Jujube variety discrimination method and system based on artificial intelligence

Abstract

The invention provides a jujube variety discrimination method and system based on artificial intelligence. The method belongs to the technical field of agricultural artificial intelligence and computer vision crossing. The method comprises the following steps: synchronously collecting multi-modal original data of jujube samples, and constructing a multi-modal original data set; preprocessing the multi-modal original data set to obtain standardized multi-modal feature data; a super-dimensional feature space is constructed based on the standardized multi-modal feature data, and a jujube super-dimensional feature vector set is generated through feature cross fusion. By synchronously acquiring the macroscopic visual image, the hyperspectral data, the three-dimensional form scanning data and the microscopic texture image, various feature information of jujube samples can be comprehensively and accurately captured, a super-dimensional feature space is constructed, and deep fusion of cross-modal features is realized, so that the precision and robustness of variety discrimination are improved.

Description

Technical Field

[0001] This invention proposes a method and system for identifying jujube varieties based on artificial intelligence, belonging to the interdisciplinary field of agricultural artificial intelligence and computer vision. Background Technology

[0002] Jujubes, as an important economic crop with both edible and medicinal value, are widely cultivated globally, with numerous varieties. Traditional jujube variety identification relies heavily on manual experience, involving observation of appearance and taste. This method is not only inefficient but also susceptible to subjective influences, making accuracy difficult to guarantee. With the large-scale development of the jujube industry, the need for precise and efficient variety identification methods is becoming increasingly urgent.

[0003] In recent years, multimodal data acquisition technologies have continued to develop, with equipment such as high-definition cameras, hyperspectral analyzers, 3D scanners, and microscopic imaging systems enabling the acquisition of rich information from jujube samples across multiple dimensions, from macroscopic to microscopic and from visual to spectral perspectives. Significant progress has also been made in the field of artificial intelligence in feature fusion and classification model construction, providing new technical pathways for jujube variety identification. However, currently, there is no mature system capable of fully integrating the advantages of multimodal data and utilizing artificial intelligence technology to achieve high-precision, intelligent jujube variety identification. Summary of the Invention

[0004] This invention provides a method and system for identifying jujube varieties based on artificial intelligence, in order to solve the problems mentioned in the background art above:

[0005] This invention proposes an artificial intelligence-based method for identifying jujube varieties, the method comprising:

[0006] S1: Synchronously collect multimodal raw data of jujube samples to construct a multimodal raw dataset; preprocess the multimodal raw dataset to obtain standardized multimodal feature data; construct a hyperdimensional feature space based on the standardized multimodal feature data, and generate a hyperdimensional feature vector set of jujube samples through feature cross-fusion;

[0007] S2: Input the hyperdimensional feature vector set into the dynamic gated fusion network, and adaptively adjust the contribution of each modality feature through the dynamic weight allocation module. Combine the cross-modal attention mechanism to strengthen the association of key features and output a robust fusion feature vector. Based on the robust fusion feature vector, perform initial variety classification to obtain preliminary variety discrimination results and corresponding confidence scores.

[0008] S3: Introduce a prior knowledge base of the genus *Ziziphus* evolutionary lineage, semantically align the preliminary variety identification results with the prior knowledge base of the genus *Ziziphus* evolutionary lineage, and correct low-confidence classification results through knowledge graph reasoning; based on the corrected results, generate a variety feature attribution graph to clarify the contribution path of each modality feature to the classification decision.

[0009] S4: Construct a blockchain-based trusted traceability system to store the life data of jujube samples on the blockchain and generate unique digital fingerprints through smart contracts; based on the distributed verification mechanism of blockchain nodes, cross-verify the identification results of jujube varieties across regions;

[0010] S5: Integrate the fused data from the dynamic gating fusion network to generate a structured jujube variety identification report.

[0011] This invention proposes an artificial intelligence-based jujube variety identification system, comprising:

[0012] One or more processors;

[0013] Memory, used to store one or more programs.

[0014] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the above.

[0015] Beneficial effects of this invention:

[0016] 1. By simultaneously acquiring macroscopic visual images, hyperspectral data, three-dimensional morphological scanning data, and microscopic texture images, it is possible to comprehensively and accurately capture various characteristic information of jujube samples, construct a hyperdimensional feature space, and achieve deep fusion of cross-modal features, thereby improving the accuracy and robustness of variety identification.

[0017] 2. By adaptively adjusting the contribution of each modal feature and combining a cross-modal attention mechanism, this network can effectively strengthen the correlation of key features, optimize the model's utilization of data from each modality, and further improve the accuracy of variety classification and the ability to handle low-confidence classification results.

[0018] 3. By introducing a prior knowledge base of the genus *Ziziphus* evolutionary lineage and combining it with knowledge graph reasoning techniques, semantic alignment correction can be performed on low-confidence classification results. This process not only improves the credibility of the discrimination results but also provides greater transparency and interpretability for the discrimination process.

[0019] 4. By constructing a blockchain traceability system, the collection information, characteristic data, and discrimination results of jujube samples are reliably stored, ensuring the immutability and traceability of the data. Smart contracts generate unique digital fingerprints, and the distributed verification mechanism of blockchain nodes is used to cross-verify discrimination results across regions, further enhancing the reliability of the data and the security of the system.

[0020] 5. This method can generate structured jujube variety identification reports, including variety name, characteristic attribution analysis, discrimination confidence level, evolutionary phylogenetic correlation, and blockchain-based evidence, helping relevant personnel to understand the identification results more intuitively and providing comprehensive support for the management, protection, and research of jujube varieties. Attached Figure Description

[0021] Figure 1 This is a diagram illustrating the steps of the method described in this invention. Detailed Implementation

[0022] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0023] One embodiment of the present invention, such as Figure 1 As shown, an artificial intelligence-based method for identifying jujube varieties includes:

[0024] S1: Simultaneously collect multimodal raw data of jujube samples, including macroscopic visual images, hyperspectral data, three-dimensional morphological scan data, and microscopic texture images to construct a multimodal raw dataset; preprocess the multimodal raw dataset to obtain standardized multimodal feature data; construct a hyperdimensional feature space based on the standardized multimodal feature data, and generate a jujube hyperdimensional feature vector set containing joint spatial-spectral-morphological-texture representations through feature cross-fusion;

[0025] S2: Input the hyperdimensional feature vector set into the dynamic gated fusion network, and adaptively adjust the contribution of each modality feature through the dynamic weight allocation module. Combine the cross-modal attention mechanism to strengthen the association of key features and output a robust fusion feature vector. Based on the robust fusion feature vector, perform initial variety classification to obtain preliminary variety discrimination results and corresponding confidence scores.

[0026] S3: Introduce a prior knowledge base of the genus *Ziziphus* evolutionary lineage, which includes the phylogenetic tree of *Ziziphus* species, genetic marker data, and ecological distribution information; semantically align the preliminary variety discrimination results with the prior knowledge base of the genus *Ziziphus* evolutionary lineage, and correct low-confidence classification results through knowledge graph reasoning; based on the corrected discrimination results, generate a variety feature attribution graph to clarify the contribution path of each modality feature to the classification decision, and realize the visual explanation of the discrimination process;

[0027] S4: Construct a blockchain-based trusted traceability system to store the life data of jujube samples on the blockchain. The life data includes collection time, geographical location, multimodal raw data, hyperdimensional feature vectors, and final variety identification results. Generate unique digital fingerprints through smart contracts. Cross-verify the cross-regional jujube variety identification results based on the distributed verification mechanism of blockchain nodes.

[0028] S5: Integrate the fusion data from the dynamic gating fusion network, which includes robust discrimination results, enhanced interpretation data of the jujube evolutionary lineage correction, and blockchain traceability verification information to generate a structured jujube variety discrimination report; the discrimination report includes variety name, biological characteristic attribution analysis, discrimination confidence level, evolutionary lineage correlation description, and blockchain evidence.

[0029] The working principle and effects of the above technical solution are as follows: By simultaneously collecting macroscopic, spectral, three-dimensional, and microscopic multi-dimensional data and constructing a hyperdimensional feature space, the problem of one-sided information from single-modal data is avoided, making the basis for variety identification more comprehensive and improving the utilization rate of multimodal data; the dynamic gating fusion network combined with a cross-modal attention mechanism can adaptively adjust the contribution of each modality feature, reduce interference from irrelevant features, and make the robust fusion feature vector more in line with the needs of variety identification, enhancing the pertinence of feature fusion; the introduction of a prior knowledge base of the jujube genus evolutionary lineage and the correction of low-confidence results through knowledge graph reasoning reduce the impact of data noise or feature similarity. The system reduces the false positive rate and improves the efficiency of correcting preliminary judgment results; by generating a variety characteristic attribution map to visualize the contribution paths of each modality feature, it reduces decision-making doubts caused by the black box problem of traditional AI judgment and enhances the interpretability of the judgment process; the blockchain system realizes on-chain storage and distributed verification of life data, avoids the risk of data tampering, reduces inconsistencies in judgment results across regions, and improves the credibility of judgment results; it integrates core data from multiple stages to generate structured reports, clarifies the biological characteristics, evolutionary relationships, and evidence information of varieties, reduces the information query cost for subsequent traceability or application, and enhances the practical value of the judgment report.

[0030] In one embodiment of the present invention, S1 includes:

[0031] S11. Screen the jujube samples, which include individual jujubes of different varieties, growth cycles and ecological environments. Remove invalid samples that are damaged by pests and diseases or have morphological defects, and determine the set of valid samples.

[0032] S12. For the effective sample set, multi-modal raw data is collected simultaneously using multiple devices, including: macroscopic visual images are collected by a high-definition camera, hyperspectral data is collected by a hyperspectral instrument, three-dimensional morphological scanning data is collected by a three-dimensional scanner, and microscopic texture images are collected by a microscopic imaging system. All data are integrated to construct a multimodal raw dataset.

[0033] S13. Preprocess the multimodal raw dataset. The preprocessing includes denoising and white balance correction of macroscopic visual images, spectral baseline correction and redundant band removal of hyperspectral data, point cloud denoising and registration of three-dimensional morphological scanning data, grayscale normalization of microscopic texture images, elimination of data interference terms, and obtaining standardized multimodal feature data.

[0034] S14. Based on standardized multimodal feature data, establish a hyperdimensional feature space coordinate system, and map macroscopic visual features, hyperspectral features, three-dimensional morphological features, and microscopic texture features to different dimensional axes to construct a hyperdimensional feature space containing multimodal dimensions.

[0035] S15. Perform feature cross-fusion in the hyperdimensional feature space, and use the attention mechanism to explore the correlation between various modal features (such as the correspondence between spectral features and microscopic texture features, and the consistency between three-dimensional morphology and macroscopic vision) to generate a hyperdimensional feature vector set of jujube containing joint representation of space-spectrum-morphology-texture.

[0036] The working principle and effects of the above technical solution are as follows: By screening jujube samples of different varieties, growth cycles, and ecological environments and removing invalid samples, the interference of incomplete or damaged samples on subsequent data processing is reduced, providing more reliable basic data for subsequent discrimination and improving the stability of sample quality; Multimodal data is collected simultaneously by multiple devices, covering macroscopic to microscopic and visual to spectral feature dimensions, avoiding the problem of one-sided information from single data types, making the data representation more closely match the true characteristics of jujube varieties, and enhancing the comprehensiveness of the original data; Targeted preprocessing eliminates interference items such as image noise and spectral redundancy, reducing... This approach eliminates interference from invalid information in feature analysis, improves the purity of standardized multimodal feature data, and reduces the impact of data interference on feature extraction. Establishing a hyperdimensional feature space maps different modal features to independent dimensional axes, avoiding dimensional confusion between different types of features, making the uniqueness of each modality more readily apparent, and improving the discriminative power of multimodal features. Furthermore, by exploiting an attention mechanism to uncover the correspondence between cross-modal features, the approach reduces the blindness in feature fusion, allowing the generated hyperdimensional feature vector set to more accurately reflect the joint features of the variety, providing more representative feature support for subsequent discrimination, and enhancing the effectiveness of feature association.

[0037] In one embodiment of the present invention, step S14 includes:

[0038] The core feature parameters of each modality are extracted from the standardized multimodal feature data. Among them, the macroscopic visual feature parameters include the color channel mean and texture entropy value, the hyperspectral feature parameters include the reflectance of the feature band and the spectral peak value, the three-dimensional morphological feature parameters include volume, surface area and curvature distribution, and the microscopic texture feature parameters include texture granularity and porosity, forming a set of core feature parameters of the modality.

[0039] Based on the physical meaning and data volume of each modal feature parameter, the modal core feature parameter set is normalized, and parameters of different dimensions are uniformly mapped to the [0,1] interval to generate a normalized feature parameter set.

[0040] Based on the normalized feature parameter set, a dimensional axis system of the hyperdimensional feature space is constructed, mapping macroscopic visual feature parameters to the X-axis cluster, hyperspectral feature parameters to the Y-axis cluster, three-dimensional morphological feature parameters to the Z-axis cluster, and microscopic texture feature parameters to the W-axis cluster, forming a four-dimensional basic coordinate system framework.

[0041] In the four-dimensional basic coordinate system framework, independent dimension sub-axis are assigned to each feature parameter. The angle between the sub-axis is adjusted by calculating the parameter correlation, so that the sub-axis of parameters with high correlation are distributed at acute angles and those with low correlation are distributed at obtuse angles, thereby optimizing the coordinate system structure.

[0042] All normalized feature parameters are spatially positioned according to their corresponding dimensional sub-axis to form a hyperdimensional feature space that includes multiple modal dimensions such as macroscopic vision, hyperspectral, three-dimensional morphology, and microscopic texture.

[0043] The working principle and effects of the above technical solution are as follows: By extracting the core feature parameters of each modality, the interference of redundant parameters on subsequent analysis is avoided, making the feature data more focused on the key information for variety identification and improving the pertinence of the feature parameters; normalization processing maps the parameters uniformly to the [0,1] interval, reducing the problem of feature weight imbalance caused by differences in data magnitude, making each parameter more comparable in analysis, and reducing the conflicting impact of parameters of different dimensions; constructing a four-dimensional basic coordinate system framework assigns different modal parameters to independent axis clusters, avoiding the confusion of modal features, making the unique attributes of each modality easier to identify, and enhancing the dimensionality of multimodal features; adjusting the angle of the sub-axis through parameter correlation makes the sub-axis of parameters with high correlation closer, reducing the interference of irrelevant parameters in the feature space, making the feature distribution more in line with the actual correlation logic, and improving the rationality of the coordinate system structure; accurately positioning all normalized parameters to the corresponding sub-axis avoids feature omission, and the resulting multimodal dimensional space can more comprehensively carry the feature information of jujube varieties, laying a solid foundation for subsequent fusion and enhancing the integrity of the hyperdimensional feature space.

[0044] In one embodiment of the present invention, S15 includes:

[0045] S151. Extract key anchor points for each modality feature in the hyperdimensional feature space. Among them, the macroscopic visual feature anchor point is the center coordinate of the color change region, the hyperspectral feature anchor point is the peak point of reflectivity of the feature band, the three-dimensional morphological feature anchor point is the curvature extreme point, and the microscopic texture feature anchor point is the texture structure inflection point, forming a multimodal feature anchor point set.

[0046] S152. Calculate the spatial correlation between feature anchor sets, measure the spatial proximity of different modal anchors by Euclidean distance, evaluate the matching degree of feature attributes by cosine similarity, and generate an anchor correlation matrix.

[0047] S153. Input the anchor point correlation matrix into the attention mechanism model, assign high weight values ​​to high correlation anchor point pairs, strengthen the intrinsic connection between cross-modal features (e.g., the correlation between three-dimensional morphological curvature anchor points and macroscopic visual color anchor points), and generate a weighted feature correlation map.

[0048] S154. Based on the weighted feature association map, cross-mapping and fusion of multimodal features in the hyperdimensional feature space are performed. The feature dimensions with high correlation are vector superimposed, while the feature dimensions with low correlation are kept independently represented to form a draft of the fused feature vector.

[0049] S155. Perform dimensionality normalization on the initial draft of the fused feature vector to eliminate the fusion bias caused by the difference in the magnitude of features of different modalities, and finally generate a set of jujube hyperdimensional feature vectors containing joint spatial-spectral-morphological-texture representations.

[0050] The working principle and effects of the above technical solution are as follows: By extracting key anchor points for each modality (such as color mutation centers and spectral reflectance peaks), invalid analysis of irrelevant feature regions is avoided, allowing subsequent correlation calculations to focus more on the core features for variety identification, thus improving feature focus; the correlation degree of anchor points is calculated using both Euclidean distance and cosine similarity, reducing misjudgments caused by a single metric, and the generated correlation degree matrix more closely reflects the true relationship between features, enhancing the accuracy of cross-modal correlation; the attention mechanism assigns high weights to highly correlated anchor point pairs, strengthening the connection between key features while weakening irrelevant features. This approach avoids confusing the discrimination direction with low-association features, making the feature association map more targeted and reducing the interference of irrelevant features. Vector superposition and independent representation are performed according to the degree of association, preserving the synergistic value of highly correlated features without losing the unique information of low-association features, reducing feature loss during the fusion process and improving the rationality of feature fusion. Finally, dimensionality normalization eliminates magnitude bias, preventing the problem of one type of modality feature masking other features due to excessively large values. This allows the generated joint representation vector to more evenly reflect the multi-dimensional features of jujube varieties, enhancing the reliability of the hyperdimensional feature vector.

[0051] In one embodiment of the present invention, S154 includes:

[0052] The weighted feature association graph is analyzed, and high-association feature dimension pairs with an association degree threshold (e.g., a preset association degree ≥ 0.8) are extracted. At the same time, low-association feature dimensions with an association degree < 0.8 are filtered out, and a high / low-association feature dimension classification table is generated.

[0053] For highly correlated feature dimension pairs, the feature vector data corresponding to each dimension is obtained, and the feature superposition between dimensions is achieved by vector dot product operation. The core feature information of the superimposed vector (such as the superposition feature of spectral reflectance and microtexture porosity) is retained, and a highly correlated fused feature vector group is generated.

[0054] For low-association feature dimensions, feature vectors of each dimension are extracted separately, and their modality types are labeled by feature identifiers (e.g., 3D morphology-volume feature vectors, macroscopic vision-color feature vectors) to ensure that the independently represented features are traceable and generate low-association independent feature vector groups.

[0055] The highly correlated fused feature vector group and the low-correlation independent feature vector group are concatenated according to the dimensional order of the hyperdimensional feature space to form a combined vector sequence containing fused features and independent features;

[0056] The integrity of the combined vector sequence is checked to confirm that no feature dimensions are missing or repeated, and finally a draft of the fused feature vector is formed.

[0057] The working principle and effects of the above technical solution are as follows: By setting a correlation threshold to distinguish between high and low correlation feature dimensions and generating a classification table, the confusion of processing features with different correlation degrees together is avoided, making the subsequent fusion direction clearer and improving the clarity of feature classification; by using vector dot product operation to superimpose high correlation dimension features, core information is preserved and the linkage effect between features is strengthened, reducing the information waste when processing high correlation features alone and enhancing the synergistic value of high correlation features; low correlation features are extracted separately and labeled with modality type to avoid them being masked or misfused by high correlation features, so that the unique value of each low correlation feature can be preserved and the risk of loss of low correlation features is reduced; by concatenating two types of vector groups according to the order of the hyperdimensional feature space dimension, the subsequent processing trouble caused by the chaotic arrangement of vectors is avoided, the structure of the combined vector sequence is clearer, and the regularity of the vector sequence is improved; by verifying the integrity to confirm that there are no missing or duplicate features, the subsequent normalization processing is avoided due to missing or redundant features, making the reliability of the initial draft of the fused feature vector higher and reducing the error risk of the initial draft of the fusion.

[0058] In one embodiment of the present invention, S2 includes:

[0059] S21. Perform modal feature alignment processing on the jujube hyperdimensional feature vector set, unify the dimension and data format of each modal feature, avoid fusion deviation caused by modal differences, and generate an aligned hyperdimensional feature vector set.

[0060] S22. Input the aligned hyperdimensional feature vector set into the dynamic gated fusion network. Through the network's built-in dynamic weight allocation module, the contribution weight of each modal feature is adaptively adjusted according to the importance of each modal feature in variety discrimination (e.g., the advantage of hyperspectral features in characterizing the intrinsic components of a variety), and a feature vector set after weight allocation is generated.

[0061] S23. For the feature vector set after weight allocation, a cross-modal attention mechanism is introduced to strengthen the correlation between key features (such as the correlation between "fruit shape index" in three-dimensional morphology and "fruit peel color" in macroscopic vision), suppress interference from irrelevant features, and output a robust fused feature vector.

[0062] S24. Input the robust fusion feature vector into a lightweight classifier (e.g., an improved convolutional neural network classifier), set the classification threshold, perform initial variety classification on the jujube samples, and obtain preliminary variety discrimination results;

[0063] S25. Based on the probability distribution output by the classifier, calculate the confidence score corresponding to the preliminary variety identification result, set a confidence threshold (e.g., 0.8), and mark the low confidence classification results below the threshold.

[0064] The working principle and effect of the above technical solution are as follows: by unifying the dimensions and formats of each modal feature through alignment processing, the fusion chaos caused by inconsistent data formats is avoided, making subsequent feature processing smoother and reducing the fusion deviation caused by modal differences.

[0065] Dynamic gating networks adjust weights based on the actual role of each mode in discrimination, such as highlighting the advantages of hyperspectral representation of intrinsic components, reducing the irrationality of one-size-fits-all weight allocation, and improving the rationality of feature weight allocation.

[0066] The cross-modal attention mechanism strengthens core related features such as fruit shape index and peel color, while suppressing interference from irrelevant information, making the output fused feature vector more focused on the key to variety identification and enhancing the correlation of key features.

[0067] Using a lightweight classifier to process robust fusion features reduces computational burden while ensuring discrimination performance, avoids the time-consuming problem caused by complex models, and improves the efficiency of initial classification.

[0068] By calculating the confidence level and marking results below the threshold, a clear direction is provided for subsequent corrections, avoiding the direct use of unreliable results that could affect the final judgment accuracy and reducing the risk of misjudgment based on low-confidence results.

[0069] In one embodiment of the present invention, step S22 includes:

[0070] After parsing the aligned hyperdimensional feature vector set, feature sub-vectors of four modalities are extracted. The four modalities include macroscopic vision, hyperspectral, three-dimensional morphology and microscopic texture. The core representation direction of each modal feature in variety identification is determined (e.g., hyperspectral corresponds to internal components, three-dimensional morphology corresponds to appearance structure), and a list of modal feature classifications is generated.

[0071] The modal feature classification list is input into the feature importance evaluation submodule of the dynamic gating fusion network. Based on the feature contribution statistics of historical discrimination samples (e.g., the accuracy of hyperspectral features in the discrimination of variety A is improved by 35%), the basic weight values ​​of each modal feature are initially calculated to form a basic weight matrix.

[0072] The real-time adjustment algorithm of the dynamic weight allocation module is invoked, and the basic weight matrix is ​​dynamically corrected by taking into account the special characteristics of the current jujube samples (e.g., the microscopic texture of some samples is blurred and the weight of this mode needs to be reduced) to generate real-time weight adjustment coefficients.

[0073] The real-time weight adjustment coefficients are weighted dimension-wise with each modal feature sub-vector to ensure that high-importance modal features (such as hyperspectral features) have a higher proportion in the fusion, while the proportion of low-importance modal features is reasonably reduced, resulting in weighted modal feature sub-vectors.

[0074] The weighted feature vectors of each modality are reorganized according to the order of the hyperdimensional feature space to form a set of feature vectors with complete structure and reasonable weight allocation.

[0075] The working principle and effects of the above technical solution are as follows: By analyzing the vector set, the core representation direction of each modality is clarified and a classification list is generated, avoiding a vague understanding of the role of modal features, making subsequent weight allocation more in line with the discrimination requirements, and improving the pertinence of modal feature analysis; the basic weights are calculated by combining the feature contribution statistics of historical samples, reducing the subjectivity of setting weights based on experience, making the initial weight allocation more data-supported, and enhancing the reliability of the basic weights; the weights are dynamically corrected according to the current sample features (e.g., blurred microscopic texture), avoiding the rigidity of using fixed weights to deal with all samples, making weight adjustment more flexible and adaptable to the actual situation, and reducing the weight bias caused by sample particularity; the weighted operation is performed dimension by dimension to ensure the proportion of high-importance modalities (e.g., hyperspectral), avoiding the imbalance of local feature weights caused by overall weighting, making the contribution of each modal feature more reasonable, and improving the accuracy of feature weighting; the weighted sub-vectors are re-integrated according to the order of the hyperdimensional space dimensions, avoiding the impact of disordered vector arrangement on subsequent processing, making the final generated feature vector set structure more regular and more usable, and reducing the structural chaos of the integrated vector set.

[0076] In one embodiment of the present invention, S3 includes:

[0077] S31. Load the prior knowledge base of the genus *Ziziphus* evolutionary lineage, perform structured analysis on the evolutionary tree of *Ziziphus* species, genetic marker data (e.g., SSR molecular markers), and ecological distribution information stored in the base, extract key knowledge that can be used for variety identification correction (e.g., genetic differences of closely related varieties, ecological distribution range of specific varieties), and generate a structured prior knowledge set.

[0078] S32. Semantically align the preliminary variety identification results with the structured prior knowledge set. Through entity matching (e.g., standardized matching of variety names) and feature mapping (e.g., matching the peel thickness in the identification results with the genetic marker association features of the variety in the knowledge base), unify the data semantic standards and generate the aligned identification-knowledge dataset.

[0079] S33. For the low-confidence classification results in the aligned discriminant-knowledge dataset, call the knowledge graph reasoning engine to perform reasoning correction based on the kinship rules of the evolutionary tree and the matching degree rules of genetic markers (e.g., genetic similarity ≥ 95% is judged as closely related varieties) to obtain the corrected variety discrimination results;

[0080] S34. Based on the corrected variety discrimination results, trace the path of each modal feature on the classification decision (e.g., reflectance of a specific band in the hyperspectrum → determination of sugar content of the variety → association with variety A), generate a variety feature attribution map, and clarify the contribution ratio of each modal feature.

[0081] S35. Use visualization tools (such as heatmaps and feature contribution line graphs) to transform the variety feature attribution map and discrimination process into a visualization interface, intuitively showing the impact of different modal features on the final discrimination result, and realizing a visual explanation of the discrimination process.

[0082] The working principle and effects of the above technical solution are as follows: By extracting key correction information from a structured knowledge base, the application difficulties caused by fragmented knowledge are avoided, allowing data such as evolutionary lineages and genetic markers to accurately support discrimination correction and improving the utilization rate of prior knowledge; by unifying semantic standards through entity matching and feature mapping, the knowledge association failure caused by non-standard variety names and inconsistent feature descriptions is avoided, making the discrimination results and knowledge base more seamlessly connected and reducing matching errors caused by data semantic bias; based on evolutionary tree kinship rules and genetic matching degree rules, the one-sidedness of judging solely by data features is reduced, the probability of misjudging closely related varieties is lowered, and the accuracy of correcting low-confidence results is enhanced; generating variety feature attribution maps clarifies the contribution paths of each modality, avoiding the problem of knowing the "what" but not the "why" in traditional discrimination, making the basis of each discrimination result clearly traceable and improving the traceability of the discrimination process; using visualization tools such as heatmaps and line graphs to display the process avoids the difficulty of understanding complex data and reasoning logic, allowing non-professionals to intuitively grasp the impact of features on discrimination and reducing the understanding threshold of discrimination results.

[0083] In one embodiment of the present invention, step S4 includes:

[0084] S41. Construct a blockchain-based trusted traceability system, design the blockchain's storage structure, determine the storage fields for life data (collection time, geographical location, hash value of multimodal raw data, superdimensional feature vector summary, and variety identification result), and use an asymmetric encryption algorithm to set data access permissions.

[0085] S42. Collect complete life data of jujube samples, format the data according to the blockchain storage structure, generate a standardized life dataset, and associate it with a unique sample identifier (e.g., sample number).

[0086] S43. Upload the standardized life dataset to the blockchain system, and complete the on-chain storage of the data through a consensus mechanism (such as PoS proof of stake) to form an immutable life data chain;

[0087] S44. Call the blockchain smart contract to generate a unique digital fingerprint for the sample based on the hash value of the standardized life data, and bind the digital fingerprint with the sample identifier as a trusted credential for the sample's identity.

[0088] S45. Activate the distributed verification mechanism of blockchain nodes. Blockchain nodes in different regions load the cross-regional jujube variety identification results and corresponding life data respectively. By comparing the consistency of digital fingerprints and the conformity of the identification logic, the cross-regional identification results are cross-verified and a unified identification result that has passed the verification is output.

[0089] The working principle and effects of the above technical solution are as follows: By designing a blockchain structure and setting access permissions with asymmetric encryption, the risk of arbitrary tampering or unauthorized access to life data is avoided, making the entire process of sample data more reliable and enhancing the security of data storage; Life data is formatted according to the blockchain storage structure and associated with a unique identifier, reducing the difficulty of traceability caused by chaotic data formats, ensuring that the data of each sample can be accurately matched, and improving the uniformity of data formats; By completing the on-chain storage through a consensus mechanism to form an immutable life data chain, the problem of human modification of data affecting the discrimination results is avoided, ensuring the originality of the data and reducing the possibility of data tampering; A unique digital fingerprint is generated based on the data hash value and bound to the sample identifier, reducing the risk of sample confusion or identity fraud, giving each sample a credible digital ID card, and enhancing the uniqueness of sample identity; By verifying and comparing fingerprints and logic through distributed nodes, the result conflict caused by different discrimination standards in different regions is avoided, reducing disputes when applying across regions and improving the consistency of cross-regional discrimination results.

[0090] In one embodiment of the present invention, S45 includes:

[0091] Extract the variety identification results and corresponding life data (including digital fingerprints and identification logic descriptions) of cross-regional jujube samples from the blockchain system, divide the data into groups by region, and generate regionalized identification-data association packages;

[0092] The system sends verification commands to nodes in each region of the blockchain and assigns corresponding regional discrimination-data association packets. After receiving the packets, the nodes parse the data packet structure, confirm the data integrity (e.g., no missing fields, identifiable hash values), and generate node data reception confirmation information.

[0093] Each regional node first performs a digital fingerprint consistency check, calculates the hash value of the locally received life data, and compares it with the digital fingerprint attached to the packet. If they match, the next step is taken; otherwise, it is marked as a fingerprint anomaly and the reason for the anomaly is recorded.

[0094] For data packets that pass fingerprint verification, the node initiates a logic conformity verification, compares the prior knowledge of the genus Ziziphus genus evolutionary lineage (e.g., the matching degree of variety genetic characteristics and ecological distribution adaptability), checks the logical relationship between the discrimination result and the life data (e.g., whether the hyperspectral features support the judgment result), and generates a logic verification report.

[0095] Summarize the verification results of each regional node (including fingerprint verification results and logical verification reports), and count the percentage of nodes that have passed verification (for example, a preset percentage of ≥80% of nodes passing is considered valid). If the standard is met, output a unified judgment result indicating that the verification has passed; otherwise, trigger a secondary verification process.

[0096] The working principle and effects of the above technical solution are as follows: Dividing associated packets by region and assigning them to corresponding nodes avoids verification chaos caused by mixed data, making the verification objectives of each node clearer and improving the systematic nature of cross-regional data verification; after receiving the data, nodes first confirm its integrity, avoiding the impact of missing fields, incorrect hash values, etc., on subsequent verification, reducing the waste of invalid verification, and lowering the risk of data loss during transmission; through digital fingerprint consistency comparison, data tampering is investigated from the source, reducing the possibility of false data participating in verification and enhancing the verification of data authenticity; after fingerprint verification is passed, logical connections are checked against prior knowledge, avoiding the problem of only looking at the surface consistency of data while ignoring the internal rationality, and improving the logical reliability of the judgment result; the final result is determined by the percentage of nodes passing the test, avoiding the one-sidedness of a single region having the final say, making the unified judgment result more credible, and reducing the controversy of cross-regional results.

[0097] In one embodiment of the present invention, step S5 includes:

[0098] S51. Collect core data from each stage, robust discrimination results output by the dynamic gating fusion network, enhanced explanatory data for the correction of the jujube evolutionary lineage (e.g. key information of the attribution map, correction reasoning process), and blockchain traceability verification information (digital fingerprint, node verification record). Standardize the data format and remove redundant information to generate an integrated discrimination dataset.

[0099] S52. Perform structured analysis on the integrated discriminant dataset to extract core information, including variety name, biological attribution of each modality feature (e.g., flavonoid content of the variety corresponding to the hyperspectral feature), final discriminant confidence, phylogenetic relationship with the target variety in the evolutionary lineage (e.g., genetic similarity with variety B is 92%, belonging to the third branch of the evolutionary tree), and blockchain evidence certificate number.

[0100] S53. Call the preset discrimination report template, fill the core information after structured analysis into the template in the logical order of variety basic information - characteristic attribution analysis - discrimination reliability description - traceability certificate, and generate a preliminary structured jujube variety discrimination report;

[0101] S54. Verify the preliminary identification report, check the accuracy of the data (e.g., the variety name is consistent with the blockchain evidence), logical coherence (e.g., the attribution analysis matches the identification results), and information completeness (e.g., missing evidence certificates), and correct any deviations or omissions in the report.

[0102] S55. Output a structured jujube variety identification report that has passed verification. The report includes the variety name, biological characteristic attribution analysis (including the contribution ratio of each modality feature), discrimination confidence (with the basis for confidence calculation), evolutionary lineage correlation description (including a schematic diagram of the evolutionary tree branches), and blockchain evidence (including a digital fingerprint query link).

[0103] The working principle and effects of the above technical solution are as follows: By collecting core data from each stage and unifying the format, eliminating redundancy, the chaotic accumulation of scattered and fragmented data is avoided, making the generated integrated dataset more refined and usable, and improving the efficiency of data integration; structured analysis extracts key information, reducing the interference of irrelevant information on core conclusions, making the report more focused and enhancing the core focus of the report content; filling the preset template according to a fixed logical order avoids reading difficulties caused by disordered content arrangement, making the initial report structure more regular and reducing the problem of chaotic report generation format; verification ensures data accuracy, logical coherence, and information completeness, reducing the decrease in report credibility due to omissions or errors, making the final output report more rigorous and improving the reliability of the report; including details such as feature contribution ratio, confidence basis, and phylogenetic tree diagram, the report content is not empty and general, allowing users to fully understand the judgment process and basis, facilitating subsequent application and traceability, and enhancing the practical value of the report.

[0104] One embodiment of the present invention provides an artificial intelligence-based jujube variety identification system, comprising:

[0105] One or more processors;

[0106] Memory, used to store one or more programs.

[0107] Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any one of the above.

[0108] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for identifying jujube varieties based on artificial intelligence, characterized in that, The method includes: S1: Synchronously collect multimodal raw data of jujube samples to construct a multimodal raw dataset; preprocess the multimodal raw dataset to obtain standardized multimodal feature data; construct a hyperdimensional feature space based on the standardized multimodal feature data, and generate a hyperdimensional feature vector set of jujube samples through feature cross-fusion; S2: Input the hyperdimensional feature vector set into the dynamic gated fusion network, and adaptively adjust the contribution of each modality feature through the dynamic weight allocation module. Combine the cross-modal attention mechanism to strengthen the association of key features and output a robust fusion feature vector. Based on the robust fusion feature vector, perform initial variety classification to obtain preliminary variety discrimination results and corresponding confidence scores. S3: Introduce a prior knowledge base of the genus *Ziziphus* evolutionary lineage, semantically align the preliminary variety identification results with the prior knowledge base of the genus *Ziziphus* evolutionary lineage, and correct low-confidence classification results through knowledge graph reasoning; based on the corrected results, generate a variety feature attribution graph to clarify the contribution path of each modality feature to the classification decision. S4: Construct a blockchain-based trusted traceability system to store the life data of jujube samples on the blockchain and generate unique digital fingerprints through smart contracts; based on the distributed verification mechanism of blockchain nodes, cross-verify the identification results of jujube varieties across regions; S5: Integrate the fused data from the dynamic gating fusion network to generate a structured jujube variety identification report.

2. The method for identifying jujube varieties based on artificial intelligence according to claim 1, characterized in that, S1 includes: S11. Screen the jujube samples, remove invalid samples damaged by pests and diseases or with incomplete morphology, and determine the set of valid samples. S12. For the effective sample set, use multiple devices to synchronously collect multimodal raw data and integrate all data to construct a multimodal raw dataset; S13. Preprocess the original multimodal dataset to obtain standardized multimodal feature data; S14. Based on standardized multimodal feature data, establish a hyperdimensional feature space coordinate system, and map macroscopic visual features, hyperspectral features, three-dimensional morphological features, and microscopic texture features to different dimensional axes to construct a hyperdimensional feature space. S15. Perform feature cross-fusion in the hyperdimensional feature space, and use the attention mechanism to explore the correlation between features of each modality to generate a hyperdimensional feature vector set of jujube.

3. The method for identifying jujube varieties based on artificial intelligence according to claim 2, characterized in that, S15 includes: S151. Extract key anchor points of each modality feature in the hyperdimensional feature space to form a multimodal feature anchor point set; S152. Calculate the spatial correlation between feature anchor sets, measure the spatial proximity of different modal anchors by Euclidean distance, evaluate the matching degree of feature attributes by cosine similarity, and generate an anchor correlation matrix. S153. Input the anchor point correlation matrix into the attention mechanism model, assign high weight values ​​to high correlation anchor point pairs, and generate a weighted feature correlation map. S154. Based on the weighted feature association map, cross-mapping and fusion of multimodal features in the hyperdimensional feature space are performed. The feature dimensions with high correlation are vector superimposed, while the feature dimensions with low correlation are kept independently represented to form a draft of the fused feature vector. S155. Perform dimensionality normalization on the initial draft of the fused feature vector to eliminate the fusion bias caused by the difference in the magnitude of features of different modalities, and finally generate a set of jujube-type hyperdimensional feature vectors.

4. The method for identifying jujube varieties based on artificial intelligence according to claim 3, characterized in that, S154 includes: The weighted feature association graph is analyzed, and high-association feature dimension pairs with an association degree above the threshold are extracted. At the same time, low-association feature dimensions with an association degree <0.8 are filtered out, and a high / low-association feature dimension classification table is generated. For highly correlated feature dimension pairs, the feature vector data corresponding to each dimension is obtained, and the vector dot product operation is used to realize the feature superposition between dimensions. The core feature information of the superimposed vector is retained to generate a highly correlated fused feature vector group. For low-association feature dimensions, extract the feature vectors of each dimension separately, and generate low-association independent feature vector groups by labeling their modality types with feature identifiers; The highly correlated fused feature vector group and the low-correlation independent feature vector group are concatenated according to the dimensional order of the hyperdimensional feature space to form a combined vector sequence containing fused features and independent features; The integrity of the combined vector sequence is verified, and the initial draft of the fused feature vector is finally formed.

5. The method for identifying jujube varieties based on artificial intelligence according to claim 1, characterized in that, S2 includes: S21. Perform modal feature alignment processing on the jujube hyperdimensional feature vector set to generate an aligned hyperdimensional feature vector set. S22. Input the aligned hyperdimensional feature vector set into the dynamic gated fusion network. Through the network's built-in dynamic weight allocation module, the contribution weight of each modality feature is adaptively adjusted according to the importance of each modality feature in variety discrimination, and a weighted feature vector set is generated. S23. For the feature vector set after weight allocation, a cross-modal attention mechanism is introduced to strengthen the correlation between key features, suppress interference from irrelevant features, and output a robust fused feature vector. S24. Input the robust fusion feature vector into the lightweight classifier, set the classification threshold, perform initial variety classification on the jujube samples, and obtain preliminary variety discrimination results; S25. Based on the probability distribution output by the classifier, calculate the confidence score corresponding to the preliminary variety identification result, set a confidence threshold, and mark the low confidence classification results below the threshold.

6. The method for identifying jujube varieties based on artificial intelligence according to claim 5, characterized in that, S22 includes: After parsing the aligned hyperdimensional feature vector set, feature sub-vectors of the four modalities are extracted, the core representation direction of each modal feature in variety discrimination is determined, and a list of modal feature classifications is generated. The modal feature classification list is input into the feature importance evaluation submodule of the dynamic gating fusion network. Based on the feature contribution statistics of historical discrimination samples, the basic weight values ​​of each modal feature are initially calculated to form a basic weight matrix. The real-time adjustment algorithm of the dynamic weight allocation module is invoked, and the basic weight matrix is ​​dynamically corrected based on the specific characteristics of the current jujube samples to generate real-time weight adjustment coefficients. The real-time weight adjustment coefficients are weighted dimension by dimension with the feature vectors of each modality to obtain the weighted feature vectors of each modality. The weighted feature vectors of each modality are reorganized according to the dimensional order of the hyperdimensional feature space to form a set of feature vectors after weight allocation.

7. The method for identifying jujube varieties based on artificial intelligence according to claim 1, characterized in that, The S3 includes: S31. Load the genus *Ziziphus* evolutionary phylogenetic prior knowledge base, perform structured analysis on the phylogenetic tree, genetic marker data and ecological distribution information of *Ziziphus* species stored in the base, extract key knowledge that can be used for variety identification and correction, and generate a structured prior knowledge set. S32. Semantically align the preliminary variety identification results with the structured prior knowledge set. Through entity matching and feature mapping, unify the data semantic standards and generate the aligned discrimination-knowledge dataset. S33. For the low-confidence classification results in the aligned discriminant-knowledge dataset, call the knowledge graph reasoning engine to perform reasoning correction based on the kinship rules of the evolutionary tree and the matching degree rules of genetic markers to obtain the corrected variety discrimination results. S34. Based on the corrected variety discrimination results, trace the role path of each modal feature in the classification decision, generate a variety feature attribution map, and clarify the contribution ratio of each modal feature. S35. Use visualization tools to transform the variety characteristic attribution map and discrimination process into a visualization interface, intuitively showing the impact of different modal characteristics on the final discrimination result, and realizing a visual explanation of the discrimination process.

8. The method for identifying jujube varieties based on artificial intelligence according to claim 1, characterized in that, The S4 includes: S41. Construct a blockchain-based trusted traceability system, design the blockchain's storage structure, and use asymmetric encryption algorithms to set data access permissions; S42. Collect complete life data of jujube samples, format the data according to the blockchain storage structure, generate a standardized life dataset, and associate it with a unique sample identifier; S43. Upload the standardized life dataset to the blockchain system, complete the on-chain storage of data through the consensus mechanism, and form an immutable life data chain. S44. Call the blockchain smart contract to generate a unique digital fingerprint for the sample based on the hash value of the standardized life data, and bind the digital fingerprint with the sample identifier as a trusted credential for the sample's identity. S45. Activate the distributed verification mechanism of blockchain nodes. Blockchain nodes in different regions load the cross-regional jujube variety identification results and corresponding life data respectively, cross-verify the cross-regional identification results, and output a unified identification result that has passed the verification.

9. The method for identifying jujube varieties based on artificial intelligence according to claim 1, characterized in that, The S5 includes: S51. Collect core data from each stage, standardize the data format and remove redundant information to generate an integrated discrimination dataset. S52. Perform structured analysis on the integrated discriminant dataset to extract core information; S53. Call the preset discrimination report template, fill the core information after structured analysis into the template in the logical order of variety basic information - characteristic attribution analysis - discrimination reliability description - traceability certificate, and generate a preliminary structured jujube variety discrimination report; S54. Verify the preliminary judgment report, check the accuracy of the data, the logical coherence, and the completeness of the information, and correct any deviations or omissions in the report; S55. Output a structured jujube variety identification report that has passed verification.

10. An artificial intelligence-based jujube variety identification system, comprising: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 9.

Images