the grain spreading process of a solid-state fermentation brewing production line, cooked grain is spread and transported via a plate chain conveyor. A vision inspection unit is deployed on the grain falling side, while an impact inspection unit is arranged as an array of accelerometer sensors on the load-bearing plate. Upon initial system startup, the system enters the initial registration phase: collecting M minutes of dual-modal operating data, recording the time and location of vision and impact detection events; estimating a reasonable time range based on the material's falling height and gravitational acceleration, searching for matching event pairs within this range, and filtering out valid event pairs with equivalent diameter deviations less than the filtering threshold; calculating the average time difference based on valid event pairs as a fixed time correspondence value, establishing a spatial correspondence table between visual positions and impact sensor numbers, and completing the determination of feature correspondences.

[0080] After entering steady-state production, the system continuously executes the following detection cycle: It receives geometric morphological feature parameters such as connected region area, aspect ratio, and dispersion uploaded by the visual inspection unit, and impact response feature parameters such as impact force amplitude, impact duration, and frequency domain energy distribution uploaded by the impact detection unit. When the visual inspection unit reports a large connected region, it initiates a query to the impact detection unit through the large connected region query signal path to retrieve the impact force amplitude for the corresponding time period for verification. When the impact detection unit reports an abnormal impact force amplitude, it initiates a reverse query to the visual inspection unit through the abnormal impact force query signal path to retrieve the morphological features of the corresponding visual region for verification. The verification results are fed back to the corresponding detection unit with confidence levels, and the confidence levels of each feature are corrected according to three levels (high, medium, and low, corresponding to mapping values ​​of 1.0, 0.6, and 0.2), outputting the verified feature parameters and confidence level labels.

[0081] Then, the system receives the first cluster size distribution independently estimated by the visual detection unit and the second cluster size distribution independently estimated by the impact detection unit, both represented as the proportions of large, medium, and small clusters. Simultaneously, based on four types of features—connected region area, dispersion, impact force amplitude, and frequency domain energy—a comprehensive cluster index is calculated using feature fusion mapping rules. This index is then used to generate a third cluster size distribution through a membership matching algorithm and normalization processing, and its third distribution confidence is calculated. The system performs hierarchical adaptive correction (zeroing, attenuation, or retention) and normalization on the confidence of the three distributions, obtains normalized weights, performs linear weighted fusion, and outputs the fusion result.

[0082] Next, a sliding time window is established by receiving the fusion agglomeration size distribution from the most recent N time points, where N is dynamically adjusted according to the chain speed. The system receives the grain variety code, grain layer thickness setting, and chain speed feedback value for the current batch, and sequentially performs thickness normalization, variety density conversion, and time normalization to output a standardized agglomeration size distribution. The proportion of large agglomeration size ranges is statistically analyzed and compared with an agglomeration judgment threshold dynamically adjusted according to the air permeability requirements of subsequent saccharification processes. When the proportion exceeds the threshold, a true agglomeration presence flag is output along with the judgment confidence level; when the proportion does not exceed the threshold, a false agglomeration presence flag is output. Production managers can intervene in the grain spreading process promptly based on the agglomeration detection results, thereby ensuring the quality of subsequent saccharification and fermentation processes.

[0083] It should be understood that the order of the steps in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the process of the embodiments of this application.

[0084] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0085] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0086] Figure 2 This is a schematic diagram illustrating the structure of a cooked grain clumping detection device, as shown in an exemplary embodiment of this application. Figure 2 As shown, the exemplary cooked grain clumping detection device includes: The feature acquisition module 210 is configured to acquire the initial geometric morphological features and initial impact response features of the cooked grain to be detected; The feature verification module 220 is configured to perform confidence-based interactive verification on the initial geometric morphology features and the initial impact response features respectively, to obtain geometric morphology features carrying a first confidence level and impact response features carrying a second confidence level. The first determining module 230 is configured to determine the first agglomeration size distribution of the cooked grain to be tested based on the initial geometric morphology features, and to determine the second agglomeration size distribution of the cooked grain to be tested based on the initial impact response features. The second determining module 240 is configured to determine the third agglomeration size distribution based on the geometric morphological features carrying a first confidence level and the impact response features carrying a second confidence level. The result detection module 250 is configured to fuse the first agglomeration size distribution, the second agglomeration size distribution, and the third agglomeration size distribution, and determine the agglomeration detection result of the cooked grain to be tested based on the fusion result.

[0087] In some embodiments, the feature verification module 220 is further configured to: if a first preset event exists in the initial geometric morphology features, determine the target impact response feature from the initial impact response features according to a predetermined feature correspondence to perform confidence interactive verification on the first preset event, and determine the first confidence level of the initial geometric morphology features based on the verification result; if a second preset event exists in the initial impact response features, determine the target geometric morphology feature from the initial geometric morphology features according to the feature correspondence to perform confidence interactive verification on the second preset event, and determine the second confidence level of the initial impact response features based on the verification result.

[0088] In some embodiments, the feature verification module 220 is further configured to acquire the sample geometric morphology features and sample impact response features of the cooked grain sample to be detected within a preset time period; determine the time boundary based on the falling height of the cooked grain sample to be detected; determine the sample event pair between the first preset event in the sample geometric morphology features and the second preset event in the sample impact response features within the time boundary; and determine the feature correspondence based on the sample event pair.

[0089] In some embodiments, the geometric morphological features include the area and dispersion of connected regions, and the impact response features include the impact force amplitude and frequency domain energy. The second determining module 240 is further configured to determine a comprehensive agglomeration index based on the area and dispersion of connected regions carrying a first confidence level, and the impact force amplitude and frequency domain energy carrying a second confidence level; determine the initial agglomeration size matching probability of the cooked grain to be tested based on the comprehensive agglomeration index and a preset agglomeration grading threshold; and normalize the initial matching probability to obtain a third agglomeration size distribution.

[0090] In some embodiments, the result detection module 250 is further configured to: determine a first sub-distribution confidence level of the first cluster size distribution based on a first confidence level; determine a second sub-distribution confidence level of the second cluster size distribution based on a second confidence level; determine a reliability index of the third cluster size distribution feature based on the first confidence level and the second confidence level; determine a third sub-distribution confidence level of the third cluster size distribution feature based on the reliability index and the feature consistency index, wherein the feature consistency index characterizes the degree of correspondence between geometric morphological features and impact response features; normalize the first sub-distribution confidence level, the second sub-distribution confidence level, and the third sub-distribution confidence level; and perform weighted fusion of the first cluster size distribution, the second cluster size distribution, and the third cluster size distribution based on the normalization result.

[0091] In some embodiments, the result detection module 250 is further configured to: if any one of the first sub-distribution confidence, the second sub-distribution confidence, and the third sub-distribution confidence is lower than a first preset threshold, then set the sub-distribution confidence below the first preset threshold to zero; if any one of the first sub-distribution confidence, the second sub-distribution confidence, and the third sub-distribution confidence is greater than or equal to the first preset threshold and less than the second preset threshold, then update the sub-distribution confidence greater than or equal to the first preset threshold and less than the second preset threshold using a preset attenuation coefficient.

[0092] In some embodiments, the result detection module 250 is further configured to acquire the fusion result within the target time window; standardize the fusion agglomeration size distribution according to the current working condition characteristics of the cooked grain to be detected, and obtain a standardized agglomeration size distribution, wherein the current working condition characteristics include grain variety, grain layer thickness on the plate chain and plate chain speed; and determine the agglomeration detection result of the cooked grain to be detected according to the standardized agglomeration size distribution and the preset agglomeration judgment rules.

[0093] Embodiments of this application also provide an electronic device, including: one or more processors; and a storage device for storing one or more programs, which, when executed by one or more processors, cause the electronic device to implement the methods provided in the above embodiments.

[0094] Figure 3A schematic diagram of a computer system suitable for implementing the embodiments of this application is shown. It should be noted that... Figure 3 The computer system 300 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0095] like Figure 3 As shown, the computer system 300 includes a Central Processing Unit (CPU) 301, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 302 or programs loaded from storage portion 308 into Random Access Memory (RAM) 303, such as executing the methods described in the above embodiments. The RAM 303 also stores various programs and data required for system operation. The CPU 301, ROM 302, and RAM 303 are interconnected via a bus 304. An Input / Output (I / O) interface 305 is also connected to the bus 304.

[0096] The following components are connected to I / O interface 305: an input section 306 including a keyboard, mouse, etc.; an output section 307 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to I / O interface 305 as needed. Removable media 311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 310 as needed so that computer programs read from them can be installed into storage section 308 as needed.

[0097] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including a computer program for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 309, and / or installed from removable medium 311. When the computer program is executed by central processing unit (CPU) 301, it performs various functions defined in the system of this application.

[0098] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. A computer-readable storage medium can be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable computer program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0099] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation that may be implemented in systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0100] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0101] Another aspect of this application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a computer's processor, causes the computer to perform the method as described above. This computer-readable storage medium may be included in the electronic device described in the above embodiments, or it may exist independently and not assembled into the electronic device.

[0102] Another aspect of this application provides a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods described in the various embodiments above.

[0103] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the steps of this application.

Claims

1. A method for detecting clumping of cooked grains, characterized in that, include: Acquire the initial geometric morphological features and initial impact response features of the cooked grain to be tested; The initial geometric morphological features and the initial impact response features are respectively subjected to confidence cross-validation to obtain geometric morphological features carrying a first confidence level and impact response features carrying a second confidence level. The first agglomeration size distribution of the cooked grain to be tested is determined based on the initial geometric morphological characteristics, and the second agglomeration size distribution of the cooked grain to be tested is determined based on the initial impact response characteristics. The third cluster size distribution is determined based on the geometric morphological characteristics carrying the first confidence level and the impact response characteristics carrying the second confidence level. The first agglomeration size distribution, the second agglomeration size distribution, and the third agglomeration size distribution are fused together, and the agglomeration detection result of the cooked grain to be tested is determined based on the fusion result.

2. The method according to claim 1, characterized in that, The confidence cross-validation of the initial geometric features and the initial impact response features includes: If a first preset event exists in the initial geometric morphology features, then the target impact response feature is determined from the initial impact response features according to the pre-determined feature correspondence relationship to perform confidence interactive verification on the first preset event, and the first confidence level of the initial geometric morphology features is determined based on the verification result. If a second preset event exists in the initial impact response features, then the target geometric morphology features are determined from the initial geometric morphology features according to the feature correspondence to perform confidence interaction verification on the second preset event, and the second confidence level of the initial impact response features is determined based on the verification result.

3. The method according to claim 2, characterized in that, Before performing confidence cross-validation on the initial geometric features and the initial impact response features respectively, the method further includes: Obtain the geometric morphological features and impact response features of the cooked grain samples to be tested within a preset time period; The time boundary is determined based on the falling height of the cooked grain sample to be tested, and within the time boundary, a sample event pair is determined between the first preset event in the geometric morphology features of the sample and the second preset event in the impact response features of the sample. The feature correspondence is determined based on the sample event pairs.

4. The method according to claim 1, characterized in that, The geometric features include the area and dispersion of the connected domains, and the impact response features include the impact force amplitude and the frequency domain energy. Determining the third agglomeration size distribution based on the geometric morphological features carrying a first confidence level and the impact response features carrying a second confidence level includes: The comprehensive clumping index is determined based on the connected region area and the dispersion carrying a first confidence level, and the impact force amplitude and the frequency domain energy carrying a second confidence level. Based on the comprehensive agglomeration index and the preset agglomeration grading threshold, the initial agglomeration size matching probability of the cooked grain to be tested is determined; The initial matching probability is normalized to obtain the third cluster size distribution.

5. The method according to claim 1, characterized in that, The fusion of the first cluster size distribution, the second cluster size distribution, and the third cluster size distribution includes: Based on the first confidence level, determine the first sub-distribution confidence level of the first cluster size distribution; based on the second confidence level, determine the second sub-distribution confidence level of the second cluster size distribution; Based on the first confidence level and the second confidence level, determine the reliability index of the third cluster size distribution characteristic; Based on the reliability index and the feature consistency index, the third distribution reliability of the third cluster size distribution feature is determined, wherein the feature consistency index characterizes the degree of correspondence between the geometric morphology feature and the impact response feature; The confidence scores of the first sub-distribution, the second sub-distribution, and the third sub-distribution are normalized. Based on the normalization results, the first cluster size distribution, the second cluster size distribution, and the third cluster size distribution are weighted and fused.

6. The method according to claim 5, characterized in that, Before normalizing the confidence scores of the first, second, and third sub-distributions, the method further includes: If any of the confidence levels of the first sub-arrangement, the second sub-arrangement, and the third sub-arrangement is lower than a first preset threshold, then the sub-arrangement confidence level that is lower than the first preset threshold is set to zero. If any one of the confidence levels of the first sub-arrangement confidence, the second sub-arrangement confidence, and the third sub-arrangement confidence is greater than or equal to the first preset threshold and less than the second preset threshold, then the sub-arrangement confidence that is greater than or equal to the first preset threshold and less than the second preset threshold is updated by a preset attenuation coefficient.

7. The method according to claim 1, characterized in that, The step of determining the clumping detection result of the cooked grain to be tested based on the fusion result includes: Obtain the fusion results within the target time window; Based on the current working condition characteristics of the cooked grain to be tested, the fusion agglomeration size distribution is standardized to obtain a standardized agglomeration size distribution, wherein the current working condition characteristics include grain variety, grain layer thickness on the plate chain, and plate chain speed. Based on the standardized agglomeration size distribution and the preset agglomeration judgment rules, the agglomeration detection result of the cooked grain to be tested is determined.

8. A device for detecting clumping of cooked grain, characterized in that, include: The feature acquisition module is configured to acquire the initial geometric morphological features and initial impact response features of the cooked grain to be detected; The feature verification module is configured to perform confidence-based interactive verification on the initial geometric morphology features and the initial impact response features respectively, to obtain geometric morphology features carrying a first confidence level and impact response features carrying a second confidence level. The first determining module is configured to determine the first agglomeration size distribution of the cooked grain to be tested based on the initial geometric morphological features, and to determine the second agglomeration size distribution of the cooked grain to be tested based on the initial impact response features. The second determining module is configured to determine the third agglomeration size distribution based on the geometric morphological features carrying a first confidence level and the impact response features carrying a second confidence level. The result detection module is configured to fuse the first agglomeration size distribution, the second agglomeration size distribution, and the third agglomeration size distribution, and determine the agglomeration detection result of the cooked grain to be detected based on the fusion result.

9. An electronic device, characterized in that, include: One or more processors and a memory, the memory storing a computer program that, when executed by the one or more processors, causes the device to perform the steps of the cooked grain clumping detection method as described in any one of claims 1 to 7.

10. A readable storage medium, characterized in that, It stores a computer program that, when executed by one or more processors, causes the device to perform the steps of the cooked grain clumping detection method as described in any one of claims 1 to 7.

Images