Trade goods classification and standardization method and system based on intelligent algorithm

Abstract

The present invention relates to the technical field of cargo classification, and in particular to a method and system for classification and standardization of trade goods based on intelligent algorithms. The method comprises the following steps: obtaining historical trade goods images; sampling the outer contours of the historical trade goods images to obtain contour sampling feature points; performing morphological structure analysis on the trade goods images based on the contour sampling feature points to obtain cargo morphological structure data; obtaining multidimensional implicit feature data based on the cargo morphological structure data; mapping the multidimensional implicit feature data to feature spaces dimension by dimension, thereby constructing a multidimensional morphological feature space of the cargo; obtaining implicit feature space distribution data based on the multidimensional morphological feature space of the cargo; identifying the main morphological scale based on the implicit feature space distribution data, thereby obtaining the main morphological scale of the cargo; and training the spatial mapping function based on the implicit feature space distribution data to obtain a nonlinear mapping function of the classification space. The present invention realizes efficient and accurate cargo classification.

Description

Technical Field

[0001] The present invention relates to the technical field of cargo classification, and in particular to a method and system for classifying and standardizing trade cargo based on an intelligent algorithm. Background Art

[0002] In the modern globalized trade environment, the classification and standardization of trade goods are important links to ensure the smooth progress of trade. With the continuous expansion of trade scale and the increasing diversification of trade methods, traditional methods of goods classification and standardization face many challenges and limitations. Traditional methods usually rely on human experience and manual classification, and have problems such as inaccurate classification and low efficiency. They cannot meet the needs of large-scale trade data processing and rapidly changing trade environment. Therefore, an intelligent method of trade goods classification and standardization is needed. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a trade goods classification and standardization method and system based on intelligent algorithm to solve at least one of the above technical problems.

[0004] To achieve the above objectives, the present invention provides a method for classifying and standardizing trade goods based on an intelligent algorithm, comprising the following steps:

[0005] Step S1: Acquire a historical trade goods image; perform outer contour sampling on the historical trade goods image to obtain contour sampling feature points; perform morphological structure analysis on the trade goods image based on the contour sampling feature points to obtain goods morphological structure data;

[0006] Step S2: obtaining multidimensional implicit feature data based on the cargo morphological structure data; mapping the multidimensional implicit feature data dimension by dimension to construct a cargo multidimensional morphological feature space;

[0007] Step S3: Obtain implicit feature space distribution data based on the multi-dimensional morphological feature space of the goods; perform morphological main scale recognition based on the implicit feature space distribution data, thereby obtaining the morphological main scale of the goods; perform spatial mapping function training based on the implicit feature space distribution data, thereby obtaining a classification space nonlinear mapping function;

[0008] Step S4: classifying the cargo form based on the main scale of the cargo form to obtain multi-dimensional implicit feature classification data; using the classification space nonlinear mapping function to perform classification space mapping on the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; performing classification area boundary analysis based on the multi-dimensional implicit feature classification space to construct an intelligent form classification map;

[0009] Step S5: Obtain the trade goods category attributes; obtain the goods fine-grained attribute classification data based on the trade goods category attributes; perform cluster analysis on the goods fine-grained attribute classification data, and construct attribute standardization classification indicators;

[0010] Step S6: Obtain the image of the trade goods to be classified; perform standardized classification on the image of the trade goods to be classified based on the attribute standardized classification index and the intelligent morphological classification map to obtain the goods classification result; input the goods classification result into the morphological-attribute classification engine for classification recognition optimization to obtain the intelligent classification optimization strategy to execute the classification and standardization of trade goods.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Acquire historical trade goods images;

[0013] Step S12: performing edge feature recognition on the historical trade goods image to obtain the goods contour line;

[0014] Step S13: sampling the outer contour of the cargo contour line to obtain contour sampling feature points;

[0015] Step S14: performing image segmentation on the trade goods image based on the contour sampling feature points to obtain a goods area map;

[0016] Step S15: Performing morphological structure analysis on the cargo area map to obtain cargo morphological structure data.

[0017] Preferably, step S2 comprises the following steps:

[0018] Step S21: performing potential dimensional decomposition on the cargo morphological structure data to obtain multi-dimensional implicit feature data; the multi-dimensional implicit feature data includes the cargo surface curvature, curvature extreme point and shape moment;

[0019] Step S22: performing local compression encoding on the multi-dimensional implicit feature data to obtain a multi-dimensional implicit feature vector;

[0020] Step S23: Map the multi-dimensional implicit feature vector to the feature space dimension by dimension, so as to construct the multi-dimensional morphological feature space of the goods.

[0021] Preferably, the specific steps of step S21 are:

[0022] Construct a three-dimensional geometric model of the cargo based on the cargo morphological structure data;

[0023] Calculate the surface curvature of the three-dimensional geometric model of the cargo to obtain the surface curvature of the cargo;

[0024] Perform curvature feature analysis based on the curvature of the cargo surface to extract the curvature extreme points;

[0025] The coordinates of the extreme point of curvature are obtained based on the extreme point of curvature;

[0026] The second-order contour feature points are calculated based on the coordinates of the curvature extreme point to obtain shape moments; the shape moments include zero-order shape moments, first-order shape moments and second-order central moments.

[0027] Preferably, the specific steps of step S3 are:

[0028] Step S31: performing multi-directional characteristic value distribution analysis on the multi-dimensional morphological characteristic space of the goods to obtain implicit characteristic space distribution data;

[0029] Step S32: quantifying the morphological difference based on the implicit feature space distribution data to obtain the quantification degree of the morphological difference;

[0030] Step S33: identifying the main morphological dimensions of the cargo morphological structure data based on the quantified degree of morphological differences, thereby obtaining the main morphological dimensions of the cargo;

[0031] Step S34: Performing spatial mapping function training based on implicit feature spatial distribution data to obtain a classification space nonlinear mapping function.

[0032] Preferably, the specific steps of step S4 are:

[0033] Step S41: classifying the multidimensional implicit feature data according to the cargo form based on the cargo form main scale to obtain multidimensional implicit feature classification data;

[0034] Step S42: using a classification space nonlinear mapping function to perform classification space mapping on the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space;

[0035] Step S43: calculating the distribution density of feature points in the multi-dimensional implicit feature classification space to obtain the implicit feature distribution density;

[0036] Step S44: performing classification region boundary analysis based on implicit feature distribution density to obtain implicit feature classification boundaries;

[0037] Step S45: Fit the knowledge graph to the multi-dimensional implicit feature classification space based on the implicit feature classification boundary to construct an intelligent morphological classification graph.

[0038] Preferably, the specific steps of step S44 are:

[0039] Performing density level analysis on implicit feature distribution density based on a preset feature distribution density threshold to obtain a feature distribution density level;

[0040] Based on the feature distribution density level, the multi-dimensional implicit feature classification space is segmented to generate multiple feature classification regions;

[0041] Perform cross-border recognition on multiple feature classification areas and locate the area boundary lines;

[0042] Boundary smoothing is performed based on the region boundary line to obtain the implicit feature classification boundary.

[0043] Preferably, the specific steps of step S5 are:

[0044] Step S51: Acquire trade goods category attributes; perform fine-grained attribute classification on the trade goods category attributes to obtain goods fine-grained attribute classification data;

[0045] Step S52: performing semantic feature encoding on the fine-grained attribute classification data of the goods to obtain attribute classification semantic encoding;

[0046] Step S53: performing cluster analysis on the attribute classification semantic coding to obtain the main category and subcategory of the goods attributes;

[0047] Step S54: construct attribute standardization classification indicators based on the main categories and subcategories of cargo attributes.

[0048] Preferably, the specific steps of step S6 are:

[0049] Step S61: Acquire the image of trade goods to be classified;

[0050] Step S62: Optimizing the attribute classification of the intelligent morphological classification map based on the attribute standardization classification index to obtain a morphological-attribute classification engine;

[0051] Step S63: using a morphology-attribute classification engine to perform standardized classification on the trade goods image to be classified, so as to obtain a goods classification result;

[0052] Step S64: Input the cargo classification results into the morphology-attribute classification engine for classification recognition optimization to obtain an intelligent classification optimization strategy to perform trade cargo classification and standardization.

[0053] In this specification, a trade goods classification and standardization system based on intelligent algorithm is provided, which is used to execute the trade goods classification and standardization method based on intelligent algorithm as described above, including:

[0054] The morphological structure module obtains the historical trade goods image; performs outer contour sampling on the historical trade goods image to obtain contour sampling feature points; performs morphological structure analysis on the trade goods image based on the contour sampling feature points to obtain goods morphological structure data;

[0055] The implicit feature module obtains multi-dimensional implicit feature data based on the cargo morphological structure data; maps the multi-dimensional implicit feature data to feature space dimension by dimension, thereby constructing a multi-dimensional morphological feature space of the cargo;

[0056] The spatial distribution module obtains implicit feature spatial distribution data based on the multi-dimensional morphological feature space of the goods; identifies the main morphological scale based on the implicit feature spatial distribution data, thereby obtaining the main morphological scale of the goods; and trains the spatial mapping function based on the implicit feature spatial distribution data to obtain the nonlinear mapping function of the classification space;

[0057] The classification atlas module classifies the cargo form based on the main scale of the cargo form to obtain multi-dimensional implicit feature classification data; uses the classification space nonlinear mapping function to map the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; performs classification area boundary analysis based on the multi-dimensional implicit feature classification space to construct an intelligent form classification atlas;

[0058] Standardized classification module, obtains the attributes of trade goods categories; obtains fine-grained attribute classification data of goods based on the attributes of trade goods categories; performs cluster analysis on the fine-grained attribute classification data of goods, and constructs attribute standardized classification indicators;

[0059] The classification optimization module obtains the images of trade goods to be classified; performs standardized classification on the images of trade goods to be classified based on attribute standardized classification indicators and intelligent morphological classification maps to obtain the goods classification results; inputs the goods classification results into the morphological-attribute classification engine for classification recognition optimization to obtain an intelligent classification optimization strategy to execute trade goods classification and standardization.

[0060] The present invention has the following beneficial effects: by acquiring historical trade goods images, a sample library containing a variety of goods forms can be established, providing a basis for subsequent classification and standardization; outer contour sampling and morphological structure analysis make it possible to extract the morphological structure characteristics of goods, further strengthening the basic data for goods classification and standardization; the acquisition of multidimensional implicit feature data makes it possible to fully grasp the multiple feature dimensions of goods form, improving the accuracy of classification and standardization; constructing a multidimensional morphological feature space of goods helps to form a more comprehensive and detailed feature representation, providing richer information for subsequent morphological classification; the acquisition of implicit feature space distribution data helps to understand the distribution of different features in the feature space, providing a more accurate basis for morphological classification; the identification of the main morphological scale makes it possible to find the most representative and distinguishing scale information in the goods form; the training of the classification space nonlinear mapping function helps to map the multidimensional features to a space that is more suitable for classification, improving the classification accuracy; the cargo morphological classification is based on the main morphological scale and the multidimensional implicit feature classification data. According to the data, goods can be effectively classified according to their morphological characteristics. Classification space mapping and classification area boundary analysis are helpful to map multi-dimensional implicit feature classification data into a more interpretable and distinguishable classification space, and to construct an intelligent morphological classification map, which provides more intuitive and understandable results for goods classification. The acquisition of trade goods category attributes enables more refined classification and standardization according to the category to which the goods belong. The generation of fine-grained attribute classification data of goods helps to classify goods more accurately according to their specific attributes. Cluster analysis can aggregate goods with similar attributes and establish attribute standardization classification indicators to provide a basis for subsequent classification and standardization. Based on attribute standardization classification indicators and intelligent morphological classification maps, standardized classification of images of classified trade goods can quickly and accurately classify them into corresponding categories and attributes. The classification recognition optimization of the morphological-attribute classification engine can further improve the classification accuracy. Through intelligent classification optimization strategies, the classification results are optimized and adjusted to meet the needs of trade goods classification and standardization. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 A schematic diagram of the steps of a method for classifying and standardizing trade goods based on an intelligent algorithm of the present invention;

[0062] Figure 2 Detailed implementation flow chart of step S1;

[0063] Figure 3 Detailed implementation flow chart of step S2;

[0064] Figure 4 Detailed implementation flow chart of step S3. DETAILED DESCRIPTION

[0065] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0066] The example of this application provides a method and system for classifying and standardizing trade goods based on intelligent algorithms. The execution subjects of the method and system for classifying and standardizing trade goods based on intelligent algorithms include but are not limited to: mechanical equipment, data processing platform, cloud server node, network upload device, etc. equipped with the system can be regarded as the general computing node of this application, and the data processing platform includes but is not limited to: at least one of an audio and image management system, an information management system, and a cloud data management system.

[0067] In the embodiment of the present invention, refer to Figure 1 , is a schematic flow chart of the steps of a trade goods classification and standardization method based on an intelligent algorithm of the present invention. In this example, the steps of the trade goods classification and standardization method based on an intelligent algorithm include:

[0068] Step S1: Acquire a historical trade goods image; perform outer contour sampling on the historical trade goods image to obtain contour sampling feature points; perform morphological structure analysis on the trade goods image based on the contour sampling feature points to obtain goods morphological structure data;

[0069] In this embodiment, after obtaining authorization, the cargo image data of completed transactions are obtained from the historical trade records, each cargo image is preprocessed, and image enhancement operations such as normalization and denoising are performed to improve the image quality. The Canny edge detection or GrabCut segmentation algorithm is applied to extract the image contour, and the contour is sampled to obtain contour feature points. The number of sampling points is appropriately set according to the complexity of the contour, generally 50-100 points. The sampled contour point set is morphologically calculated, the fixed points of the contour and the polygonal approximation contour are calculated, the main curvature and secondary curvature information on the contour are extracted, the contour moment and other overall feature parameters are calculated, and after all images are processed, the sampling point coordinates and morphological feature values ​​are organized into a structured data matrix, i.e., the cargo morphological structure data.

[0070] Step S2: obtaining multidimensional implicit feature data based on the cargo morphological structure data; mapping the multidimensional implicit feature data dimension by dimension to construct a cargo multidimensional morphological feature space;

[0071] In this embodiment, a large amount of high-quality three-dimensional scanning data or CAD model data of different types of goods are collected, and a series of implicit morphological features are extracted from these three-dimensional data using computer vision and geometric processing technology. The value range of the feature dimension is divided into several discrete intervals, and the frequency of each product falling into different intervals on the feature dimension is counted to form a one-dimensional histogram feature. These histogram features are projected onto a one-dimensional coordinate axis to construct a one-dimensional feature space of the feature dimension. The above steps are repeated for all morphological feature dimensions to finally obtain a multidimensional feature space. In this multidimensional feature space, each product corresponds to a specific coordinate point, reflecting its comprehensive performance in various morphological features.

[0072] Step S3: Obtain implicit feature space distribution data based on the multi-dimensional morphological feature space of the goods; perform morphological main scale recognition based on the implicit feature space distribution data, thereby obtaining the morphological main scale of the goods; perform spatial mapping function training based on the implicit feature space distribution data, thereby obtaining a classification space nonlinear mapping function;

[0073] In this embodiment, clustering algorithms such as k-means, DBSCAN, etc. are used to automatically identify dense clustering areas in the feature space, analyze the statistical characteristics of each cluster, such as the center position, variance, covariance, etc., understand the distribution of implicit features in space, and use visualization tools such as PCA, t-SNE and other dimensionality reduction technologies to intuitively observe the overall distribution of the feature space. Based on the analysis of the distribution of the implicit feature space, the main feature dimensions that contribute most to the difference in the shape of the goods are identified: the contribution rate of each feature dimension in the overall variance is calculated, and the dimension with a higher contribution rate is selected as the main scale of the shape. The main scale dimension in the feature space is analyzed. The distribution characteristics of the goods, such as cluster centers, variance, covariance, etc., are highlighted in the visualized feature space to observe its relationship with the overall morphological distribution. Based on the distribution of the multidimensional morphological feature space, a nonlinear mapping function is trained to divide the feature space into several category areas, each representing a typical cargo morphology. Nonlinear models such as neural networks or kernel methods are used to learn the mapping function from the high-dimensional feature space to the classification space. During the training process, the cluster distribution information in the feature space is used to guide the model to learn a reasonable nonlinear mapping, and finally a nonlinear function that can map any cargo morphological features to the classification space is obtained.

[0074] Step S4: classifying the cargo form based on the main scale of the cargo form to obtain multi-dimensional implicit feature classification data; using the classification space nonlinear mapping function to perform classification space mapping on the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; performing classification area boundary analysis based on the multi-dimensional implicit feature classification space to construct an intelligent form classification map;

[0075] In this embodiment, according to the distribution of the main scale feature, the appropriate number of classifications and boundary thresholds are determined. For all cargo samples, they are divided into different morphological categories according to their main scale feature values. Through this main scale-based classification method, the morphological category label of each cargo sample is obtained, and the multi-dimensional implicit feature data is mapped to the classification space: for each cargo sample, its multi-dimensional implicit feature vector is input into the nonlinear mapping function, and the function outputs the corresponding classification space coordinates, indicating the position of the cargo in the classification space. Through this nonlinear mapping, the high-dimensional feature space is projected into the low-dimensional classification space, and cluster analysis is performed on the cargo samples mapped to the classification space to identify the regional boundaries of different morphological categories. Clustering algorithms such as K-Means and DBSCAN are used to automatically divide the regional clusters in the classification space, analyze the statistical characteristics of each regional cluster, such as the center position, variance, boundary, etc., describe its morphological characteristics, visualize the regional boundaries in the classification space, and construct an intelligent morphological classification map.

[0076] Step S5: Obtain the trade goods category attributes; obtain the goods fine-grained attribute classification data based on the trade goods category attributes; perform cluster analysis on the goods fine-grained attribute classification data, and construct attribute standardization classification indicators;

[0077] In this embodiment, detailed attribute information of various types of trade goods is collected, including basic information: name, code, brand, specification model, etc., material composition: material, textile composition, chemical composition, etc., functional attributes: purpose, scope of application, functional characteristics, etc., quality indicators: size, weight, capacity, strength, etc. A comprehensive cargo category attribute system is sorted out, covering the characteristic indicators of various categories. According to the granularity and universality of the attributes, they are divided into classification indicators of different levels. For example, material, purpose, size, etc. are first-level indicators, and specific ingredients, functions, and parameters are second-level indicators. Each cargo sample is marked with its attribution in each level of attribute classification. Clustering algorithms such as K-Means and DBSCAN are used to automatically aggregate similar attribute values ​​into standardized classifications. The statistical characteristics of each classification cluster are analyzed to determine its representative value range and standard. The classification results are sorted into a complete set of standardized classification systems for cargo attributes.

[0078] Step S6: Obtain the image of the trade goods to be classified; perform standardized classification on the image of the trade goods to be classified based on the attribute standardized classification index and the intelligent morphological classification map to obtain the goods classification result; input the goods classification result into the morphological-attribute classification engine for classification recognition optimization to obtain the intelligent classification optimization strategy to execute the classification and standardization of trade goods.

[0079] In this embodiment, image data of trade goods that need to be standardized and classified is obtained, and the goods image is input into the morphological classification engine. The morphological category to which it belongs is determined according to the morphological features. Then, according to the morphological category, combined with the attribute standardized classification index, the specific category of the goods is further determined. Through this dual judgment of attributes and morphology, the final standardized classification result of the goods is obtained, and the classification result is fed back to the morphological-attribute classification engine for optimization learning. Errors or abnormal situations that occur during the classification process are analyzed to understand their causes, and the classification rules and algorithm parameters are adjusted to optimize the performance of morphological classification and attribute association, continuously accumulate classification samples, and continuously improve the morphological classification map and attribute standard system.

[0080] In this embodiment, refer to Figure 2 , is a flowchart of detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0081] Step S11: Acquire historical trade goods images;

[0082] Step S12: performing edge feature recognition on the historical trade goods image to obtain the goods contour line;

[0083] Step S13: sampling the outer contour of the cargo contour line to obtain contour sampling feature points;

[0084] Step S14: performing image segmentation on the trade goods image based on the contour sampling feature points to obtain a goods area map;

[0085] Step S15: Performing morphological structure analysis on the cargo area map to obtain cargo morphological structure data.

[0086] In this embodiment, after obtaining authorization, the image data of the goods of the completed transaction is obtained from the historical trade records, and the computer vision technology, such as the edge detection algorithm, is used to extract the clear contour line from the goods image: the edge detection algorithm such as Canny and Sobel is applied to identify the edge contour of the goods in the image, the extracted contour line is smoothed, the noise and burrs are removed, and a stable contour of the goods is obtained, and the contour line of the goods is uniformly sampled to obtain a series of feature points describing the contour shape: a number of feature points are selected at equal intervals along the contour line, and their coordinate positions are recorded. By adjusting the sampling density, the number of feature points and the expression accuracy of the contour shape can be controlled, and the contour sampling feature can be used to obtain the feature points. Points are used to segment the cargo image and extract the cargo area: the cargo area is segmented from the image according to the contour described by the feature points through methods such as polygon fitting or image masking, the boundary of the segmented cargo area is smoothed, noise and burrs are removed, and a stable cargo area is obtained. The cargo area image is deeply analyzed in geometry to extract rich structural features: the basic size features of the cargo area, such as length, width, area, etc., are calculated, the contour curve of the cargo area is analyzed, shape features such as curvature and bending are extracted, structural features such as holes and depressions inside the cargo area are detected, and their topological properties are described. The various morphological structural features extracted are integrated into a comprehensive cargo morphological dataset.

[0087] In this embodiment, refer to Figure 3 , is a flowchart of detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0088] Step S21: performing potential dimensional decomposition on the cargo morphological structure data to obtain multi-dimensional implicit feature data; the multi-dimensional implicit feature data includes the cargo surface curvature, curvature extreme point and shape moment;

[0089] Step S22: performing local compression encoding on the multi-dimensional implicit feature data to obtain a multi-dimensional implicit feature vector;

[0090] Step S23: Map the multi-dimensional implicit feature vector to the feature space dimension by dimension, so as to construct the multi-dimensional morphological feature space of the goods.

[0091] In this embodiment, for the geometric contour of the cargo area, its surface curvature and curvature extreme point and other features are calculated, based on the basic features such as the size and area of ​​the cargo area, high-order geometric features such as shape moments are extracted, and the above multi-dimensional morphological structural features are locally compressed and encoded to obtain a compact feature vector representation: the autoencoder, principal component analysis and other technologies are used to reduce the dimension of each feature dimension to ensure that the high-dimensional feature is compressed into a more concise feature vector while retaining sufficient information, and the compressed multi-dimensional implicit feature vector is used as the digital expression of the cargo morphology, and the above multi-dimensional implicit feature vector is mapped to a high-dimensional feature space to construct the morphological feature space of the cargo: the multi-dimensional feature vector of each cargo sample is converted to a high-dimensional feature space through linear or nonlinear mapping. In this high-dimensional feature space, each cargo sample corresponds to a unique position coordinate point, and the high-dimensional feature space is projected into a low-dimensional space for observation through visualization technology, such as PCA, t-SNE, etc.

[0092] In this embodiment, the specific steps of step S21 are:

[0093] Construct a three-dimensional geometric model of the cargo based on the cargo morphological structure data;

[0094] Calculate the surface curvature of the three-dimensional geometric model of the cargo to obtain the surface curvature of the cargo;

[0095] Perform curvature feature analysis based on the curvature of the cargo surface to extract the curvature extreme points;

[0096] The coordinates of the extreme point of curvature are obtained based on the extreme point of curvature;

[0097] The second-order contour feature points are calculated based on the coordinates of the curvature extreme point to obtain shape moments; the shape moments include zero-order shape moments, first-order shape moments and second-order central moments.

[0098] In this embodiment, contour lines and area information are extracted from the segmented cargo area image, and a 3D model is constructed using a 3D reconstruction algorithm (such as a voxel method, a grid method, etc.) in combination with the size data of the cargo. The constructed 3D model is smoothed to eliminate noise and topological errors to obtain a high-quality geometric model. The principal curvature value of each point on the surface of the 3D model is calculated using a differential geometry method. The convex and concave characteristics and the degree of curvature of the surface are determined by analyzing the magnitude and sign of the principal curvature. The surface curvature field is analyzed and derived to find feature points where the curvature value reaches a maximum or minimum. These curvature extreme value points often correspond to significant morphological structures such as convexities and concavities on the cargo surface. Based on the position coordinates of the curvature extreme value points, the zero-order shape moment (area) and the first-order shape moment (center of mass) are calculated, and the second-order central moment is further calculated to describe the characteristics of the cargo shape such as the degree of expansion and the degree of distortion. The calculated shape moments of each order are saved as a statistical feature description of the cargo shape.

[0099] In this embodiment, refer to Figure 4 , is a flowchart of detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0100] Step S31: performing multi-directional characteristic value distribution analysis on the multi-dimensional morphological characteristic space of the goods to obtain implicit characteristic space distribution data;

[0101] Step S32: quantifying the morphological difference based on the implicit feature space distribution data to obtain the quantification degree of the morphological difference;

[0102] Step S33: identifying the main morphological dimensions of the cargo morphological structure data based on the quantified degree of morphological differences, thereby obtaining the main morphological dimensions of the cargo;

[0103] Step S34: Performing spatial mapping function training based on implicit feature spatial distribution data to obtain a classification space nonlinear mapping function.

[0104] In this embodiment, the distribution of eigenvalues ​​in the feature space is observed from different angles (such as the direction of the principal component, the direction of the standard deviation, etc.). The statistical characteristics of the distribution of eigenvalues, such as the central tendency, degree of dispersion, and skew direction, are analyzed. The distance measurement between any two cargo samples in the feature space is calculated as a quantitative indicator of their morphological differences. Methods such as Euclidean distance and cosine similarity can be used to select a suitable measurement method according to actual needs. The morphological difference quantification matrix is ​​analyzed to find the feature dimensions that contribute most to the difference. The morphological structural features corresponding to these dominant feature dimensions are the main scales of the cargo morphology. Using technologies such as deep learning and samples distributed in the feature space, a complex mapping relationship from morphological features to classification space is fitted. This mapping function can map the original high-dimensional morphological features to a low-dimensional space that is more conducive to classification.

[0105] In this embodiment, step S4 includes the following steps:

[0106] Step S41: classifying the multidimensional implicit feature data according to the cargo form based on the cargo form main scale to obtain multidimensional implicit feature classification data;

[0107] Step S42: using a classification space nonlinear mapping function to perform classification space mapping on the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space;

[0108] Step S43: calculating the distribution density of feature points in the multi-dimensional implicit feature classification space to obtain the implicit feature distribution density;

[0109] Step S44: performing classification region boundary analysis based on implicit feature distribution density to obtain implicit feature classification boundaries;

[0110] Step S45: Fit the knowledge graph to the multi-dimensional implicit feature classification space based on the implicit feature classification boundary to construct an intelligent morphological classification graph.

[0111] In this embodiment, the key feature dimensions corresponding to the main scale of the morphology are selected as the basis for classification. According to the value range of these main scale features, the cargo samples are divided into different morphological categories. The multidimensional implicit feature classification data is mapped to the classification space using the previously trained classification space nonlinear mapping function. The multidimensional implicit features of each cargo sample are input and converted to the classification space coordinates through the mapping function. In this way, each cargo sample has a corresponding coordinate point in the classification space. The data density of each area in the classification space is calculated by using methods such as kernel density estimation and Gaussian mixture. The calculated implicit feature distribution density information is saved to provide a basis for boundary analysis, find out the area with large density gradient changes, and determine the dividing line between categories. By adjusting the density threshold, classification boundaries of different granularities can be obtained. According to the geometric characteristics of the classification boundary, the boundary of the morphological category is modeled by polygons, surfaces, etc. In the classification map, the main scale features of the morphology of each category, the implicit feature distribution and other information are recorded. The constructed intelligent morphological classification map is saved as a knowledge base for subsequent intelligent classification of goods.

[0112] In this embodiment, the specific steps of step S44 are:

[0113] Performing density level analysis on implicit feature distribution density based on a preset feature distribution density threshold to obtain a feature distribution density level;

[0114] Based on the feature distribution density level, the multi-dimensional implicit feature classification space is segmented to generate multiple feature classification regions;

[0115] Perform cross-border recognition on multiple feature classification areas and locate the area boundary lines;

[0116] Boundary smoothing is performed based on the region boundary line to obtain the implicit feature classification boundary.

[0117] In this embodiment, several different density thresholds are set, such as high density, medium density, low density, etc., and each area in the classification space is divided into different density levels according to its implicit feature distribution density. For areas of different density levels, clustering, watershed and other algorithms are used to perform detailed regional division to ensure that the implicit feature distribution density within each area is relatively uniform and there are obvious differences between areas. The above-mentioned multiple feature classification areas are cross-compared, the boundary lines between the various areas are identified, and the significant jump points of the implicit feature distribution density between adjacent areas are found to determine the location of the regional boundary. The boundary line contours of each area are depicted using the coordinate information of the feature points on the regional boundary. The above-mentioned regional boundary lines are smoothed to obtain smoother and more stable implicit feature classification boundaries: curve fitting, polygonal approximation and other methods are used to smooth and optimize the original regional boundary lines to ensure that the smoothed boundary lines can better describe the classification boundaries in the implicit feature classification space.

[0118] In this embodiment, the specific steps of step S5 are:

[0119] Step S51: Acquire trade goods category attributes; perform fine-grained attribute classification on the trade goods category attributes to obtain goods fine-grained attribute classification data;

[0120] Step S52: performing semantic feature encoding on the fine-grained attribute classification data of the goods to obtain attribute classification semantic encoding;

[0121] Step S53: performing cluster analysis on the attribute classification semantic coding to obtain the main category and subcategory of the goods attributes;

[0122] Step S54: construct attribute standardization classification indicators based on the main categories and subcategories of cargo attributes.

[0123] In this embodiment, detailed attribute information of goods is collected, such as various detailed characteristics such as material, color, and purpose. According to the specific content of the attribute, expert knowledge or machine learning methods are used to divide it into more detailed attribute categories. According to the semantic content of the attribute category, a unique semantic code is assigned to each attribute category. One-hot encoding, word embedding, etc. can be used to convert the attribute category into a semantic representation in vector form. Cluster analysis is performed on the attribute classification semantic code to extract the main category and subcategory of the goods attribute. K-means, hierarchical clustering, etc. are used to divide the attribute semantic code into different main category clusters. Within each main category cluster, fine-grained clustering is further performed to identify the subcategory of the attribute. Based on the above main category and subcategory information of the goods attribute, a set of standardized attribute classification index system is constructed: a hierarchical attribute classification system is designed, the main category is used as the first-level classification, and the subcategory is used as the second-level classification. A unique number or label is defined for each main category and subcategory as a standardized attribute classification index. The constructed attribute standardized classification index system is saved for subsequent goods attribute management and application.

[0124] In this embodiment, the specific steps of step S6 are:

[0125] Step S61: Acquire the image of trade goods to be classified;

[0126] Step S62: Optimizing the attribute classification of the intelligent morphological classification map based on the attribute standardization classification index to obtain a morphological-attribute classification engine;

[0127] Step S63: using a morphology-attribute classification engine to perform standardized classification on the trade goods image to be classified, so as to obtain a goods classification result;

[0128] Step S64: Input the cargo classification results into the morphology-attribute classification engine for classification recognition optimization to obtain an intelligent classification optimization strategy to perform trade cargo classification and standardization.

[0129] In this embodiment, the morphological classification map and the attribute classification system are integrated to establish a joint classification model that can simultaneously identify the morphology and attributes of goods. The morphological-attribute classification engine is trained and optimized using machine learning or deep learning techniques to have high classification accuracy. The image of the goods to be classified is input into the classification engine to obtain the recognition results of its morphological category and attribute category. According to the classification results, the goods are classified into corresponding morphological categories and attribute categories to obtain standardized classification labels. The correctness of the preliminary classification results is analyzed. According to the correctness of the preliminary classification results, the structure and parameters of the classification model are regressed and updated, and the model is re-optimized on the new training set to generate an iteratively updated intelligent classification strategy.

[0130] In this embodiment, a trade goods classification and standardization system based on an intelligent algorithm is provided, which is used to execute the trade goods classification and standardization method based on an intelligent algorithm as described above, including:

[0131] The morphological structure module obtains the historical trade goods image; performs outer contour sampling on the historical trade goods image to obtain contour sampling feature points; performs morphological structure analysis on the trade goods image based on the contour sampling feature points to obtain goods morphological structure data;

[0132] The implicit feature module obtains multi-dimensional implicit feature data based on the cargo morphological structure data; maps the multi-dimensional implicit feature data to feature space dimension by dimension, thereby constructing a multi-dimensional morphological feature space of the cargo;

[0133] The spatial distribution module obtains implicit feature spatial distribution data based on the multi-dimensional morphological feature space of the goods; identifies the main morphological scale based on the implicit feature spatial distribution data, thereby obtaining the main morphological scale of the goods; and trains the spatial mapping function based on the implicit feature spatial distribution data to obtain the nonlinear mapping function of the classification space;

[0134] The classification atlas module classifies the cargo form based on the main scale of the cargo form to obtain multi-dimensional implicit feature classification data; uses the classification space nonlinear mapping function to map the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; performs classification area boundary analysis based on the multi-dimensional implicit feature classification space to construct an intelligent form classification atlas;

[0135] Standardized classification module, obtains the attributes of trade goods categories; obtains fine-grained attribute classification data of goods based on the attributes of trade goods categories; performs cluster analysis on the fine-grained attribute classification data of goods, and constructs attribute standardized classification indicators;

[0136] The classification optimization module obtains the images of trade goods to be classified; performs standardized classification on the images of trade goods to be classified based on attribute standardized classification indicators and intelligent morphological classification maps to obtain the goods classification results; inputs the goods classification results into the morphological-attribute classification engine for classification recognition optimization to obtain an intelligent classification optimization strategy to execute trade goods classification and standardization.

[0137] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0138] The above is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A method for classifying and standardizing trade goods based on intelligent algorithms, characterized in that: The following steps are involved: Step S1: Acquire a historical trade goods image; perform outer contour sampling on the historical trade goods image to obtain contour sampling feature points; perform morphological structure analysis on the trade goods image based on the contour sampling feature points to obtain goods morphological structure data; Step S2: obtaining multidimensional implicit feature data based on the cargo morphological structure data; mapping the multidimensional implicit feature data dimension by dimension to construct a cargo multidimensional morphological feature space; Step S3: Obtain implicit feature space distribution data based on the multi-dimensional morphological feature space of the goods; perform morphological main scale recognition based on the implicit feature space distribution data, thereby obtaining the morphological main scale of the goods; perform spatial mapping function training based on the implicit feature space distribution data, so as to obtain a classification space nonlinear mapping function; the specific steps of step S3 are: Step S31: performing multi-directional characteristic value distribution analysis on the multi-dimensional morphological characteristic space of the goods to obtain implicit characteristic space distribution data; Step S32: quantifying the morphological difference based on the implicit feature space distribution data to obtain the quantification degree of the morphological difference; Step S33: identifying the main morphological scale of the cargo morphological structure data based on the morphological difference quantification degree, thereby obtaining the main morphological scale of the cargo; the main morphological scale of the cargo is specifically: analyzing the morphological difference quantification matrix to find out the characteristic dimensions that contribute most to the difference, and the morphological structural features corresponding to these dominant characteristic dimensions are the main scales of the cargo morphology; Step S34: performing spatial mapping function training based on implicit feature spatial distribution data to obtain a classification space nonlinear mapping function; Step S4: classifying the cargo form based on the cargo form main dimensions to obtain multi-dimensional implicit feature classification data; The classification space nonlinear mapping function is used to map the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; Based on the multi-dimensional implicit feature classification space, the classification area boundary analysis is carried out to construct an intelligent morphological classification map; Step S5: Obtain the trade goods category attributes; obtain the goods fine-grained attribute classification data based on the trade goods category attributes; perform cluster analysis on the goods fine-grained attribute classification data, and construct attribute standardization classification indicators; Step S6: Acquire the image of trade goods to be classified; Based on the attribute standardization classification index and the intelligent morphology classification map, the images of the trade goods to be classified are standardized and classified to obtain the goods classification results; the goods classification results are input into the morphology-attribute classification engine for classification recognition optimization to obtain the intelligent classification optimization strategy to perform trade goods classification and standardization; the specific steps of step S6 are: Step S61: Acquire the image of trade goods to be classified; Step S62: Optimizing the attribute classification of the intelligent morphological classification map based on the attribute standardization classification index to obtain a morphological-attribute classification engine; The morphology-attribute classification engine specifically comprises: integrating the morphology classification map and the attribute classification system to establish a joint classification model that can simultaneously identify the morphology and attributes of goods, and using machine learning or deep learning and other technologies to train and optimize the morphology-attribute classification engine; Step S63: using a morphology-attribute classification engine to perform standardized classification on the trade goods image to be classified, so as to obtain a goods classification result; Step S64: Input the cargo classification results into the morphology-attribute classification engine for classification recognition optimization to obtain an intelligent classification optimization strategy to perform trade cargo classification and standardization.

2. The method for classifying and standardizing trade goods based on intelligent algorithms according to claim 1 is characterized in that: The specific steps of step S1 are: Step S11: Acquire historical trade goods images; Step S12: performing edge feature recognition on the historical trade goods image to obtain the goods contour line; Step S13: sampling the outer contour of the cargo contour line to obtain contour sampling feature points; Step S14: performing image segmentation on the trade goods image based on the contour sampling feature points to obtain a goods area map; Step S15: Performing morphological structure analysis on the cargo area map to obtain cargo morphological structure data.

3. The method for classifying and standardizing trade goods based on intelligent algorithms according to claim 1 is characterized in that: The specific steps of step S2 are: Step S21: performing potential dimensional decomposition on the cargo morphological structure data to obtain multi-dimensional implicit feature data; the multi-dimensional implicit feature data includes the cargo surface curvature, curvature extreme point and shape moment; Step S22: performing local compression encoding on the multi-dimensional implicit feature data to obtain a multi-dimensional implicit feature vector; Step S23: Map the multi-dimensional implicit feature vector to the feature space dimension by dimension, so as to construct the multi-dimensional morphological feature space of the goods.

4. The method for classifying and standardizing trade goods based on intelligent algorithms according to claim 3 is characterized in that: The specific steps of step S21 are: Construct a three-dimensional geometric model of the cargo based on the cargo morphological structure data; Calculate the surface curvature of the three-dimensional geometric model of the cargo to obtain the surface curvature of the cargo; Perform curvature feature analysis based on the curvature of the cargo surface to extract the curvature extreme points; The coordinates of the extreme point of curvature are obtained based on the extreme point of curvature; The second-order contour feature points are calculated based on the coordinates of the curvature extreme point to obtain shape moments; the shape moments include zero-order shape moments, first-order shape moments and second-order central moments.

5. The method for classifying and standardizing trade goods based on intelligent algorithms according to claim 1 is characterized in that: The specific steps of step S4 are: Step S41: classifying the multidimensional implicit feature data according to the cargo form based on the cargo form main scale to obtain multidimensional implicit feature classification data; Step S42: using a classification space nonlinear mapping function to perform classification space mapping on the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; Step S43: calculating the distribution density of feature points in the multi-dimensional implicit feature classification space to obtain the implicit feature distribution density; Step S44: performing classification region boundary analysis based on implicit feature distribution density to obtain implicit feature classification boundaries; Step S45: Fit the knowledge graph to the multi-dimensional implicit feature classification space based on the implicit feature classification boundary to construct an intelligent morphological classification graph.

6. The method for classifying and standardizing trade goods based on intelligent algorithms according to claim 5 is characterized in that: The specific steps of step S44 are: Performing density level analysis on implicit feature distribution density based on a preset feature distribution density threshold to obtain a feature distribution density level; Based on the feature distribution density level, the multi-dimensional implicit feature classification space is segmented to generate multiple feature classification regions; Perform cross-border recognition on multiple feature classification areas and locate the area boundary lines; Boundary smoothing is performed based on the region boundary line to obtain the implicit feature classification boundary.

7. The method for classifying and standardizing trade goods based on intelligent algorithms according to claim 1 is characterized in that: The specific steps of step S5 are: Step S51: Acquire trade goods category attributes; perform fine-grained attribute classification on the trade goods category attributes to obtain goods fine-grained attribute classification data; Step S52: performing semantic feature encoding on the fine-grained attribute classification data of the goods to obtain attribute classification semantic encoding; Step S53: performing cluster analysis on the attribute classification semantic coding to obtain the main category and subcategory of the goods attributes; Step S54: construct attribute standardization classification indicators based on the main categories and subcategories of cargo attributes.

8. A trade goods classification and standardization system based on intelligent algorithms, characterized in that: Used to implement the trade goods classification and standardization method based on intelligent algorithm as claimed in claim 1, comprising: The morphological structure module obtains the historical trade goods image; performs outer contour sampling on the historical trade goods image to obtain contour sampling feature points; performs morphological structure analysis on the trade goods image based on the contour sampling feature points to obtain goods morphological structure data; The implicit feature module obtains multi-dimensional implicit feature data based on the cargo morphological structure data; maps the multi-dimensional implicit feature data to feature space dimension by dimension, thereby constructing a multi-dimensional morphological feature space of the cargo; The spatial distribution module obtains implicit feature spatial distribution data based on the multi-dimensional morphological feature space of the goods; identifies the main morphological scale based on the implicit feature spatial distribution data, thereby obtaining the main morphological scale of the goods; and trains the spatial mapping function based on the implicit feature spatial distribution data to obtain the nonlinear mapping function of the classification space; The classification atlas module classifies the cargo form based on the main scale of the cargo form to obtain multi-dimensional implicit feature classification data; uses the classification space nonlinear mapping function to map the multi-dimensional implicit feature classification data to obtain a multi-dimensional implicit feature classification space; performs classification area boundary analysis based on the multi-dimensional implicit feature classification space to construct an intelligent form classification atlas; Standardized classification module, obtains the attributes of trade goods categories; obtains fine-grained attribute classification data of goods based on the attributes of trade goods categories; performs cluster analysis on the fine-grained attribute classification data of goods, and constructs attribute standardized classification indicators; The classification optimization module obtains the images of trade goods to be classified; performs standardized classification on the images of trade goods to be classified based on attribute standardized classification indicators and intelligent morphological classification maps to obtain the goods classification results; inputs the goods classification results into the morphological-attribute classification engine for classification recognition optimization to obtain an intelligent classification optimization strategy to execute trade goods classification and standardization.

Images