Multi-valley physical index ai intelligent accurate judgment analysis method and system

Abstract

The application discloses an AI intelligent accurate judgment analysis method and system for multiple valley physical and chemical indexes, and comprises the following steps: based on the intelligent control unit, multi-source heterogeneous data collection is cooperatively triggered to obtain an associated data set carrying a unified timestamp; the associated data set is subjected to multi-modal adaptive preprocessing to obtain a standardized feature set; based on the standardized feature set, a multi-crop adaptive AI detection model group is called to obtain single-item physical and chemical index detection results; the single-item physical and chemical index detection results are subjected to multi-dimensional fusion judgment to obtain a grain comprehensive quality grade and settlement parameters; the grain comprehensive quality grade and settlement parameters are uploaded to a grain quality inspection platform, and based on artificial review data, closed-loop self-optimization of the multi-crop adaptive AI detection model group is realized.

Description

Technical Field

[0001] This invention discloses an AI-powered intelligent and precise analysis method, device, and medium for multi-valley physical and chemical indicators, belonging to the field of multi-valley physical and chemical analysis technology. Background Technology

[0002] Currently, the testing of physicochemical indicators of grain mainly adopts a combination of manual sensory inspection and laboratory physicochemical analysis. Although some enterprises have introduced single machine vision or near-infrared spectroscopy detection equipment, there are still many insurmountable shortcomings: manual inspection is highly subjective and inefficient, and multi-indicator testing of a single batch takes 30-60 minutes, which is difficult to meet the needs of rapid collection and inspection during the peak grain purchasing season; a single detection device can only obtain information on the appearance or internal composition of grain, and cannot achieve complementary verification of multi-source data, resulting in a high rate of missed detection of abnormal grain conditions such as hidden mold and internal insect infestation; data from weighing, sampling, and testing are collected independently, lacking a unified timestamp and business association, making it impossible to achieve accurate traceability of one file per vehicle and posing a risk of data tampering.

[0003] Existing AI detection models are mostly trained for single grain types or single indicators, lacking the ability to adapt to multiple grain types, and there is no online model update mechanism. As a result, the detection accuracy gradually decreases as the quality of grain changes. At the same time, the detection system is isolated from business systems such as weighing, settlement, and warehousing. The detection results need to be entered manually, which not only increases labor costs but also creates opportunities for human fraud such as favoritism and cronyism, seriously affecting the digital transformation and quality and safety management of the grain industry.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides an AI-powered intelligent and precise analysis method and system for multiple grain physicochemical indicators, which at least solves the technical problem that existing grain physicochemical indicator detection methods cannot simultaneously achieve accurate and efficient detection of multiple grain types and multiple indicators, full-process data traceability, and continuous self-optimization of the model.

[0006] According to one aspect of the present invention, in order to achieve the above-mentioned objective, an AI-powered intelligent and precise analysis method for multi-valley physical and chemical indicators is provided, comprising the following steps:

[0007] Based on the collaborative triggering of multi-source heterogeneous data acquisition by intelligent control unit, a related dataset with a unified timestamp is obtained;

[0008] Multimodal adaptive preprocessing is performed on the associated dataset to obtain a standardized feature set;

[0009] Based on the standardized feature set, the multi-grain adaptive AI detection model group is called to obtain the detection results of each individual physicochemical index.

[0010] By integrating and analyzing the test results of each individual physicochemical indicator from multiple dimensions, the comprehensive quality grade of grain and settlement parameters are obtained.

[0011] The overall quality grade of the grain and the settlement parameters are uploaded to the grain quality inspection platform.

[0012] Furthermore, the steps for collaboratively triggering multi-source heterogeneous data acquisition based on the intelligent control unit specifically include:

[0013] In response to the vehicle entry signal, the unattended weighing system collects vehicle identification information and weighing data to generate a unique business serial number.

[0014] Based on a unique business serial number, the intelligent sampling unit and the multispectral detection unit are triggered to simultaneously acquire sample image data, sample spectral data, and detection environment data;

[0015] By associating vehicle identification information, weighing data, sample image data, sample spectral data, and testing environment data with a unique business serial number and a unified timestamp, an associated dataset is obtained.

[0016] Furthermore, the steps for performing multimodal adaptive preprocessing on the associated dataset specifically include:

[0017] Adaptive morphological segmentation and noise filtering were applied to the sample image data in the associated dataset to extract grain visual features.

[0018] Multivariate scattering correction and characteristic wavelength screening were applied to the sample spectral data in the associated dataset to extract spectral features;

[0019] Normalize the weighing data and detection environment data in the associated dataset and extract statistical features;

[0020] By fusing visual, spectral, and statistical features of grains, a standardized feature set is obtained.

[0021] Furthermore, the steps for invoking the multi-grain adaptive AI detection model group based on the standardized feature set include:

[0022] Automatic grain type identification is performed based on standardized feature sets to obtain grain type identification results;

[0023] The corresponding appearance index detection sub-model and physicochemical index prediction sub-model are automatically matched based on the grain type identification results.

[0024] The standardized feature set is input into the matched appearance index detection sub-model and physicochemical index prediction sub-model to obtain the detection results of each individual physicochemical index.

[0025] Furthermore, the appearance index detection sub-model adopts a combination of the improved YOLOv8 target detection algorithm and the U-Net image segmentation algorithm to identify and count imperfect particles, moldy particles, insect-eaten particles and damaged particles, and calculate the area ratio of abnormal regions.

[0026] Furthermore, the physicochemical index prediction sub-model adopts a fusion algorithm of support vector regression and deep neural network. The input is spectral features and detection environment data, and the output is the predicted values ​​of moisture, protein, fat and starch content.

[0027] Furthermore, the specific steps for multi-dimensional integration and analysis of the test results of each individual physicochemical indicator include:

[0028] Data-level weighted fusion is performed on the multi-source detection results of the same indicator to obtain the corrected single indicator result;

[0029] The corrected individual indicator results are fused at the feature level to obtain a comprehensive quality feature vector.

[0030] Based on the comprehensive quality feature vector and the preset quality evaluation standard, a decision-level fusion analysis is performed to obtain the comprehensive quality grade of grain and settlement parameters.

[0031] Furthermore, it also includes:

[0032] Regularly collect manually reviewed data and laboratory test data to build an incrementally updated dataset;

[0033] Based on the incrementally updated dataset, the multi-grain adaptive AI detection model group is updated online using an incremental learning algorithm;

[0034] Evaluate the detection accuracy of the updated model, and activate the updated model when the accuracy improvement exceeds a preset threshold.

[0035] According to one embodiment of the present invention, an AI-powered intelligent and precise analysis system for multiple valley physical and chemical indicators is also provided, comprising:

[0036] The acquisition unit is used to collaboratively trigger the acquisition of multi-source heterogeneous data based on the intelligent control unit to obtain a related dataset with a unified timestamp;

[0037] The preprocessing unit is used to perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set;

[0038] The calling unit is used to call the multi-grain adaptive AI detection model group based on the standardized feature set to obtain the detection results of each individual physicochemical index.

[0039] The fusion unit is used to perform multi-dimensional fusion analysis on the test results of each individual physicochemical indicator to obtain the comprehensive quality grade of grain and settlement parameters.

[0040] The upload unit is used to upload the comprehensive quality grade of grain and settlement parameters to the grain quality inspection platform.

[0041] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of the present invention.

[0042] In this embodiment of the invention, by using an intelligent control unit to collaboratively trigger the acquisition of multi-source heterogeneous data and generate a correlated dataset with a unified timestamp, synchronous acquisition and precise correlation of multi-source data such as weighing, images, spectra, and environmental data are achieved, effectively preventing data tampering and ensuring data authenticity and traceability. By performing multimodal adaptive preprocessing on the correlated dataset and obtaining a standardized feature set, noise interference and dimensional differences between different types of data are eliminated, improving the accuracy and stability of subsequent AI detection. By calling a multi-grain adaptive AI detection model group to obtain the detection results of each individual physicochemical index, automatic detection of multiple indicators for various major grains is achieved, solving the problem of poor adaptability of a single model. Through the analysis of the detection results of each individual physicochemical index... By conducting multi-dimensional fusion analysis to obtain the comprehensive quality grade and settlement parameters of grain, complementary verification of appearance information and internal component information was achieved, significantly reducing the missed detection rate of abnormal grain conditions. By uploading the comprehensive quality grade and settlement parameters of grain to the grain quality inspection platform and realizing the closed-loop self-optimization of the multi-grain adaptive AI detection model group, not only was centralized management and full-process traceability of quality inspection data achieved, but the model was also able to continuously adapt to changes in grain quality and maintain high detection accuracy over a long period of time. At the same time, it was deeply integrated with unattended weighing systems and intelligent control systems, realizing the unmanned and automated operation of the entire grain procurement process, greatly improving detection efficiency, eliminating human fraud, and providing strong support for the digital transformation and quality and safety assurance of the grain industry. Attached Figure Description

[0043] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0044] Figure 1 This is a flowchart of an AI-powered intelligent and precise judgment and analysis method for multi-valley physical and chemical indicators according to one embodiment of the present invention.

[0045] Figure 2 This is a structural block diagram of an AI-powered intelligent and precise judgment and analysis system for multi-valley physical and chemical indicators according to one embodiment of the present invention. Detailed Implementation

[0046] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. 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 should fall within the scope of protection of the present invention.

[0047] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0048] According to an embodiment of the present invention, an embodiment of an AI-powered intelligent and precise judgment and analysis method for multi-valley physical indicators is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0049] This method embodiment can be executed in an electronic device or similar computing device that includes a memory and a processor. Taking operation on a vehicle terminal as an example, the vehicle terminal may include one or more processors (processors may include, but are not limited to, central processing units (CPUs), graphics processing units (GPUs), digital signal processing (DSP) chips, microcontroller units (MCUs), field-programmable gate arrays (FPGAs), neural network processors (NPUs), tensor processors (TPUs), artificial intelligence (AI) type processors, etc.) and a memory for storing data. Optionally, the vehicle terminal may also include transmission devices, input / output devices, and display devices for communication functions. Those skilled in the art will understand that the above structural description is merely illustrative and does not limit the structure of the vehicle terminal. For example, the vehicle terminal may include more or fewer components than described above, or have a different configuration than described above.

[0050] The memory can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the AI-powered intelligent and precise analysis method for multi-valley physical indicators in this embodiment of the invention. The processor executes various functional applications and data processing by running the computer program stored in the memory, thereby realizing the aforementioned AI-powered intelligent and precise analysis method for multi-valley physical indicators. The memory may include high-speed random access memory and non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the mobile terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0051] The transmission device is used to receive or send data via a network. Specific examples of the network mentioned above may include a wireless network provided by the mobile terminal's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0052] Display devices can be, for example, touchscreen liquid crystal displays (LCDs) and touch displays (also referred to as "touchscreens" or "touch displays"). The LCD allows users to interact with the user interface of the mobile terminal. In some embodiments, the mobile terminal has a graphical user interface (GUI), which allows users to interact with the GUI through finger contact and / or gestures on a touch-sensitive surface. Optional human-computer interaction functions include: creating web pages, drawing, word processing, creating electronic documents, playing games, video conferencing, instant messaging, sending and receiving emails, call interfaces, playing digital video, playing digital music, and / or web browsing, etc. Executable instructions for performing the above human-computer interaction functions are configured / stored in one or more processor-executable computer program products or readable storage media.

[0053] Figure 1 This invention relates to an AI-powered intelligent and precise analysis method for multi-valley physical and chemical indicators, comprising:

[0054] Step S110: Based on the intelligent control unit's collaborative triggering of multi-source heterogeneous data acquisition, a related dataset carrying a unified timestamp is obtained, the specific content of which is as follows:

[0055] In step S110, when a vehicle enters the grain depot, the unmanned weighing system deployed at the entrance first triggers the identity verification process. The unmanned weighing system collects vehicle license plate information via a high-definition license plate recognition device, reads basic information such as vehicle type and rated load capacity from the vehicle's built-in electronic tag via an RFID reader, and simultaneously obtains business information such as grain type, origin, and estimated weight declared by the cargo owner through an on-site touch terminal or remote reservation system. The unmanned weighing system integrates and compares the collected information, generates a globally unique business transaction number after successful verification, and transmits the encrypted business transaction number to the intelligent control unit. This business transaction number serves as the unique identifier for this quality inspection transaction and will be used throughout the entire process from vehicle entry to settlement and delivery, achieving unified association of all business data.

[0056] Upon receiving the business serial number, the intelligent control unit immediately sends a synchronization trigger command to each data acquisition device, initiating a parallel acquisition process of multi-source heterogeneous data. Based on the grain type declaration information uploaded by the unattended weighing system, the intelligent control unit automatically invokes a preset equipment scheduling strategy, controlling the intelligent sampling unit to perform automatic sampling operations according to the national standard sampling specifications for the corresponding grain type. The collected grain samples are then evenly transported to the multispectral imaging unit and near-infrared spectral detection unit via pneumatic conveying pipes. Simultaneously, the multispectral imaging unit activates a high-resolution industrial camera and multi-band light source to continuously acquire images of the grain samples transported to the testing station; the near-infrared spectral detection unit initiates a spectral scanning program to acquire full-band spectral data for the same batch of samples; and the environmental monitoring unit deployed at the testing site simultaneously collects environmental parameters such as temperature, humidity, and air pressure.

[0057] All data acquisition devices are synchronized in real time with the master clock of the intelligent control unit, ensuring that the data collected by each device has a unified time reference. The intelligent control unit assigns a high-precision timestamp synchronized with the master clock to each acquired data segment. The timestamp is accurate to the millisecond level, accurately reflecting the order and time interval of each data acquisition. The intelligent control unit binds the acquired vehicle identification information, weighing data, sample image data, sample spectral data, and testing environment data to the business serial number and the corresponding timestamp, forming a one-to-one mapping relationship to ensure that each set of data can uniquely correspond to a specific vehicle and a specific testing batch.

[0058] After data collection and timestamp binding are completed, the intelligent control unit performs integrity and consistency verification on the generated preliminary associated dataset. Integrity verification checks for missing or failed data collection at each stage. If missing data is detected at any stage, the intelligent control unit automatically triggers a re-collection process on the corresponding device. If re-collection fails, the batch of data is marked as abnormal and pushed to administrators for manual processing. Consistency verification cross-references multi-source data. By analyzing the inherent logical relationships between different data sources, it identifies potential data tampering or abnormal data. For verified data, blockchain encryption technology is used for evidence storage to ensure that the data cannot be tampered with once generated.

[0059] After the aforementioned integrity and consistency verifications, a compliant associated dataset with a unified timestamp is finally generated. This associated dataset is transmitted via industrial Ethernet to the subsequent data preprocessing module, providing input for adaptive preprocessing of multimodal data and standardized feature extraction. This step, through the coordinated scheduling of the intelligent control unit, achieves synchronous acquisition, accurate association, and secure storage of multi-source heterogeneous data, ensuring the accuracy and reliability of subsequent AI-powered intelligent judgment results from the source, while laying a solid data foundation for the full-process traceability management of grain quality.

[0060] Step S120: Perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set, the details of which are as follows:

[0061] In step S120, the multimodal data processing engine first performs modal splitting on the associated dataset, dividing it into three independent but interconnected data subsets: a sample image data subset, a sample spectral data subset, and a structured data subset. Based on the grain type identification information carried in the associated dataset, the multimodal data processing engine automatically loads the preprocessing parameter configuration file corresponding to the grain type, achieving differentiated adaptive processing of different grain type data and avoiding the problem of poor adaptability to different grain type data in traditional fixed-parameter preprocessing methods.

[0062] For a subset of sample image data, the multimodal data processing engine first performs an image quality assessment. By calculating indicators such as image sharpness, contrast, and brightness, it automatically filters out valid images that meet quality requirements, discarding blurry, overexposed, underexposed, or invalid images with severe background interference. For the filtered valid images, the multimodal data processing engine uses an adaptive median filtering algorithm to remove salt-and-pepper noise and Gaussian noise generated during image acquisition, effectively suppressing noise while preserving the edge details of the grains to the greatest extent. Subsequently, the multimodal data processing engine automatically selects the corresponding image segmentation algorithm based on the grain type: for round grains such as corn and soybeans, an improved circular Hough transform algorithm is used for grain instance segmentation; for long grains such as rice and wheat, a combined segmentation algorithm based on edge detection and morphological opening and closing operations is used to achieve accurate separation of adhering grains. After segmentation, the multimodal data processing engine performs mask extraction on each independent grain region to eliminate the interference of background regions on subsequent feature extraction.

[0063] After image segmentation and mask extraction, the multimodal data processing engine extracts multi-dimensional visual features from each grain region. These visual features fall into three main categories: color features, texture features, and shape features. Color features include the mean, variance, skewness, and kurtosis of the RGB three channels, as well as histogram features in the HSV color space. Texture features include the contrast, correlation, energy, entropy, and local binary mode features of the gray-level co-occurrence matrix. Shape features include the grain's area, perimeter, aspect ratio, roundness, rectangularity, and Hu invariant moments. The multimodal data processing engine normalizes all extracted visual features, mapping feature values ​​of different dimensions to the [0,1] interval, thus eliminating the impact of dimensional differences on subsequent model training and prediction.

[0064] For a subset of sample spectral data, the multimodal data processing engine first performs spectral data preprocessing to eliminate various interference factors generated during spectral acquisition. The engine employs a multivariate scattering correction algorithm to eliminate the influence of grain particle size, uneven distribution, and light scattering effects on the spectral data. It uses a standard normal variable transformation algorithm to eliminate spectral shifts caused by differences in sample packing density. First- and second-derivative algorithms are used to remove spectral baseline drift and background interference, highlighting characteristic absorption peaks in the spectrum. After preprocessing, the engine uses a feature wavelength selection method combining principal component analysis and continuous projection algorithms to select the feature wavelengths with the highest correlation to the physicochemical indicators of grain from the full-band spectral data. This significantly reduces the feature dimensionality while retaining effective spectral information, minimizing redundant information interference with subsequent models and improving the model's computational speed and prediction accuracy.

[0065] For the structured data subset, the multimodal data processing engine first performs data cleaning on the weighing data and the detection environment data, using the 3σ principle to identify and remove outliers, and using linear interpolation to complete missing data. After data cleaning, the multimodal data processing engine performs min-max normalization on the weighing data and the detection environment data, mapping them to the [0,1] interval. Simultaneously, the multimodal data processing engine extracts statistical features from the structured data subset, including the average weight of a single batch of samples, the standard deviation of weight, the average temperature of the detection environment, and the average humidity, providing auxiliary information for subsequent comprehensive quality assessment.

[0066] After completing independent preprocessing and feature extraction for each modality of data, the multimodal data processing engine performs a multimodal feature fusion operation. The engine concatenates the extracted visual, spectral, and statistical features in a preset order to form a unified high-dimensional feature vector. To further optimize the quality of the feature vector, the engine employs a feature selection algorithm to perform a secondary screening of the concatenated high-dimensional feature vector, removing redundant features with low relevance and retaining the core features that contribute most to the detection of grain physicochemical indicators. Finally, through the above multimodal adaptive preprocessing and feature fusion operations, a standardized feature set with uniform format, reasonable dimensions, and rich information is generated.

[0067] The standardized feature set is transmitted to the subsequent AI detection module via the internal data bus, serving as input data for the multi-grain adaptive AI detection model group. This step employs differentiated adaptive preprocessing methods tailored to the characteristics of different modalities and effectively fuses multi-modal features, effectively eliminating various noise interferences and data heterogeneity issues. This significantly improves the quality and representativeness of the feature data, laying a solid foundation for subsequent AI-powered intelligent and accurate detection of multiple grain types and indicators.

[0068] Step S130: Based on the standardized feature set, the multi-grain adaptive AI detection model group is invoked to obtain the detection results of each individual physicochemical index, as detailed below:

[0069] In step S130, the multi-model fusion inference engine first performs automatic grain type identification. Based on the visual and spectral features fused from the standardized feature set, it accurately classifies the grain types to be detected. This embodiment pre-constructs a multi-dimensional feature database covering major grain types such as corn, rice, soybeans, and wheat in the main producing areas of Northeast China. This database contains hundreds of thousands of grain type samples from different origins, years, grades, and storage periods, covering the characteristic variation patterns of each grain type under different growth environments and storage conditions. The multi-model fusion inference engine uses a support vector machine classifier. The standardized feature set is input into the support vector machine classifier for calculation to obtain the grain type identification result. The grain type identification result includes the grain type name, variety type, and confidence level information. When the identification confidence level is higher than a preset threshold, the system automatically confirms the grain type; when the identification confidence level is lower than the preset threshold, the system triggers a manual review process to ensure the accuracy of grain type identification.

[0070] Based on the grain type identification results, the multi-model fusion inference engine automatically matches the corresponding dedicated detection sub-model from the multi-grain adaptive AI detection model group. The multi-grain adaptive AI detection model group adopts a modular architecture design, consisting of two parts: an appearance index detection sub-model library and a physicochemical index prediction sub-model library. Each sub-model library contains dedicated sub-models trained separately for different grain types. The appearance index detection sub-model library stores appearance detection sub-models for corn, rice, soybeans, and wheat, etc.; the physicochemical index prediction sub-model library stores moisture prediction sub-models for corn, protein prediction sub-models for corn, amylose prediction sub-models for rice, and fat prediction sub-models for soybeans, etc. This modular model architecture design allows the system to flexibly call upon the corresponding sub-models according to the detection needs of different grain types, avoiding the problem of decreased accuracy of a single model in multi-grain detection, and also facilitating independent updates and maintenance of subsequent models.

[0071] After matching, the multi-model fusion inference engine inputs the standardized feature set into the appearance index detection sub-model and the physicochemical index prediction sub-model in parallel, achieving simultaneous detection of appearance and internal physicochemical indicators. The appearance index detection sub-model adopts a technical solution combining an improved YOLOv8 object detection algorithm and a U-Net image segmentation algorithm. The improved YOLOv8 object detection algorithm is used to quickly locate and classify various abnormal grains such as imperfect grains, moldy grains, insect-damaged grains, broken grains, and grains of different varieties. The U-Net image segmentation algorithm is used to perform pixel-level segmentation of the detected abnormal grains and accurately calculate the area ratio of moldy and insect-damaged regions. This embodiment specifically optimizes the YOLOv8 object detection algorithm by introducing an attention mechanism to enhance the model's ability to identify small abnormal regions and by improving the loss function to improve the model's classification accuracy for different types of abnormal grains. At the same time, the encoder and decoder structures of the U-Net image segmentation algorithm are lightweighted, improving the model's inference speed while ensuring segmentation accuracy.

[0072] The physicochemical index prediction sub-model adopts an algorithm architecture that integrates support vector regression and deep neural networks. This fully combines the advantages of support vector regression in small-sample learning and generalization capabilities with the advantages of deep neural networks in fitting complex nonlinear relationships. The input to the physicochemical index prediction sub-model consists of spectral features from a standardized feature set and detection environment data. The output is the predicted values ​​of core physicochemical indicators such as moisture, protein, fat, starch, and bulk density. This embodiment trains corresponding dedicated physicochemical index prediction sub-models for different grain varieties and production areas, and introduces an environmental parameter correction mechanism. This mechanism can automatically adjust the model's prediction parameters based on environmental data such as temperature and humidity at the detection site, eliminating the influence of environmental factors on the spectral detection results and further improving the prediction accuracy of physicochemical indicators.

[0073] After completing the detection of all individual indicators, the multi-model fusion inference engine integrates and formats the detection results output by each sub-model. The engine summarizes the number, proportion, and area percentage of abnormal grains obtained from appearance indicator detection, as well as the predicted values ​​of moisture, protein, fat, starch, and other indicators from physicochemical indicators, generating complete detection results for each individual physicochemical indicator according to a preset format. Each individual physicochemical indicator detection result includes the detection value, detection time, and corresponding sub-model version number for each indicator, providing comprehensive and accurate basic data for subsequent multi-dimensional fusion analysis.

[0074] The test results of each individual physicochemical indicator are transmitted to the subsequent comprehensive evaluation module via an internal data bus, serving as input data for multi-dimensional fusion evaluation. This step, through the adoption of a modular model architecture that adapts to multiple grain varieties, and a technical solution that combines target detection with image segmentation and support vector regression with deep neural networks, achieves simultaneous and accurate detection of multiple indicators for various major grain varieties. This effectively solves the problems of poor adaptability and insufficient detection accuracy of traditional single models, providing reliable technical support for the comprehensive evaluation of grain quality.

[0075] Step S140 involves multi-dimensional fusion analysis of the test results of each individual physicochemical indicator to obtain the comprehensive grain quality grade and settlement parameters. The specific details are as follows:

[0076] In step S140, a data-level fusion operation is first performed to weight and correct the multi-source detection results of the same indicator, thereby further reducing detection errors and improving the accuracy of individual indicators. In actual testing, the same physicochemical indicator may yield different results through multiple detection methods; for example, moisture content can be obtained simultaneously through near-infrared spectroscopy prediction and rapid moisture meter detection. This embodiment employs a dynamic weighted fusion algorithm, automatically calculating the weight coefficients of different detection results based on the historical detection accuracy and stability data of each detection method. Specifically, the system calculates the error variance between each detection method and the laboratory standard detection results over a past period; the smaller the error variance, the higher the corresponding weight coefficient. By weighting and averaging the multi-source detection results of the same indicator, the corrected individual indicator result is obtained, effectively eliminating random and systematic errors of a single detection method.

[0077] After data-level fusion, the system performs feature-level fusion, transforming the corrected individual indicator results into a comprehensive quality feature vector that fully reflects grain quality. The comprehensive quality feature vector consists of two parts: appearance quality feature components and internal component feature components. The appearance quality feature components include indicators such as imperfect grain rate, moldy grain rate, insect-damaged grain rate, broken grain rate, and heterogeneous grain rate. The internal component feature components include indicators such as moisture, protein, fat, starch, and bulk density. This embodiment introduces an attention mechanism to assign weights to different feature components. By learning from a large amount of historical detection data and expert evaluation results, it automatically identifies key features that have a significant impact on the overall quality of grain and assigns them higher weights. For example, the moldy grain rate has a decisive impact on the edible safety and storage stability of grain, so the system automatically assigns it a higher weight; while the slight broken grain rate has a relatively small impact on grain quality, so the system assigns it a lower weight. Through feature-level fusion, complementary verification of appearance and internal component information is achieved, significantly improving the ability to identify abnormal grain conditions such as hidden mold and internal insect damage that are difficult to detect using a single detection method.

[0078] Based on the generated comprehensive quality feature vector, the system performs a decision-level fusion operation to automatically determine the comprehensive quality grade of the grain. This embodiment pre-constructs a multi-index comprehensive evaluation model that conforms to national grain quality standards. This model is based on various national grain quality grading standards and incorporates general practices in the grain industry with the specific needs of different application scenarios, establishing a feature vector threshold matrix corresponding to each quality grade. The system compares the comprehensive quality feature vector with the threshold matrix and calculates the membership degree between the batch of grain and each quality grade using a fuzzy comprehensive evaluation algorithm. Finally, the grade with the highest membership degree is determined as the comprehensive quality grade of the batch of grain. Simultaneously, the system allows users to customize evaluation standards and grade thresholds according to their actual needs, flexibly adapting to the grain quality evaluation requirements of different regions, enterprises, and uses.

[0079] After determining the overall quality grade of the grain, the system automatically calculates the corresponding settlement parameters. These parameters include the moisture deduction ratio, impurity deduction ratio, base unit price, settlement unit price, and total settlement amount. The system calculates the moisture deduction ratio according to a preset formula based on the difference between the corrected moisture index result and the standard moisture requirement; it also calculates the impurity deduction ratio according to a preset formula based on the difference between the corrected impurity index result and the standard impurity requirement. Based on this, the system automatically calculates the actual settlement unit price for the batch of grain, combining this with the previously collected vehicle weighing data, and calculates the final total settlement amount. All calculation formulas and parameters strictly adhere to relevant national regulations and industry standards to ensure transparency and fairness in the settlement process.

[0080] This step also integrates an intelligent early warning function for abnormal grain conditions. The system has pre-established an abnormal grain condition early warning threshold database, covering various key indicator thresholds affecting grain safety and quality. When the detection results of one or more indicators exceed the preset warning threshold, the system will automatically issue an audible and visual alarm signal and push the abnormal information to the management personnel's terminal. For example, when the detected grain mold rate exceeds the safety threshold, the system will immediately issue a high-level warning, prompting management personnel that the batch of grain has a quality and safety risk, prohibiting its storage or requiring special handling. At the same time, the system will statistically analyze the quality data of grain from the same production area and the same batch, identify quality change trends, and provide early warnings of potential regional quality problems.

[0081] The final generated comprehensive grain quality grade and settlement parameters will be fully recorded and associated with the corresponding business serial number, and transmitted to the subsequent closed-loop optimization module and related business systems via the internal data bus. This step, through a three-level progressive multi-dimensional integrated assessment system, achieves a leap from single-indicator testing to comprehensive evaluation of grain quality, effectively avoiding the one-sidedness and subjectivity of single-indicator evaluation, and providing a scientific, accurate, and reliable decision-making basis for quality-based pricing, graded storage, and scientific processing in the grain procurement process.

[0082] Step S150: Upload the comprehensive grain quality grade and settlement parameters to the grain quality inspection platform. The specific details are as follows:

[0083] In step S150, after generating the comprehensive grain quality grade and settlement parameters, the system first performs data encapsulation and encryption. The system integrates core data such as the comprehensive grain quality grade, test results of each individual physicochemical indicator, moisture deduction ratio, impurity deduction ratio, settlement unit price, and final settlement amount with corresponding business transaction numbers, vehicle identification information, weighing data, and collection timestamps. This data is then standardized and encapsulated according to a preset JSON format to form a unified data upload message. To ensure the security and integrity of data transmission, the system uses the national cryptographic algorithm SM2 to perform asymmetric encryption on the data message. Simultaneously, it generates a digital digest of the data message and signs it to ensure that the data is not stolen, tampered with, or forged during transmission.

[0084] After data encapsulation and encryption, the system uploads the encrypted data packets to the grain quality inspection platform via industrial Ethernet or a dedicated 5G network. The system employs a breakpoint resume mechanism; when the network is interrupted or transmission fails, it automatically saves the transmission progress and resumes transmission from the point of interruption once the network is restored, avoiding duplicate data uploads or data loss. During the data upload process, the system monitors the transmission status in real time and performs integrity verification on the uploaded data. By comparing the verification code returned by the receiving end with the original verification code from the sending end, it confirms whether the data has arrived at the platform completely and accurately. If the verification fails, the system automatically triggers a retransmission mechanism until the data upload is successful.

[0085] Upon receiving encrypted data packets, the grain quality inspection platform first performs data decryption and signature verification. The platform uses the corresponding private key to decrypt the data packets and simultaneously verifies the validity of the digital signature, confirming the legality and integrity of the data's source. After successful signature verification, the platform parses and converts the data, extracting structured and unstructured data, and stores them separately in corresponding databases. Structured data such as vehicle information, weighing data, test results, and settlement parameters are stored in a relational database for easy retrieval and statistical analysis; unstructured data such as sample images and spectral data are stored in a distributed file system, with an associated index established between the unstructured data and the structured data.

[0086] The platform deeply binds all uploaded data with a globally unique business transaction number, automatically constructing an electronic quality file for each vehicle. This electronic quality file comprehensively records information from vehicle entry, identity verification, intelligent sampling, multi-source data collection, AI intelligent testing, comprehensive quality assessment, to final settlement. Data at each stage can be traced and queried using the business transaction number. The platform supports file retrieval based on multiple dimensions, including business transaction number, license plate number, grain type, owner information, and acquisition time. Users only need to enter any search criteria to quickly retrieve the corresponding complete electronic file, achieving full traceability of grain quality throughout the entire process.

[0087] The grain quality inspection platform provides multi-terminal data display and query services, achieving transparent information sharing. At the purchasing site, the platform pushes information such as the overall grain quality grade and settlement amount to the on-site LED display screen and voice broadcast system in real time, allowing drivers to clearly understand the test results and settlement information of the current purchase without leaving their vehicles. At the grain depot management end, managers can view detailed data for all purchased batches in real time through the web management interface, including test results for each individual indicator, distribution of overall quality grades, and daily purchase volume statistics. Simultaneously, the platform also provides a mobile APP service, allowing cargo owners to check the test results and settlement progress of their sold grain anytime, anywhere via a mobile app, effectively avoiding disputes caused by information asymmetry.

[0088] The grain quality inspection platform possesses open system integration capabilities, achieving seamless integration with other business systems through standardized API interfaces. The platform can automatically synchronize the comprehensive quality grade of grain with settlement parameters to the financial settlement system, triggering an automated settlement process. This eliminates the need for manual data entry by financial personnel, significantly improving settlement efficiency and accuracy. The platform can also interface with warehouse management systems, automatically allocating corresponding storage locations based on the comprehensive quality grade of grain, enabling tiered storage and categorized management. Furthermore, the platform supports data integration with third-party systems such as grain trading platforms and regulatory information platforms, achieving cross-platform data sharing and collaborative application.

[0089] The platform has also established a comprehensive data security management system to ensure the security of data storage and access. It employs a multi-level access control mechanism, assigning different data access permissions based on user roles and responsibilities, ensuring that users can only view data relevant to their work. Simultaneously, the platform logs all data operations, recording information such as the operator, time, and content of the operation, enabling traceability of data operations. Furthermore, the platform utilizes data backup and disaster recovery mechanisms, regularly performing full and incremental backups and storing backup data in an off-site disaster recovery center to ensure rapid data recovery in the event of natural disasters or system failures.

[0090] This step, through a standardized and secure data upload process and centralized and intelligent platform data management, achieves unified aggregation and efficient utilization of grain quality inspection data, breaks down data silos between various business systems, and realizes automatic data flow and business collaboration throughout the entire process. This not only significantly improves the efficiency and transparency of grain procurement, but also provides a rich, accurate, and reliable data foundation for grain quality and safety supervision, grain quality analysis, and market forecasting.

[0091] Based on steps S110 to S150 above, in this embodiment of the invention, by using an intelligent control unit to collaboratively trigger the acquisition of multi-source heterogeneous data and generate a correlated dataset with a unified timestamp, synchronous acquisition and precise correlation of multi-source data such as weighing, images, spectra, and environment are achieved, effectively preventing data tampering and ensuring the authenticity and traceability of the data. By performing multimodal adaptive preprocessing on the correlated dataset and obtaining a standardized feature set, noise interference and dimensional differences between different types of data are eliminated, improving the accuracy and stability of subsequent AI detection. By calling the multi-grain adaptive AI detection model group to obtain the detection results of each individual physicochemical index, automatic detection of multiple indicators for various major grains is achieved, solving the problem of poor adaptability of a single model. The results of single physicochemical index tests are integrated and analyzed from multiple dimensions to obtain the comprehensive quality grade and settlement parameters of grain. This achieves complementary verification of appearance information and internal composition information, significantly reducing the missed detection rate of abnormal grain conditions. By uploading the comprehensive quality grade and settlement parameters of grain to the grain quality inspection platform and realizing the closed-loop self-optimization of the multi-grain adaptive AI detection model group, not only is centralized management and full-process traceability of quality inspection data achieved, but the model can also continuously adapt to changes in grain quality and maintain high detection accuracy over a long period of time. At the same time, it is deeply integrated with unattended weighing systems and intelligent control systems, realizing the unmanned and automated operation of the entire grain procurement process, greatly improving detection efficiency, eliminating human fraud, and providing strong support for the digital transformation and quality and safety assurance of the grain industry.

[0092] The method of the embodiments of the present invention further includes the step of triggering multi-source heterogeneous data acquisition based on the intelligent control unit: in response to the vehicle entry signal, the vehicle identification information and weighing data are collected through the unattended weighing system to generate a unique business serial number; based on the unique business serial number, the intelligent sampling unit and the multispectral detection unit are triggered to simultaneously collect sample image data, sample spectral data and detection environment data; the vehicle identification information, weighing data, sample image data, sample spectral data and detection environment data are associated with the unique business serial number and a unified timestamp to obtain an associated dataset.

[0093] Furthermore, the steps for multimodal adaptive preprocessing of the associated dataset specifically include: applying adaptive morphological segmentation and noise filtering to the sample image data in the associated dataset to extract grain visual features; applying multivariate scattering correction and feature wavelength screening to the sample spectral data in the associated dataset to extract spectral features; normalizing the weighing data and detection environment data in the associated dataset to extract statistical features; and fusing the grain visual features, spectral features, and statistical features to obtain a standardized feature set.

[0094] Furthermore, the steps of calling the multi-grain adaptive AI detection model group based on the standardized feature set include: automatically identifying grain types based on the standardized feature set to obtain grain type identification results;

[0095] Based on the grain type identification results, the corresponding appearance index detection sub-model and physicochemical index prediction sub-model are automatically matched; the standardized feature set is input into the matched appearance index detection sub-model and physicochemical index prediction model to obtain the detection results of each individual physicochemical index.

[0096] Furthermore, the appearance index detection sub-model adopts a combination of the improved YOLOv8 target detection algorithm and the U-Net image segmentation algorithm to identify and count imperfect particles, moldy particles, insect-eaten particles and damaged particles, and calculate the area ratio of abnormal regions.

[0097] Furthermore, the physicochemical index prediction sub-model adopts a fusion algorithm of support vector regression and deep neural network. The input is spectral features and detection environment data, and the output is the predicted values ​​of moisture, protein, fat and starch content.

[0098] Furthermore, the steps for multi-dimensional fusion and analysis of the test results of each individual physicochemical indicator specifically include: performing data-level weighted fusion of the multi-source test results of the same indicator to obtain the corrected individual indicator results; performing feature-level fusion of the corrected individual indicator results to obtain the comprehensive quality feature vector; and performing decision-level fusion and analysis based on the comprehensive quality feature vector and the preset quality evaluation standard to obtain the comprehensive quality grade of grain and settlement parameters.

[0099] Furthermore, it also includes: regularly collecting manually reviewed data and laboratory test data to construct an incrementally updated dataset; using an incremental learning algorithm to update the multi-grain adaptive AI detection model group online based on the incrementally updated dataset; evaluating the detection accuracy of the updated model, and activating the updated model when the accuracy improvement exceeds a preset threshold.

[0100] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0101] This invention also provides an AI-powered intelligent and precise analysis system for multi-valley physical and chemical indicators. This device is used to implement the above embodiments and preferred embodiments, and details already described will not be repeated. As used below, the term "module" can refer to a combination of software and / or hardware that performs a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0102] Figure 2 According to one embodiment of the present invention, an AI-powered intelligent and precise analysis system for multi-valley physical and chemical indicators includes:

[0103] The acquisition unit 201 is used to collaboratively trigger the acquisition of multi-source heterogeneous data based on the intelligent control unit to obtain a related dataset with a unified timestamp;

[0104] Preprocessing unit 202 is used to perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set;

[0105] Calling unit 203 is used to call the multi-grain adaptive AI detection model group based on the standardized feature set to obtain the detection results of each individual physicochemical index.

[0106] The fusion unit 204 is used to perform multi-dimensional fusion analysis on the test results of each individual physicochemical indicator to obtain the comprehensive quality grade of grain and settlement parameters.

[0107] Upload unit 205 is used to upload the comprehensive quality grade of grain and settlement parameters to the grain quality inspection platform.

[0108] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0109] According to one embodiment of the present invention, an electronic device is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the above-mentioned AI intelligent and accurate judgment and analysis method for multi-valley physical indicators when it runs.

[0110] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program:

[0111] Step S1: In collaboration with the intelligent control unit, multi-source heterogeneous data acquisition is triggered to obtain a related dataset carrying a unified timestamp;

[0112] Step S2: Perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set;

[0113] Step S3: Based on the standardized feature set, call the multi-grain adaptive AI detection model group to obtain the detection results of each individual physicochemical index;

[0114] Step S4: Perform multi-dimensional fusion analysis on the test results of each individual physicochemical indicator to obtain the comprehensive quality grade of grain and settlement parameters;

[0115] Step S5: Upload the comprehensive quality grade of the grain and the settlement parameters to the grain quality inspection platform.

[0116] According to one embodiment of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the storage medium is located to execute the above-mentioned AI intelligent and accurate judgment and analysis method for multi-valley physical and chemical indicators.

[0117] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0118] Step S1: In collaboration with the intelligent control unit, multi-source heterogeneous data acquisition is triggered to obtain a related dataset carrying a unified timestamp;

[0119] Step S2: Perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set;

[0120] Step S3: Based on the standardized feature set, call the multi-grain adaptive AI detection model group to obtain the detection results of each individual physicochemical index;

[0121] Step S4: Perform multi-dimensional fusion analysis on the test results of each individual physicochemical indicator to obtain the comprehensive quality grade of grain and settlement parameters;

[0122] Step S5: Upload the comprehensive quality grade of the grain and the settlement parameters to the grain quality inspection platform.

[0123] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0124] According to one embodiment of the present invention, a computer program product is also provided, including a computer program that, when executed by a processor, implements the above-mentioned AI intelligent and accurate judgment and analysis method for multi-valley physical indicators.

[0125] Optionally, in this embodiment, the above-mentioned computer program product can be configured as a computer program that performs the following steps:

[0126] Step S1: In collaboration with the intelligent control unit, multi-source heterogeneous data acquisition is triggered to obtain a related dataset carrying a unified timestamp;

[0127] Step S2: Perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set;

[0128] Step S3: Based on the standardized feature set, call the multi-grain adaptive AI detection model group to obtain the detection results of each individual physicochemical index;

[0129] Step S4: Perform multi-dimensional fusion analysis on the test results of each individual physicochemical indicator to obtain the comprehensive quality grade of grain and settlement parameters;

[0130] Step S5: Upload the comprehensive quality grade of the grain and the settlement parameters to the grain quality inspection platform.

[0131] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0132] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0133] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.

[0134] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0135] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0136] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0137] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. An AI-powered intelligent and precise analysis method for multiple valley physical and chemical indicators, characterized in that, include: Based on the collaborative triggering of multi-source heterogeneous data acquisition by intelligent control unit, a related dataset with a unified timestamp is obtained; The associated dataset is subjected to multimodal adaptive preprocessing to obtain a standardized feature set; Based on the standardized feature set, the multi-grain adaptive AI detection model group is invoked to obtain the detection results of each individual physicochemical index. The test results of each individual physicochemical indicator are analyzed from multiple dimensions to obtain the comprehensive quality grade of grain and settlement parameters; The comprehensive quality grade of the grain and the settlement parameters are uploaded to the grain quality inspection platform.

2. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 1, characterized in that, The steps for collaboratively triggering multi-source heterogeneous data acquisition based on intelligent control units specifically include: In response to the vehicle entry signal, the unattended weighing system collects vehicle identification information and weighing data to generate a unique business serial number. Based on the unique business serial number, the intelligent sampling unit and the multispectral detection unit are triggered to simultaneously acquire sample image data, sample spectral data, and detection environment data; The vehicle identification information, the weighing data, the sample image data, the sample spectral data, and the detection environment data are associated with the unique business serial number and the unified timestamp to obtain the associated dataset.

3. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 2, characterized in that, The steps of performing multimodal adaptive preprocessing on the associated dataset specifically include: Adaptive morphological segmentation and noise filtering were applied to the sample image data in the associated dataset to extract grain visual features; The sample spectral data in the associated dataset are subjected to multivariate scattering correction and characteristic wavelength screening to extract spectral features; The weighing data and detection environment data in the associated dataset are normalized, and statistical features are extracted. The standardized feature set is obtained by fusing the visual features of the grains, the spectral features, and the statistical features.

4. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 3, characterized in that, The steps of calling the multi-grain adaptive AI detection model group based on the standardized feature set include: Automatic grain type identification is performed based on the standardized feature set to obtain the grain type identification result; The corresponding appearance index detection sub-model and physicochemical index prediction sub-model are automatically matched based on the grain type identification results. The standardized feature set is input into the matched appearance index detection sub-model and the physicochemical index prediction sub-model to obtain the detection results of each individual physicochemical index.

5. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 4, characterized in that, The appearance index detection sub-model uses a combination of the improved YOLOv8 target detection algorithm and the U-Net image segmentation algorithm to identify and count imperfect particles, moldy particles, insect-eaten particles and broken particles, and calculate the area ratio of abnormal regions.

6. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 4, characterized in that, The physicochemical index prediction sub-model adopts a fusion algorithm of support vector regression and deep neural network. The input is the spectral features and the detection environment data, and the output is the predicted values ​​of moisture, protein, fat and starch content.

7. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 6, characterized in that, The step of performing multi-dimensional fusion analysis on the test results of each individual physicochemical indicator specifically includes: Data-level weighted fusion is performed on the multi-source detection results of the same indicator to obtain the corrected single indicator result; The modified individual indicator results are fused at the feature level to obtain a comprehensive quality feature vector. Based on the comprehensive quality feature vector and the preset quality evaluation standard, a decision-level fusion analysis is performed to obtain the comprehensive quality grade of the grain and the settlement parameters.

8. The AI-powered intelligent and precise analysis method for multi-valley physicochemical indicators according to claim 1, characterized in that, Also includes: Regularly collect manually reviewed data and laboratory test data to build an incrementally updated dataset; Based on the incrementally updated dataset, the multi-grain adaptive AI detection model group is updated online using an incremental learning algorithm; The detection accuracy of the updated model is evaluated, and the updated model is activated when the accuracy improvement exceeds a preset threshold.

9. An AI-powered intelligent and precise analysis system for multiple physicochemical indicators, characterized in that, include: The acquisition unit is used to collaboratively trigger the acquisition of multi-source heterogeneous data based on the intelligent control unit to obtain a related dataset with a unified timestamp; The preprocessing unit is used to perform multimodal adaptive preprocessing on the associated dataset to obtain a standardized feature set; The calling unit is used to call the multi-grain adaptive AI detection model group based on the standardized feature set to obtain the detection results of each individual physicochemical index. The fusion unit is used to perform multi-dimensional fusion analysis on the test results of each individual physicochemical indicator to obtain the comprehensive quality grade of grain and settlement parameters. The uploading unit is used to upload the comprehensive quality grade of the grain and the settlement parameters to the grain quality inspection platform.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the AI-powered intelligent and precise judgment and analysis method for multi-valley physical indicators as described in any one of claims 1 to 8.

Images