Micro-mineral organic fertilizer quality intelligent detection method and system

Abstract

The application relates to the technical field of material testing, and specifically discloses a quality intelligent detection method and system for micro-mineral biological organic fertilizer, which comprises the following steps: applying local thermal excitation to a fertilizer sample through a micro-zone thermal pulse excitation device, synchronously collecting a gas release kinetics curve and a spectral change track, and constructing a gas spectrum coupling data cube; then performing differential transformation and modal decomposition on a time sequence feature set, extracting a transient response characteristic vector, and generating an activity response distribution graph; then establishing component-activity correlation analysis, decoupling biological metabolic activity and matrix background interference through characteristic separation and contrast learning strategies, and obtaining a dynamic metabolic fingerprint; further constructing a multi-dimensional feature space by combining the dynamic metabolic fingerprint with apparent characteristic parameters, adopting a graph convolution network to perform information aggregation and feature reconstruction, and generating a comprehensive quality index; and finally realizing quality grade determination based on a quality feature pyramid structure, and simultaneously establishing a dynamic early warning mechanism by analyzing time sequence evolution characteristics of the activity response distribution graph.

Description

Technical Field

[0001] This invention relates to the field of materials testing technology, specifically to a method and system for intelligent quality detection of micro-mineral bio-organic fertilizer. Background Technology

[0002] Micro-mineral bio-organic fertilizer, as a novel fertilizer integrating organic matter, microbial agents, and trace elements, relies heavily on the quantity and metabolic capacity of active microorganisms for its quality. However, due to the complex composition of the fertilizer, including peat moss, humic acid, attapulgite, and other additives, traditional detection methods relying on chemical analysis or single sensors (such as near-infrared spectroscopy or gas sensors) are insufficient for accurate and non-destructive assessment of its microbial activity. Existing technologies typically rely on offline sampling, laboratory cultivation, or static spectral analysis for quality judgment, which suffers from cumbersome processes, poor timeliness, and susceptibility to subjective experience, failing to meet the demands of modern agricultural production for real-time monitoring and intelligent management of fertilizer quality.

[0003] The existing technology has the following shortcomings:

[0004] Porous adsorbent additives in fertilizers (such as zeolite and attapulgite) strongly adsorb trace characteristic gases (such as volatile organic acids) produced by microbial metabolism, leading to severe attenuation of gas sensor signals and causing false negatives. Simultaneously, organic matter such as peat moss and humic acid creates strong background spectral signals in the near-infrared band, completely obscuring the weak spectral changes caused by microbial metabolism. This matrix interference makes traditional data analysis methods (such as static threshold judgment or simple regression models) unable to effectively distinguish between the true metabolic signals of microorganisms and matrix background noise. This causes the detection system to misclassify highly active fertilizers as "unqualified" or to misclassify inferior fertilizers that produce gas due to contamination by other microorganisms as "highly active," severely limiting the reliability and applicability of quality testing. Summary of the Invention

[0005] The purpose of this invention is to provide a method and system for intelligent quality detection of micro-mineral bio-organic fertilizer, so as to solve the problems mentioned above.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A smart quality detection method for micro-mineral bio-organic fertilizer includes the following steps:

[0008] S1: By applying energy excitation in a preset mode and simultaneously acquiring multi-source dynamic response signals of fertilizer samples during the excitation period, a time-series feature set containing gas release kinetic curves and spectral change trajectories is constructed.

[0009] S2: Perform differential transformation and mode decomposition on the time-series feature set to extract transient response feature vectors characterizing the intensity of biological metabolism and generate corresponding activity response distribution maps;

[0010] S3: Based on the activity response distribution map, a component-activity correlation model is established. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained.

[0011] S4: Construct a multi-dimensional feature space by combining dynamic metabolic fingerprints and sample apparent characteristic parameters, and use graph convolutional networks to aggregate information and reconstruct features of feature nodes to generate comprehensive quality indicators.

[0012] S5: Generate quality level judgment results based on comprehensive quality indicators, and analyze the temporal evolution characteristics of the activity response distribution map to establish a dynamic quality early warning threshold.

[0013] As a further aspect of the present invention: the construction of a time-series feature set including gas release kinetic curves and spectral change trajectories specifically includes:

[0014] A micro-area thermal pulse excitation device was used to apply local thermal excitation to the sample surface with a duration of milliseconds and a temperature change range of 3 to 5 degrees Celsius. Hyperspectral image sequences of specific areas of the sample surface were acquired before and after the excitation.

[0015] By arranging an array of gas sensors around the sample space, the instantaneous changes in the concentration of various characteristic gases are monitored synchronously at a millisecond sampling frequency, and the gas release rate curves and their spectral characteristics are recorded.

[0016] The hyperspectral image sequence was time-stamped with the gas release rate curve, and the gradient of spectral reflectance change at each sampling time was extracted to construct a gas-spectral coupled data cube containing spatiotemporal features.

[0017] The gas-spectral coupled data cube is decomposed into multi-scale components to extract fast response components and slow-varying background components, and the fast response components are reconstructed into a time-series feature set reflecting the dynamics of microbial metabolism.

[0018] As a further aspect of the present invention: the step of performing differential transformation and mode decomposition on the time-series feature set to extract transient response feature vectors characterizing the intensity of biological metabolism specifically includes:

[0019] Higher-order differential processing is performed on gas release kinetic curves with time-series feature sets to calculate their first-order and second-order differential sequences, thereby capturing abrupt changes and trends in gas release rates.

[0020] Adaptive intrinsic mode decomposition is performed on the spectral change trajectory to decompose the non-stationary signal into multiple intrinsic mode components at different time scales, and feature modes that match the microbial metabolic cycle are screened.

[0021] Extract the amplitude features of abrupt change points and the energy distribution parameters of characteristic modes, and construct a multi-dimensional transient response feature vector containing time-domain abrupt change features and frequency-domain energy features;

[0022] Dimensionality reduction is performed on the multidimensional transient response feature vector to obtain the core feature subset characterizing the intensity of biological metabolism.

[0023] As a further aspect of the present invention: the generation of the corresponding activity response distribution map specifically includes:

[0024] The transient response feature vectors are reorganized according to the sampling spatial location to construct a feature intensity matrix based on a spatial grid;

[0025] The radial basis function interpolation method is used to spatially interpolate the characteristic intensity matrix to generate a continuously distributed contour map of activity intensity.

[0026] Based on the gradient change characteristics of each region in the activity intensity contour map, response regions of different activity levels are divided, and metabolic hotspot regions are marked.

[0027] By superimposing and fusing the spatial distribution of gas release characteristics and spectral change characteristics, an activity response distribution map that comprehensively reflects the spatial distribution and metabolic intensity of microorganisms is generated.

[0028] As a further aspect of the present invention: the establishment of the component-activity correlation model based on the activity response distribution map specifically includes:

[0029] Using the spatial gradient features and intensity distribution features in the activity response distribution map as input, a dual-channel feature matrix containing matrix background features and metabolic activity features is constructed.

[0030] An adversarial Siamese network model is established. The generator network learns to separate matrix background features from the dual-channel feature matrix, while the discriminator network is used to identify the integrity of metabolic activity features.

[0031] The separated metabolic activity features are input into the residual mapping network to establish a nonlinear mapping relationship between metabolic activity intensity and microbial quantity, and output a pure activity feature vector that removes matrix interference.

[0032] The residual mapping network is optimized by a contrastive learning strategy to ensure that the output features of the same activity level remain consistent under different matrix backgrounds.

[0033] As a further aspect of the present invention: obtaining the dynamic metabolic fingerprint that reflects the true activity of microorganisms specifically includes:

[0034] The pure activity feature vectors are recombined according to the time series to construct a dynamic metabolic trajectory matrix;

[0035] Singular value decomposition is performed on the dynamic metabolic trajectory matrix to extract intrinsic metabolic patterns that characterize the main features of the metabolic process.

[0036] The intrinsic metabolic pattern is dynamically time-warped and matched with the standard metabolic template to calculate the metabolic pathway similarity index.

[0037] Based on the metabolic pathway similarity index and the energy distribution of intrinsic metabolic patterns, a dynamic metabolic fingerprint with unique identifiers is generated.

[0038] As a further aspect of the present invention: the generation of the comprehensive quality index specifically includes:

[0039] The various feature dimensions of the dynamic metabolic fingerprint and the apparent characteristic parameters of the sample are used to construct feature nodes, and node connection edges are established based on the mutual information between features to form a dynamic feature topology graph.

[0040] By aggregating the neighborhood information of each feature node layer by layer through graph convolutional layers, global related features are fused while local features are preserved.

[0041] The weights in the information transmission process are dynamically adjusted based on the contribution of feature nodes to the quality assessment.

[0042] Global pooling is performed on feature nodes that have undergone multi-layer graph convolution processing, and the output is a comprehensive quality index that includes local features and global correlations.

[0043] As a further aspect of the present invention: the generation of quality grade determination results based on comprehensive quality indicators specifically includes:

[0044] Construct a hierarchical decision tree based on the quality feature pyramid, and input the comprehensive quality indicators into the corresponding decision nodes according to the hierarchical structure of the quality feature pyramid;

[0045] At each decision node, an adaptive membership function is used to perform fuzzy quantization on the input features, and the probability distribution of each quality level is calculated.

[0046] The high-frequency fluctuations of the underlying micro-characteristics are integrated with the stable trends of the top-level macro-characteristics for analysis.

[0047] Based on the comparison between the final confidence score and the preset level threshold, the quality level determination result with confidence assessment is output.

[0048] As a further aspect of the present invention: the analysis of the temporal evolution characteristics of the active response distribution map and the establishment of a dynamic quality early warning threshold specifically includes:

[0049] Spatial distribution characteristic parameters of the activity response distribution map during continuous detection were extracted, including activity intensity gradient, area of ​​metabolic hotspot region and spatial heterogeneity index;

[0050] Calculate the differential characteristics of spatial distribution feature parameters over time to identify trend changes and abnormal fluctuation patterns in the activity evolution process;

[0051] Based on the acceleration characteristics of trend changes and the statistical characteristics of abnormal fluctuations, a dynamic early warning threshold curve is constructed.

[0052] By comparing and analyzing historical early warning data with current monitoring data, the morphological parameters of the early warning threshold curve are optimized. When the activity evolution characteristics exceed the early warning threshold, a graded early warning signal is triggered and the abnormal feature map is recorded.

[0053] A smart quality detection system for micro-mineral bio-organic fertilizer includes:

[0054] The dynamic excitation and multi-source signal acquisition module applies energy excitation in a preset mode and simultaneously acquires multi-source dynamic response signals of fertilizer samples during the excitation period, constructing a time-series feature set including gas release kinetic curves and spectral change trajectories.

[0055] The transient feature extraction and visualization module is used to perform differential transformation and mode decomposition on the time-series feature set, extract transient response feature vectors that characterize the intensity of biological metabolism, and generate corresponding activity response distribution maps.

[0056] The metabolic fingerprint decoupling and analysis module establishes a component-activity correlation model based on the activity response distribution map. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained.

[0057] The multimodal feature fusion and quality index generation module constructs a multidimensional feature space from dynamic metabolic fingerprints and sample phenotypic parameters, and uses a graph convolutional network to aggregate information and reconstruct features from feature nodes to generate a comprehensive quality index.

[0058] The intelligent decision-making and dynamic early warning module generates quality level judgment results based on comprehensive quality indicators, and analyzes the temporal evolution characteristics of the active response distribution map to establish a dynamic quality early warning threshold.

[0059] The beneficial effects of this invention are:

[0060] (1) This invention uses micro-region thermal pulse excitation to stimulate microbial specific responses, combined with gas-spectral coupling data cube construction and multi-scale signal decomposition, to effectively separate microbial metabolic signals from matrix background noise. Through differential transformation and mode decomposition of temporal feature sets, transient response features characterizing the intensity of biological metabolism are further extracted. Then, through adversarial learning and contrastive learning strategies in component-activity correlation analysis, accurate assessment of the true activity of microorganisms under different formulation conditions is achieved, solving the detection distortion problem caused by matrix shielding effect in traditional methods.

[0061] (2) This invention deeply integrates dynamic metabolic fingerprints and apparent characteristic parameters through graph convolutional networks to construct a comprehensive quality index, overcoming the limitations of single-index evaluation. The hierarchical decision structure based on the quality feature pyramid comprehensively considers microscopic metabolic characteristics and macroscopic physical properties, ensuring the scientific nature of quality level determination through multi-level information fusion and confidence assessment. Simultaneously, a dynamic early warning mechanism is established through temporal evolution analysis of the activity response distribution map, enabling early detection of abnormal quality trends and achieving a shift from passive detection to proactive early warning, providing effective technical support for quality control in the production process. Attached Figure Description

[0062] The invention will now be further described with reference to the accompanying drawings.

[0063] Figure 1 This is a flowchart of the method of the present invention;

[0064] Figure 2 This is a flowchart of the system of the present invention. Detailed Implementation

[0065] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0066] Please see Figure 1 As shown, this invention provides an intelligent quality detection method for micro-mineral bio-organic fertilizer, comprising the following steps:

[0067] S1: By applying energy excitation in a preset mode and simultaneously acquiring multi-source dynamic response signals of fertilizer samples during the excitation period, a time-series feature set containing gas release kinetic curves and spectral change trajectories is constructed.

[0068] S2: Perform differential transformation and mode decomposition on the time-series feature set to extract transient response feature vectors characterizing the intensity of biological metabolism and generate corresponding activity response distribution maps;

[0069] S3: Based on the activity response distribution map, a component-activity correlation model is established. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained.

[0070] S4: Construct a multi-dimensional feature space by combining dynamic metabolic fingerprints and sample apparent characteristic parameters, and use graph convolutional networks to aggregate information and reconstruct features of feature nodes to generate comprehensive quality indicators.

[0071] S5: Generate quality level judgment results based on comprehensive quality indicators, and analyze the temporal evolution characteristics of the activity response distribution map to establish a dynamic quality early warning threshold.

[0072] In S1, by applying an energy excitation mode with a preset pattern and simultaneously acquiring multi-source dynamic response signals of fertilizer samples during the excitation period, a time-series feature set containing gas release kinetic curves and spectral change trajectories is constructed, specifically including:

[0073] A micro-area thermal pulse excitation device was used to apply localized thermal excitation to the sample surface. This device, consisting of a precision temperature controller and a miniature heating probe, generates pulsed thermal excitation lasting 50-100 milliseconds, raising the temperature of a specific area on the sample surface by 3-5 degrees Celsius in a short time. The diameter of the thermal excitation area was controlled within the range of 5-8 millimeters to ensure concentrated excitation energy without damaging the sample. Before the thermal excitation began, a reference spectral image of the sample surface was acquired using a hyperspectral imaging system. During and after the thermal excitation, a sequence of hyperspectral images of the sample surface was continuously acquired at a rate of 20 frames per second, recording changes in spectral reflectance within the wavelength range of 400-1000 nanometers.

[0074] Secondly, the release dynamics of characteristic gases are simultaneously monitored by a gas sensor array deployed within the sample detection chamber. This array includes carbon dioxide, ammonia, and volatile organic compound sensors, each with millisecond-level response capabilities. Simultaneously with the thermal excitation, all sensors record gas concentration data at a sampling frequency of 100 Hz. By analyzing the rate of change of gas concentration over time, the gas release rate curve is calculated; simultaneously, a Fast Fourier Transform is performed on the release rate data to extract its spectral characteristics, including the dominant frequency component and frequency band energy distribution.

[0075] Then, the acquired hyperspectral image sequences and gas release rate curves were time-synchronized. A unified time axis coordinate system was established, and the gas concentration data at each sampling moment was registered with the corresponding spectral image. From the registered data, the gradient of spectral reflectance change at each spatial location at each sampling moment was extracted. The calculation method was to compare the difference between the reflectance at the current moment and the reference reflectance, and to combine the changing trends of adjacent sampling points. These spatiotemporally correlated data were organized into a three-dimensional structure, where two dimensions represent the spatial coordinates of the sample surface and the third dimension represents time, thus constructing a gas-spectral coupled data cube.

[0076] Finally, multi-scale signal decomposition was performed on the gas-spectral coupled data cube. A separation method based on signal change rate was used to divide the signal components in the data cube into two parts: a fast response component and a slow-varying background component. The fast response component mainly corresponds to the instantaneous changes in signals generated by microbial metabolic activity, with a time constant below the second level; the slow-varying background component mainly originates from the physicochemical properties of the matrix material, and its change rate is relatively slow. By setting appropriate time windows and change rate thresholds, the fast response component was extracted from the original data and reorganized into a time-series feature set reflecting the dynamics of microbial metabolism. This time-series feature set contains the instantaneous metabolic response characteristics of microorganisms under thermal stimulation, providing a reliable data foundation for subsequent activity assessment.

[0077] In S2, differential transformation and mode decomposition are performed on the time-series feature set to extract transient response feature vectors characterizing the intensity of biological metabolism, and corresponding activity response distribution maps are generated, specifically including:

[0078] The gas release kinetic curves in the time-series feature set are processed by differentiation. The first-order differential sequence of the gas concentration data is calculated using the central difference method, reflecting the changes in the gas release rate. Based on this, the second-order differential sequence is calculated, reflecting the acceleration changes in the release rate. By setting appropriate thresholds, extreme points in the first-order differential sequence and zero-crossing points in the second-order differential sequence are identified; these points correspond to abrupt changes in the gas release process. The time coordinates, magnitudes, and trends before and after each abrupt change point are recorded, forming a set of abrupt change features of gas release.

[0079] The spectral trajectory was decomposed. An adaptive decomposition method based on local extrema was used to decompose the non-stationary spectral signal into multiple intrinsic modal components. The specific process included: identifying local maxima and minima of the spectral signal; constructing upper and lower envelopes using cubic spline interpolation; calculating the mean of the upper and lower envelopes and subtracting this mean from the original signal; repeating this process until the conditions for intrinsic modal components were met. All decomposed components were sorted according to their average period length, and components with periods ranging from 0.5 seconds to 5 seconds were selected, as these components correspond to the timescale of microbial metabolic activity.

[0080] Feature parameters are extracted from the processing results. For abrupt gas release points, parameters such as relative amplitude, duration, and slope of change before and after are extracted. For the selected spectral feature modes, the energy value, center frequency, and bandwidth parameters of each mode are calculated. These time-domain and frequency-domain features are combined in a fixed order to form a multi-dimensional transient response feature vector. Since this feature vector may have a high dimensionality, principal component analysis is used for dimensionality reduction. By calculating the covariance matrix and eigenvectors of the features, principal components with a cumulative contribution rate exceeding 85% are selected to form the core feature subset.

[0081] To generate the activity response distribution map, the transient response feature vector is first reorganized according to the spatial coordinates at the time of sampling. A two-dimensional grid coordinate system is established based on the distribution of sampling points on the sample surface, and the feature intensity value of each sampling point is filled into the corresponding grid position to form a feature intensity matrix. For locations with missing data in the grid, a distance-weighted radial basis function interpolation method is used for filling, specifically using a Gaussian function as the basis function, whose shape parameters are adaptively determined based on the spatial distribution density of the data points.

[0082] An activity intensity contour map is generated based on the interpolated feature intensity matrix. In a two-dimensional plane, points with the same feature intensity are connected to form a smooth curve, with the spacing between contour lines uniformly distributed according to the global distribution range of the feature intensity. By analyzing the gradient distribution characteristics of the contour map, i.e., the density of the contour lines, the sample region is divided into high-activity, medium-activity, and low-activity regions. Areas with dense contour lines and high intensity values ​​are marked as metabolic hotspots.

[0083] The spatial distribution maps of gas release characteristics and spectral variation characteristics are overlaid. After coordinate registration of the two feature maps, a weighted average method is used for fusion, with the weights determined based on the signal-to-noise ratio of each feature. The intensity of the fused features is visualized through color mapping, with red representing high-activity regions and blue representing low-activity regions, generating an activity response distribution map that comprehensively reflects the spatial distribution and metabolic intensity of microorganisms. This distribution map can intuitively display the spatial heterogeneity of microbial activity in the sample, providing a visual basis for quality assessment.

[0084] In S3, a component-activity correlation model is established based on the activity response distribution map. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained.

[0085] Using the spatial gradient and intensity distribution features from the activity response distribution map as input data, a dual-channel feature matrix incorporating matrix background features and metabolic activity features was constructed. Spatial gradient features were obtained by calculating the feature intensity difference between adjacent sampling points, while intensity distribution features were obtained by statistically analyzing the feature intensity distribution histogram for each region. These two types of features were organized into two independent data channels: the first channel primarily contains slowly changing background features related to the matrix material, and the second channel primarily contains rapidly changing features related to microbial metabolism.

[0086] An analytical framework comprising a feature separation network and a feature discrimination network is constructed. The feature separation network is responsible for learning to distinguish between matrix background features and metabolic activity features from a dual-channel feature matrix, achieving layer-by-layer feature separation through a multi-layer neural network. The feature discrimination network is responsible for evaluating the integrity of the separated metabolic activity features, ensuring that complete metabolic activity information is preserved while removing matrix interference. The two networks work collaboratively through alternating training. During training, the feature separation network continuously optimizes its separation ability, while the feature discrimination network continuously improves the accuracy of its judgment on feature integrity.

[0087] The separated metabolic activity features are input into a deep network with a skip connection structure for further processing. This network contains multiple processing layers, each performing a nonlinear transformation on the input features while preserving the original feature information through skip connections. In this way, a complex nonlinear mapping relationship is established between metabolic activity intensity and microbial abundance. The network output is a pure activity feature vector, free from matrix interference, which more accurately reflects the true metabolic state of the microorganisms.

[0088] To improve the stability of the analysis results under different matrix conditions, a contrastive learning strategy was employed to optimize the above processing procedure. Specifically, during training, samples with the same activity level but different matrix backgrounds were treated as positive sample pairs, and samples with different activity levels were treated as negative sample pairs. By optimizing the network parameters, the feature vectors output by the positive sample pairs after network processing exhibited high similarity, while the feature vectors of the negative sample pairs showed significant differences. This approach effectively enhances the system's adaptability to different fertilizer formulations.

[0089] Based on the obtained pure activity feature vectors, a dynamic metabolic fingerprint is further constructed. First, the pure activity feature vectors from multiple consecutive sampling times are organized chronologically to construct a dynamic metabolic trajectory matrix. The rows of this matrix correspond to different feature dimensions, and the columns correspond to different sampling time points. By performing time series analysis on this matrix, the dynamic changes in microbial metabolic activities can be obtained.

[0090] The dynamic metabolic trajectory matrix is ​​subjected to matrix decomposition to extract intrinsic metabolic patterns that characterize the main features of the metabolic process. Specifically, the covariance matrix of the matrix is ​​calculated, and then the eigenvectors and eigenvalues ​​of the covariance matrix are solved. The top few eigenvectors with larger eigenvalues ​​are selected as intrinsic metabolic patterns, representing the most significant changes in microbial metabolism.

[0091] The extracted intrinsic metabolic patterns are matched and compared with pre-established standard metabolic templates. These standard metabolic templates are reference patterns established through extensive experiments on standard samples with known activity levels. The matching process employs a dynamic time warping algorithm, which can handle pattern matching problems at different time scales. The metabolic pathway similarity index, reflecting the degree of similarity between the test sample and the standard template, is obtained by calculating the minimum alignment distance between the two pattern sequences.

[0092] A unique dynamic metabolic fingerprint is generated based on the metabolic pathway similarity index and the energy distribution of intrinsic metabolic patterns. The energy distribution is obtained by calculating the proportion of the feature value corresponding to each intrinsic metabolic pattern to the total feature value. The similarity index and energy distribution parameters are combined according to a fixed coding rule to form a feature vector containing multiple dimensions, which is the dynamic metabolic fingerprint. This fingerprint can comprehensively and accurately reflect the characteristic state of microbial metabolic activities, providing a reliable basis for subsequent quality assessment.

[0093] In S4, dynamic metabolic fingerprints and sample phenotypic parameters are constructed into a multidimensional feature space. A graph convolutional network is used to aggregate information and reconstruct features from feature nodes, generating a comprehensive quality index, which specifically includes:

[0094] The dynamic metabolic fingerprint is constructed by combining its various feature dimensions with sample phenotypic parameters to form feature nodes. The dynamic metabolic fingerprint comprises 12 feature dimensions, derived from the metabolic pathway similarity index and energy distribution parameters of intrinsic metabolic patterns. Sample phenotypic parameters include six parameters: temperature, humidity, color value, and particle size distribution. These features are organized into 18 feature nodes, each containing three attributes: feature name, feature value, and feature type. Node connections are established based on the mutual information between features. Mutual information is obtained by calculating the statistical correlation between any two features, specifically using a histogram-based statistical method to calculate the ratio of the joint probability to the marginal probability of two feature distributions. When the mutual information between two feature nodes exceeds a preset threshold of 0.15, a connection is established between the corresponding nodes, ultimately forming a dynamic feature topology graph with 18 nodes and numerous connections.

[0095] The graph convolutional processing layer aggregates the neighborhood information of each feature node layer by layer. In the first layer, each feature node collects feature information from its immediate neighbors and fuses the neighborhood features with its own features through a weighted summation, with the weights determined by the mutual information between nodes. In the second layer, each feature node can collect neighborhood information from a greater distance, extending the information transmission range to the two-hop neighborhood. Through this multi-layer processing mechanism, each feature node retains its local features while incorporating globally related features. In each layer, a combination of linear transformations and non-linear activation functions is used to update node features, ensuring that the complexity of feature transformations fully uncovers the deep relationships between features.

[0096] The weights of feature nodes in the information transmission process are dynamically adjusted based on their contribution to the quality assessment. An attention allocation mechanism is introduced during feature information transmission, which determines the weights by calculating the correlation between feature nodes and the quality assessment target. The specific calculation process includes: first, calculating the cross-correlation coefficient between each feature node and sample features of known quality levels; then, dynamically adjusting the node's influence in the information aggregation process based on the magnitude of the correlation coefficient. Feature nodes highly correlated with quality assessment are assigned larger weights, allowing them to play a more significant role in feature fusion; feature nodes with low correlation have their weights appropriately reduced. This dynamic weight adjustment mechanism adaptively highlights the role of key features, improving the targeting of feature fusion.

[0097] Global pooling is performed on feature nodes processed by multi-layer graph convolution to output a comprehensive quality index. Global pooling includes two steps: max pooling and average pooling. Max pooling selects the maximum value for each feature dimension from all feature nodes, while average pooling calculates the average value for each feature dimension across all feature nodes. The results of max pooling and average pooling are concatenated to form a 36-dimensional feature vector. This feature vector is then standardized to a value range of 0 to 1. A fully connected layer then maps the 36-dimensional feature vector to three quality index dimensions: microbial activity, nutritional components, and physical properties. These three indices together constitute the comprehensive quality index, with a weight of 0.5 for microbial activity, 0.3 for nutritional components, and 0.2 for physical properties. A weighted sum is then used to obtain the final comprehensive quality score, ranging from 0 to 100.

[0098] In S5, quality level determination results are generated based on comprehensive quality indicators. Simultaneously, the temporal evolution characteristics of the activity response distribution map are analyzed to establish dynamic quality early warning thresholds, specifically including:

[0099] In terms of quality level determination, a hierarchical decision structure based on a quality feature pyramid is first constructed. This structure consists of three levels: the bottom layer represents microbial activity characteristics, including five indicators such as metabolic rate and activity intensity; the middle layer represents nutritional component characteristics, including four indicators such as organic matter content and trace elements; and the top layer represents physical property characteristics, including three indicators such as particle size and color. The comprehensive quality indicators are decomposed according to this hierarchical structure and input into the corresponding decision nodes. Each decision node performs preliminary classification using a piecewise linear function based on the value range of the input features, outputting the quality level for that level.

[0100] Secondly, an adaptive membership function is used to quantify the input features at each decision node. The parameters of the membership function are automatically adjusted based on the feature distribution patterns of different quality levels in historical data. For microbial activity features, a Gaussian membership function is used, with its mean and variance obtained through statistical learning of the feature values ​​of qualified samples. For nutrient composition features, a trapezoidal membership function is used, with its boundary points set according to the content range specified in national standards. Through membership function calculation, the probability distribution of each feature corresponding to the three levels of superior, qualified, and unqualified products is obtained, with the sum of the probability values ​​being 1.

[0101] Then, the high-frequency fluctuation information of the bottom-level micro-features is fused and analyzed with the stable trend information of the top-level macro-features. The bottom-level micro-features include the instantaneous rate of change of activity intensity and the movement speed of metabolic hotspots; these features change at a high frequency, fluctuating 3–5 times per second. The top-level macro-features include the overall distribution trend of active regions and the stability of metabolic patterns; these features change more slowly, approximately 1–2 times per minute. Through time scale transformation, high-frequency features are processed by moving average, and low-frequency features are refined by difference, allowing the two to be fused at the same time scale. The fusion weights are dynamically adjusted according to the signal-to-noise ratio (SNR) of the features, which is determined by calculating the ratio of the variance to the mean of the feature values.

[0102] Finally, the final confidence score is calculated based on the fused features. The confidence score consists of three parts: the matching degree between each level of features and the corresponding quality level, the consistency of judgment results between different levels, and the continuity between the current judgment result and historical judgment results. The matching degree is obtained by calculating the Euclidean distance between the feature value and the grade standard value; the consistency is obtained by calculating the correlation coefficient of judgment results at different levels; and the continuity is obtained by calculating the rate of change of judgment results at adjacent time points. These three parts are weighted and summed with weights of 0.5, 0.3, and 0.2 respectively to obtain a confidence score ranging from 0 to 100. This score is compared with a preset grade threshold: scores above 85 are considered excellent, scores between 60 and 85 are considered acceptable, and scores below 60 are considered unacceptable. A confidence assessment report is also output.

[0103] For dynamic quality early warning, spatial distribution characteristic parameters of the activity response distribution map are first extracted during continuous detection. The activity intensity gradient is obtained by calculating the rate of change of activity intensity per unit distance, with a sampling point spacing of 2 mm. The area of ​​metabolic hotspot regions is obtained by statistically analyzing the proportion of regions with activity intensity exceeding 1.5 times the overall average. The spatial heterogeneity index is obtained by calculating the coefficient of variation of activity intensity at each sampling point, i.e., the ratio of standard deviation to mean. These parameters are collected every 30 seconds to form time series data.

[0104] Next, the differential characteristics of the spatial distribution feature parameters over time are calculated. First-order differential calculations are performed on each feature parameter sequence to obtain the rate of change sequence; second-order differential calculations are performed to obtain the acceleration sequence. By setting thresholds, abnormal upward or downward trends in the rate of change sequence and abrupt changes in the acceleration sequence are identified. Abnormal fluctuation patterns are identified by calculating the consistency of the change direction of five consecutive sampling points; when all five points show an upward or downward trend, abnormal fluctuations are determined to exist.

[0105] A dynamic early warning threshold curve is constructed based on the acceleration characteristics of trend changes and the statistical characteristics of abnormal fluctuations. The trend acceleration characteristic is obtained by calculating the change in the rate of change per unit time; a trend change is defined as when the acceleration value exceeds twice the standard deviation of the historical average for three consecutive sampling points. The statistical characteristics of abnormal fluctuations include fluctuation amplitude, duration, and frequency, obtained by statistically analyzing the occurrence of abnormal fluctuations over the past hour. The early warning threshold curve consists of a base threshold and an adjustment coefficient. The base threshold is determined based on historical normal data statistics, while the adjustment coefficient is dynamically adjusted based on the real-time monitored trend acceleration and abnormal fluctuation characteristics.

[0106] Finally, by comparing and analyzing historical early warning data with current monitoring data, the morphological parameters of the early warning threshold curve are optimized. An early warning case library is established, recording the characteristic parameter combinations triggered each time an early warning is issued and the subsequent quality changes. When new monitoring data arrives, it is matched with historical data in the case library to find the 10 most similar cases. The sensitivity of the current early warning threshold is adjusted based on the subsequent development of these cases. When the activity evolution characteristics exceed the optimized early warning threshold, different levels of early warning signals are triggered according to the degree of exceedance: Level 1 for mild anomalies, Level 2 for moderate anomalies, and Level 3 for severe anomalies. Simultaneously, anomaly feature maps are recorded, including the time, location, duration, and evolution trend of the anomaly, providing a basis for quality traceability.

[0107] Please see Figure 2 As shown, a smart quality detection system for micro-mineral bio-organic fertilizer includes:

[0108] The dynamic excitation and multi-source signal acquisition module applies energy excitation in a preset mode and simultaneously acquires multi-source dynamic response signals of fertilizer samples during the excitation period, constructing a time-series feature set including gas release kinetic curves and spectral change trajectories.

[0109] The transient feature extraction and visualization module is used to perform differential transformation and mode decomposition on the time-series feature set, extract transient response feature vectors that characterize the intensity of biological metabolism, and generate corresponding activity response distribution maps.

[0110] The metabolic fingerprint decoupling and analysis module establishes a component-activity correlation model based on the activity response distribution map. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained.

[0111] The multimodal feature fusion and quality index generation module constructs a multidimensional feature space from dynamic metabolic fingerprints and sample phenotypic parameters, and uses a graph convolutional network to aggregate information and reconstruct features from feature nodes to generate a comprehensive quality index.

[0112] The intelligent decision-making and dynamic early warning module generates quality level judgment results based on comprehensive quality indicators, and analyzes the temporal evolution characteristics of the active response distribution map to establish a dynamic quality early warning threshold.

[0113] The working principle of this invention is as follows: By applying a preset pattern of micro-area thermal pulse excitation and simultaneously acquiring the gas release kinetic curve and spectral change trajectory of fertilizer samples, a gas-spectrum coupled data cube is constructed. Then, differential transformation and mode decomposition are performed on the time-series feature set to extract transient response feature vectors and generate an activity response distribution map. Next, component-activity correlation analysis is established, and biological metabolic activity is decoupled from matrix background interference through feature separation and contrastive learning strategies to obtain a dynamic metabolic fingerprint. Furthermore, a multi-dimensional feature space is constructed by combining the dynamic metabolic fingerprint with apparent characteristic parameters, and a graph convolutional network is used for feature fusion to generate a comprehensive quality index. Finally, a quality level is determined based on a multi-level feature pyramid structure, and a dynamic early warning mechanism is established by analyzing the temporal evolution characteristics of the activity response distribution map to achieve early detection and warning of quality anomalies.

[0114] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the claims of this invention should still fall within the patent coverage of this invention.

Claims

1. A method for intelligent quality detection of micro-mineral bio-organic fertilizer, characterized in that, Includes the following steps: S1: By applying energy excitation in a preset mode and simultaneously acquiring multi-source dynamic response signals of fertilizer samples during the excitation period, a time-series feature set containing gas release kinetic curves and spectral change trajectories is constructed. S2: Perform differential transformation and mode decomposition on the time-series feature set to extract transient response feature vectors characterizing the intensity of biological metabolism and generate corresponding activity response distribution maps; S3: Establish a component-activity correlation model based on the activity response distribution map. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained. The establishment of the component-activity correlation model based on the activity response distribution map specifically includes: Using the spatial gradient features and intensity distribution features in the activity response distribution map as input, a dual-channel feature matrix containing matrix background features and metabolic activity features is constructed. An adversarial Siamese network model is established. The generator network learns to separate matrix background features from the dual-channel feature matrix, while the discriminator network is used to identify the integrity of metabolic activity features. The separated metabolic activity features are input into the residual mapping network to establish a nonlinear mapping relationship between metabolic activity intensity and microbial quantity, and output a pure activity feature vector that removes matrix interference. The residual mapping network is optimized by a contrastive learning strategy to ensure that the output features of the same activity level remain consistent under different matrix backgrounds. S4: Construct a multi-dimensional feature space by combining dynamic metabolic fingerprints and sample apparent characteristic parameters, and use graph convolutional networks to aggregate information and reconstruct features of feature nodes to generate comprehensive quality indicators. S5: Generate quality level judgment results based on comprehensive quality indicators, and analyze the temporal evolution characteristics of the activity response distribution map to establish a dynamic quality early warning threshold.

2. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The construction of the time-series feature set, which includes gas release kinetic curves and spectral change trajectories, specifically includes: A micro-area thermal pulse excitation device was used to apply local thermal excitation to the sample surface with a duration of milliseconds and a temperature change range of three to five degrees Celsius. Hyperspectral image sequences of specific areas of the sample surface were acquired before and after the excitation. By arranging an array of gas sensors around the sample space, the instantaneous changes in the concentration of various characteristic gases are monitored synchronously at a millisecond sampling frequency, and the gas release rate curves and their spectral characteristics are recorded. The hyperspectral image sequence was time-stamped with the gas release rate curve, and the gradient of spectral reflectance change at each sampling time was extracted to construct a gas-spectral coupled data cube containing spatiotemporal features. The gas-spectral coupled data cube is decomposed into multi-scale components to extract fast response components and slow-varying background components, and the fast response components are reconstructed into a time-series feature set reflecting the dynamics of microbial metabolism.

3. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The step of performing differential transformation and mode decomposition on the time-series feature set to extract transient response feature vectors characterizing the intensity of biological metabolism specifically includes: Higher-order differential processing is performed on gas release kinetic curves with time-series feature sets to calculate their first-order and second-order differential sequences, thereby capturing abrupt changes and trends in gas release rates. Adaptive intrinsic mode decomposition is performed on the spectral change trajectory to decompose the non-stationary signal into multiple intrinsic mode components at different time scales, and feature modes that match the microbial metabolic cycle are screened. Extract the amplitude features of abrupt change points and the energy distribution parameters of characteristic modes, and construct a multi-dimensional transient response feature vector containing time-domain abrupt change features and frequency-domain energy features; Dimensionality reduction is performed on the multidimensional transient response feature vector to obtain the core feature subset characterizing the intensity of biological metabolism.

4. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The generation of the corresponding activity response distribution map specifically includes: The transient response feature vectors are reorganized according to the sampling spatial location to construct a feature intensity matrix based on a spatial grid; The radial basis function interpolation method is used to spatially interpolate the characteristic intensity matrix to generate a continuously distributed contour map of activity intensity. Based on the gradient change characteristics of each region in the activity intensity contour map, response regions of different activity levels are divided, and metabolic hotspot regions are marked. By superimposing and fusing the spatial distribution of gas release characteristics and spectral change characteristics, an activity response distribution map that comprehensively reflects the spatial distribution and metabolic intensity of microorganisms is generated.

5. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The obtained dynamic metabolic fingerprint, which reflects the true activity of microorganisms, specifically includes: The pure activity feature vectors are recombined according to the time series to construct a dynamic metabolic trajectory matrix; Singular value decomposition is performed on the dynamic metabolic trajectory matrix to extract intrinsic metabolic patterns that characterize the main features of the metabolic process. The intrinsic metabolic pattern is dynamically time-warped and matched with the standard metabolic template to calculate the metabolic pathway similarity index. Based on the metabolic pathway similarity index and the energy distribution of intrinsic metabolic patterns, a dynamic metabolic fingerprint with unique identifiers is generated.

6. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The generation of comprehensive quality indicators specifically includes: The various feature dimensions of the dynamic metabolic fingerprint and the apparent characteristic parameters of the sample are used to construct feature nodes, and node connection edges are established based on the mutual information between features to form a dynamic feature topology graph. By aggregating the neighborhood information of each feature node layer by layer through graph convolutional layers, global related features are fused while local features are preserved. The weights in the information transmission process are dynamically adjusted based on the contribution of feature nodes to the quality assessment. Global pooling is performed on feature nodes that have undergone multi-layer graph convolution processing, and the output is a comprehensive quality index that includes local features and global correlations.

7. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The generation of quality grade determination results based on comprehensive quality indicators specifically includes: Construct a hierarchical decision tree based on the quality feature pyramid, and input the comprehensive quality indicators into the corresponding decision nodes according to the hierarchical structure of the quality feature pyramid; At each decision node, an adaptive membership function is used to perform fuzzy quantization on the input features, and the probability distribution of each quality level is calculated. The high-frequency fluctuations of the underlying micro-characteristics are integrated with the stable trends of the top-level macro-characteristics for analysis. Based on the comparison between the final confidence score and the preset level threshold, the quality level determination result with confidence assessment is output.

8. The intelligent quality detection method for micro-mineral bio-organic fertilizer according to claim 1, characterized in that, The analysis of the temporal evolution characteristics of the active response distribution map and the establishment of a dynamic quality early warning threshold specifically include: Spatial distribution characteristic parameters of the activity response distribution map during continuous detection were extracted, including activity intensity gradient, area of ​​metabolic hotspot region and spatial heterogeneity index; Calculate the differential characteristics of spatial distribution feature parameters over time to identify trend changes and abnormal fluctuation patterns in the activity evolution process; Based on the acceleration characteristics of trend changes and the statistical characteristics of abnormal fluctuations, a dynamic early warning threshold curve is constructed. By comparing and analyzing historical early warning data with current monitoring data, the morphological parameters of the early warning threshold curve are optimized. When the activity evolution characteristics exceed the early warning threshold, a graded early warning signal is triggered and the abnormal feature map is recorded.

9. A quality intelligent detection system for micro-mineral bio-organic fertilizer, characterized in that, A method for intelligent quality detection of a micro-mineral bio-organic fertilizer according to any one of claims 1-8 includes: The dynamic excitation and multi-source signal acquisition module applies energy excitation in a preset mode and simultaneously acquires multi-source dynamic response signals of fertilizer samples during the excitation period, constructing a time-series feature set including gas release kinetic curves and spectral change trajectories. The transient feature extraction and visualization module is used to perform differential transformation and mode decomposition on the time-series feature set, extract transient response feature vectors that characterize the intensity of biological metabolism, and generate corresponding activity response distribution maps. The metabolic fingerprint decoupling and analysis module establishes a component-activity correlation model based on the activity response distribution map. By decoupling biological metabolic activity from matrix background interference, a dynamic metabolic fingerprint reflecting the true activity of microorganisms is obtained. The multimodal feature fusion and quality index generation module constructs a multidimensional feature space from dynamic metabolic fingerprints and sample phenotypic parameters, and uses a graph convolutional network to aggregate information and reconstruct features from feature nodes to generate a comprehensive quality index. The intelligent decision-making and dynamic early warning module generates quality level judgment results based on comprehensive quality indicators, and analyzes the temporal evolution characteristics of the active response distribution map to establish a dynamic quality early warning threshold.

Images