Multi-dimensional code intelligent analysis method and system based on visual information for processing results

Abstract

The present invention discloses a method and system for intelligent analysis of multi-dimensional codes based on processing results of visual information, aiming to simulate the process of rapid understanding and recognition of complex scenes by humans. The method encodes multi-dimensional scene grid information with depth information formed based on the processing results of visual spatiotemporal information. The encoding process uses feature extraction, multi-dimensional code generation and redundancy check multi-dimensional code encoding technology to convert complex scene information into an efficient and compressed encoding form, and store these codes in a database; after the scene is quickly encoded, it is matched with the stored multi-dimensional code information to achieve rapid understanding and accurate recognition of the scene. The present invention provides a new technical framework for intelligent and rapid scene processing through efficient multi-dimensional code encoding and processing methods, improves the scene recognition and understanding capabilities of complex environments, and provides strong technical support for humanoid robots, driverless cars, smart glasses, and inspection equipment.

Description

Technical Field

[0001] The present invention relates to the field of spatiotemporal artificial intelligence technology, and in particular to a multi-dimensional code intelligent parsing method and a multi-dimensional code intelligent parsing system based on visual spatiotemporal information for processing results. Background Art

[0002] With the rapid development of computer vision technology, intelligent scene information processing methods are increasingly being used in intelligent devices in various fields, such as humanoid robots, driverless cars, smart glasses, and inspection equipment. However, with the further development of artificial intelligence, humans have put forward higher requirements for scene recognition and understanding capabilities in complex environments. These devices need to accurately and quickly recognize and understand the surrounding environment in different scenarios in order to make corresponding decisions and actions.

[0003] At present, traditional scene recognition technology mainly relies on two-dimensional image processing methods. Although it can achieve good recognition results under certain conditions, it is often difficult to meet the requirements of real-time and reliability in complex and changeable actual environments. Traditional methods have great limitations in processing image information with depth of field and cannot fully capture the depth of field features in the scene. Although technologies such as image segmentation can segment objects in static images, their performance is often insufficient when processing scenes with similar images.

[0004] In addition, the performance of these technologies under complex lighting conditions is not satisfactory and they are easily affected by changes in lighting in the environment, which affects the accuracy of recognition. In recent years, the development of deep learning has provided a new solution for scene recognition and understanding, namely, automatically extracting high-level features from large-scale data through convolutional neural networks (CNNs). However, deep learning models usually require a lot of computing resources and storage space, and the model training and reasoning process are complex, which has certain limitations for application scenarios with high real-time requirements. In addition, the black box characteristics of deep learning models also make them insufficient in terms of explainability and transparency, which is not conducive to further debugging and optimization. Summary of the invention

[0005] In view of the above problems, the present invention proposes a multi-dimensional code intelligent parsing method and a multi-dimensional code intelligent parsing system based on visual spatiotemporal information of processing results.

[0006] The embodiment of the present invention provides a multi-dimensional code intelligent parsing method for processing results based on visual information, and the multi-dimensional code intelligent parsing method includes:

[0007] The processing result of the input visual spatiotemporal information with depth of field information is efficiently encoded to generate and store scene multi-dimensional code information containing spatiotemporal features, wherein the processing result of the visual spatiotemporal information with depth of field information is multi-dimensional scene grid information;

[0008] The new scene multi-dimensional code information is compared and matched with the stored scene multi-dimensional code information to achieve accurate scene recognition and understanding.

[0009] Optionally, the input visual spatiotemporal information is efficiently encoded to generate scene multi-dimensional code information containing spatiotemporal features, including:

[0010] Extracting key visual features and depth of field information features from the processing results of the visual spatiotemporal information pair;

[0011] Generate the scene multi-dimensional code information according to the key visual features and the depth of field information features;

[0012] Redundant check information is embedded in the scene multi-dimensional code information to ensure data integrity and data resistance.

[0013] Optionally, extracting key visual features and depth information features from the processing result of the visual spatiotemporal information pair includes:

[0014] Extracting key visual feature data of edge, texture, color, grayscale and shape representing the image content from the processing results of the visual spatiotemporal information using visual feature extraction technology, wherein the visual feature extraction technology includes: Canny edge detection algorithm, Gabor filter, color histogram or shape descriptor;

[0015] Depth feature extraction technology is used to extract depth information feature data related to depth information, such as distance, object contour and spatial distribution relationship, from the processing results of the visual spatiotemporal information. The depth feature extraction technology includes: stereo ranging technology.

[0016] Optionally, the scene multi-dimensional code information including a multi-dimensional code based on vision and depth information is constructed according to the key visual feature data and the depth information feature data.

[0017] Optionally, constructing the scene multi-dimensional code information including multi-dimensional coding based on vision and depth information according to the key visual features and the depth information feature data includes:

[0018] Mapping the key visual feature data and the depth information feature data into a multi-dimensional space coordinate system, and constructing scene multi-dimensional code information of different dimensions according to different visual features and different depth features;

[0019] The weight information of the base point pixel block and its surrounding pixel blocks is embedded in the scene multi-dimensional code information. The base point pixel block refers to the intersection of the visual spatiotemporal information and the visual intersection line when observing the target. If the surrounding pixel blocks are closer to the base point pixel block in the same section, the larger the retrieval weight coefficient assigned to them is. If the surrounding pixel blocks are closer to the base point pixel block in the direction from the base point pixel block to the midpoint of the dual camera baseline in different sections, the larger the retrieval weight coefficient assigned to them is. The retrieval weight coefficient increases continuously as the distance from the base point pixel block decreases until it reaches the maximum at the base point pixel block. The scene or target within the spatial range is processed based on the weight information of the base point pixel block and its surrounding pixel blocks to enhance the accuracy of spatial processing and intelligent recognition.

[0020] Optimize the structure and data layout of the scene multi-dimensional code to improve coding efficiency and data compression rate; use compression algorithms to compress feature data to reduce the size of the scene multi-dimensional code, while controlling the proportion of redundant information in the scene multi-dimensional code to balance data integrity and coding efficiency; dynamically adjust the coding method according to the complexity of each feature data to achieve efficient coding;

[0021] Real-time verification is performed during the scene multi-dimensional code generation process to ensure the correctness and stability of the encoding; real-time data verification, integrity verification and damage resistance testing are performed respectively to ensure that the generated scene multi-dimensional code can still be correctly decoded in the event of partial damage.

[0022] Optionally, redundant check information is embedded in the scene multi-dimensional code information to ensure data integrity and data damage resistance, including:

[0023] The redundant check information is embedded in the scene multi-dimensional code information by using a redundant coding method for data verification and recovery, wherein the redundant coding method includes: a parity check code or a hash check code;

[0024] During the decoding process, a verification method is used to detect error information in the scene multi-dimensional code information to ensure that it is not damaged during transmission or storage, and the verification method includes: a cyclic redundancy check;

[0025] If there is erroneous information, the redundant check information and error correction technology are used to correct the detected erroneous information to achieve automatic correction of erroneous data and ensure data integrity and reliability. The error correction technology includes: Hamming code and Reed-Solomon code.

[0026] Optionally, the method for storing the scene multi-dimensional code information includes:

[0027] The scene multi-dimensional code information is stored and retrieved by optimizing the database structure, and the optimized database structure includes index optimization, data partitioning and parallel processing technology; specifically, the query speed is accelerated by using B-tree and hash index technology, the data storage layout is optimized by horizontal and vertical partitioning technology, and the data access performance is improved by using parallel processing technology to ensure a fast response to the scene multi-dimensional code information; and, in the storage process, the multi-dimensional code encoding information is spatially divided, and based on the spatial position information of the base point pixel block, a higher retrieval weight coefficient is given to the data of the pixel block close to the base point, so as to facilitate fast retrieval and processing;

[0028] The scene multi-dimensional code information is stored in a dispersed manner through multiple storage nodes; specifically, when writing data, the scene multi-dimensional code information is distributed to different storage nodes using data sharding and consistent hashing algorithms to improve storage efficiency; when reading data, the target data is quickly located and retrieved through intelligent routing and load balancing strategies, and nodes are allocated based on base point pixel blocks, making retrieval based on spatial range more efficient;

[0029] Local processing and rapid response of the scene multi-dimensional code information are achieved through edge computing nodes; specifically, the scene multi-dimensional code information is stored locally on the edge device, and lightweight database or distributed storage technology is used to reduce the load of the central server and network transmission delay, thereby ensuring storage efficiency and reliability; incremental synchronization and delayed synchronization technology are used to synchronize data between the edge device and the central server, thereby achieving efficient data transmission and synchronization and ensuring seamless collaboration; at the same time, containerization technology is used to dynamically adjust edge computing and storage resources according to data scale and business needs, thereby achieving elastic expansion and automated management of edge computing nodes, thereby ensuring efficient storage and computing performance when the amount of data increases significantly.

[0030] Optionally, comparing and matching the new scene multi-dimensional code information with the stored scene multi-dimensional code information includes:

[0031] Calculating the similarity between the new scene multi-dimensional code information and the stored scene multi-dimensional code information to improve the overall matching efficiency, and verifying and evaluating the matching results, specifically including:

[0032] A variety of similarity measurement methods are used to comprehensively evaluate the similarity, and a deep learning model based on a convolutional neural network, a recurrent neural network, or a generative adversarial network is used to learn feature representation and matching patterns, and transfer learning technology is used to perform fine-tuning in different application scenarios. The multiple similarity measurement methods include: Euclidean distance, cosine similarity, Hamming distance, and Mahalanobis distance;

[0033] Use dynamic programming algorithms to record and utilize historical matching results during multiple matching processes, optimize matching paths and results, and reduce repeated calculations; or use graph embedding algorithms to embed complex scene information into low-dimensional vector space to improve matching efficiency and accuracy. The graph embedding algorithms use graph theory and embedding technology to convert node and edge relationships in the scene into low-dimensional vector representations;

[0034] Similarity verification, result verification and consistency verification methods are used respectively to verify the similarity calculation results, the logical consistency of the matching results and the overall matching effect to ensure the accuracy and reliability of data matching.

[0035] The embodiment of the present invention provides a multi-dimensional code intelligent parsing system based on visual information processing results, and the multi-dimensional code intelligent parsing system includes:

[0036] The encoding module is used to efficiently encode the processing results of the input visual spatiotemporal information to generate scene multi-dimensional code information containing spatiotemporal features;

[0037] A storage module, used to store the scene multi-dimensional code information;

[0038] The matching module is used to compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information, so as to achieve accurate scene recognition and understanding.

[0039] Optionally, the encoding module includes: a feature extraction unit, a multi-dimensional code generation unit and a redundancy check unit;

[0040] The feature extraction unit is used to extract key visual features and depth of field information features from the processing results of the visual spatiotemporal information pair;

[0041] The multi-dimensional code generating unit is used to generate the scene multi-dimensional code information according to the key visual features and the depth of field information features;

[0042] The redundant check unit is used to embed redundant check information in the scene multi-dimensional code information to ensure data integrity and data damage resistance.

[0043] Optionally, the feature extraction unit includes:

[0044] A visual feature extraction subunit, for extracting key visual feature data of edge, texture, color, grayscale and shape representing the image content from the processing result of the visual spatiotemporal information pair using visual feature extraction technology, wherein the visual feature extraction technology includes: Canny edge detection algorithm, Gabor filter, color histogram or shape descriptor;

[0045] The depth feature extraction subunit is used to extract the depth information feature data related to the depth information, such as distance, object contour and spatial distribution relationship, from the processing results of the visual spatiotemporal information using the depth feature extraction technology, wherein the depth feature extraction technology includes: stereo ranging technology.

[0046] Optionally, the multi-dimensional code generating unit is specifically used for:

[0047] Constructing the scene multi-dimensional code information including the multi-dimensional code based on the visual and depth information according to the key visual feature data and the depth information feature data, specifically comprising:

[0048] A multi-dimensional encoding subunit, used for mapping the key visual feature data and the depth of field feature data into a multi-dimensional space coordinate system, and constructing scene multi-dimensional code information of different dimensions according to different visual features and different depth of field features;

[0049] The embedding subunit is used to embed the information of the base point pixel block in the scene multi-dimensional code information, and embed the weight information of the base point pixel block and its surrounding pixel blocks in the scene multi-dimensional code information. The base point pixel block refers to the intersection of the visual spatiotemporal information and the visual intersection line when observing the target. If the surrounding pixel blocks are closer to the base point pixel block in the same section, the greater the retrieval weight coefficient assigned to them. If the distance from the base point pixel block to the midpoint of the camera baseline is in different sections, the closer the base point pixel block is, the greater the retrieval weight coefficient assigned to it. In general, the retrieval weight coefficient increases as the distance from the base point pixel block decreases until it reaches the maximum at the base point pixel block; processing scenes or targets within a spatial range based on the weight information of the base point pixel block and its surrounding pixel blocks can enhance the accuracy of spatial processing and intelligent recognition;

[0050] The coding optimization subunit is used to optimize the structure and data layout of the scene multi-dimensional code to improve the coding efficiency and data compression rate; compress the feature data using the compression algorithm to reduce the size of the scene multi-dimensional code, while controlling the proportion of redundant information in the scene multi-dimensional code to balance data integrity and coding efficiency; dynamically adjust the coding method according to the complexity of each feature data to achieve efficient coding;

[0051] The multi-dimensional code verification sub-unit is used to perform real-time verification during the scene multi-dimensional code generation process to ensure the correctness and stability of the encoding; it performs real-time data verification, integrity verification and anti-damage testing respectively to ensure that the generated scene multi-dimensional code can still be correctly decoded in the case of partial damage.

[0052] Optionally, the redundancy check unit includes:

[0053] A redundant information embedding subunit, used to embed the redundant check information in the scene multi-dimensional code information by using a redundant coding method for data verification and recovery, wherein the redundant coding method includes: a parity check code or a hash check code;

[0054] The error detection subunit is used to detect error information in the scene multi-dimensional code information by using a verification method during the decoding process to ensure that it is not damaged during transmission or storage, and the verification method includes: cyclic redundancy check;

[0055] The error correction subunit is used to correct the detected erroneous information using the redundant check information and error correction technology if there is erroneous information, so as to realize automatic correction of erroneous data and ensure data integrity and reliability. The error correction technology includes: Hamming code and Reed-Solomon code.

[0056] The multi-dimensional code intelligent parsing method based on the processing results of visual spatiotemporal information pairs proposed in the present invention is intended to simulate the human's rapid recognition and understanding process of complex scenes. Different from the traditional method of image recognition that relies on deep learning neural networks, this method adopts a novel processing algorithm based on visual spatiotemporal information pairs. Visual spatiotemporal information pairs are similar to stereo pairs. By providing multi-angle images of the scene and their corresponding depth of field information, they can capture rich spatiotemporal information, which can then be used to accurately describe and analyze the spatial structure and various target objects in the scene.

[0057] The processing results of visual spatiotemporal information are used as input to identify and encode each target in the scene. In the multi-dimensional code encoding process, the visual features and depth information of the scene are combined, and multi-dimensional code encoding technology similar to QR code or 3D code is used to convert complex scene information into an efficient and compressed encoding form. This encoding process can not only retain the key features of the scene, but also ensure the integrity and anti-corruption of the data through the redundant verification mechanism of the multi-dimensional code.

[0058] After the encoding is completed, the encoded information is stored in an efficient database structure. When the same or similar scene is encountered again, the scene can be quickly identified and understood by quickly encoding the current scene and matching and comparing it with the stored multi-dimensional code scene information. The matching process relies on the corresponding algorithm to ensure the speed and accuracy of the matching, while greatly reducing the reliance on complex computing resources.

[0059] The scene multi-dimensional code design and processing method of the present invention realizes rapid recognition and understanding of scenes through innovative coding and matching methods, and is particularly suitable for application scenarios with high real-time requirements, such as intelligent monitoring, autonomous driving, and augmented reality. The advantages of this method are its efficient coding structure, strong fault tolerance, and excellent recognition speed, which provides a new technical framework for intelligent scene processing and significantly improves the efficiency and reliability of scene information processing. This method improves the scene recognition and understanding capabilities of complex environments, and provides strong technical support for humanoid robots, driverless cars, smart glasses, and inspection equipment.

[0060] The innovation of the present invention is that a multi-dimensional code encoding and processing method is proposed, which integrates image, depth of field and spatiotemporal features into a unified encoding framework, thereby improving the expression ability and processing efficiency of scene information. Through the redundant verification technology, the integrity and damage resistance of the encoded data are enhanced, ensuring reliable performance in complex environments. The method adopts an efficient database structure and a distributed storage system to improve the storage and retrieval efficiency of multi-dimensional code information, and provides a strong guarantee for intelligent scene processing in large-scale data environments. In addition, a similarity calculation and comparison algorithm is designed, combined with optimization and auxiliary algorithms to ensure the accuracy and real-time performance of the matching results, and improve the scene recognition and understanding capabilities of complex environments. It has important application value and broad market prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0062] Figure 1 It is a flow chart of a multi-dimensional code intelligent parsing method for processing results based on visual spatiotemporal information proposed in an embodiment of the present invention;

[0063] Figure 2 It is a schematic diagram of one of the distributions of various coding features exemplified in the embodiment of the present invention;

[0064] Figure 3 It is a schematic diagram of one of the base point pixel block collection examples in the embodiment of the present invention;

[0065] Figure 4 It is a block diagram of a multi-dimensional code intelligent parsing system based on visual spatiotemporal information processing results proposed in an embodiment of the present invention;

[0066] Figure 5 It is a better architecture diagram of the multi-dimensional code intelligent parsing system based on visual spatiotemporal information processing results of an embodiment of the present invention. DETAILED DESCRIPTION

[0067] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, are only part of the embodiments of the present invention, not all of the embodiments, and are not used to limit the present invention.

[0068] The present invention proposes a multi-dimensional code intelligent parsing method based on visual spatiotemporal information for processing results, referring to Figure 1 As shown in the flowchart, the multi-dimensional code intelligent parsing method includes:

[0069] Step 101: efficiently encode the input visual spatiotemporal information with depth information to generate and store scene multi-dimensional code information containing spatiotemporal features. The processing result of the visual spatiotemporal information with depth information is multi-dimensional scene raster information.

[0070] In the multi-dimensional code intelligent parsing method proposed in the present invention, it is first necessary to efficiently encode the processing results of the input visual spatiotemporal information pairs with depth of field information, which can be understood as two steps. The first step is to receive the visual spatiotemporal information pairs with depth of field information, and the second step is to process the received visual spatiotemporal information pairs to obtain the processing results. Then, based on the processing results, efficient encoding processing is performed to generate and store the scene multi-dimensional code information containing spatiotemporal features.

[0071] A method for efficiently encoding the processing results of the input visual spatiotemporal information to generate scene multi-dimensional code information containing spatiotemporal features includes:

[0072] Firstly, the key visual features and depth information features are extracted from the processing results of visual spatiotemporal information. Secondly, the scene multi-dimensional code information is generated according to the extracted key visual features and depth information features. Finally, redundant verification information is embedded in the scene multi-dimensional code information to ensure data integrity and data resistance.

[0073] Specifically, for extracting key visual features and depth information features from the processing results of visual spatiotemporal information, a better method includes:

[0074] The key visual feature data of edge, texture, color, grayscale and shape representing the image content are extracted from the processing results of visual spatiotemporal information using visual feature extraction technology. The visual feature extraction technology includes: Canny edge detection algorithm, Gabor filter, color histogram or shape descriptor. For example: when processing an image containing multiple colors, the color histogram can identify the color distribution in the image, while the Gabor filter is used to extract the texture features in the image, etc.

[0075] The depth feature extraction technology is used to extract the distance, object outline and spatial distribution relationship from the visual spatiotemporal information processing results. The depth feature extraction technology includes stereo ranging technology. For example, the depth information of the object can be calculated through the stereo image pair ranging method, which helps to understand the distribution of the object in the three-dimensional space.

[0076] For scene multi-dimensional code information, scene multi-dimensional code information including multi-dimensional code based on vision and depth of field information can be constructed based on key visual feature data and depth of field information feature data. Specifically, for constructing scene multi-dimensional code information including multi-dimensional code based on vision and depth of field information, a better method includes:

[0077] First, the extracted key visual feature data and depth information feature data are mapped to a multidimensional space coordinate system, and at the same time, different dimensions of scene multidimensional code information are constructed based on different visual features and different depth features. Among them, different visual features include black and white, grayscale or color, etc.; different depth features include distance, contour or spatial distribution, etc. By dynamically adjusting the encoding strategy, multidimensional code information of different dimensions can be constructed.

[0078] Then, the weight information of the base pixel block and its surrounding pixel blocks is embedded in the scene multi-dimensional code information. The so-called base pixel block refers to the intersection of the visual spatiotemporal information and the visual cross line when observing the target. If the surrounding pixel blocks are closer to the base pixel block in the same section, the greater the retrieval weight coefficient assigned to them. If the distance from the base pixel block to the midpoint of the camera baseline is in different sections, the closer the base pixel block is, the greater the retrieval weight coefficient assigned to it. In general, the retrieval weight coefficient increases as the distance from the base pixel block decreases until it reaches the maximum at the base pixel block; processing scenes or targets within the spatial range based on the weight information of the base pixel block and its surrounding pixel blocks can enhance the accuracy of spatial processing and intelligent recognition.

[0079] Then, the structure and data layout of the scene multi-dimensional code are optimized to improve the coding efficiency and data compression rate; the compression algorithm is used to compress the feature data to reduce the size of the scene multi-dimensional code, while controlling the proportion of redundant information in the scene multi-dimensional code to balance data integrity and coding efficiency; the coding method is dynamically adjusted according to the complexity of each feature data to achieve efficient coding.

[0080] Finally, real-time verification is performed during the scene multi-dimensional code generation process to ensure the correctness and stability of the encoding; real-time data verification, integrity verification and damage resistance testing are performed respectively to ensure that the generated scene multi-dimensional code can still be correctly decoded in the event of partial damage.

[0081] In order to better understand the above coding features and the meaning of base point pixel block collection, refer to Figure 2 One of the schematic diagrams of the distribution of the various coding features cited in the example; refer to Figure 3 One of the schematic diagrams of the base point pixel block acquisition exemplified. Figure 2 The medium visual features correspond to the key visual feature data, the deep features correspond to the depth of field information feature data, and the spatial-temporal features contained in the visual spatial-temporal information pair are combined to form the distribution of various coding features, and the scene multi-dimensional code information is constructed based on these data.

[0082] Figure 3 The visual space-time information pair acquisition device 301 is used to acquire visual space-time information pairs. It has two cameras C1302 and C2303, and 304 is the midpoint M of the camera baseline. The base point pixel block 305 is the observation target, and the section A1306 is the section where the observation target is located. The weight information of the base point pixel block and its surrounding pixel blocks will be embedded in the scene multi-dimensional code information. If in the same section, the closer the surrounding pixel blocks are to the base point pixel blocks, the greater the retrieval weight coefficient assigned to them. If in different sections, along the direction from the base point pixel block to the midpoint of the camera baseline, the closer the base point pixel block is to the base point pixel block, the greater the retrieval weight coefficient assigned to it. In general, the retrieval weight coefficient increases continuously as the distance from the base point pixel block decreases until it reaches the maximum at the base point pixel block. Figure 3 The darker the color of the middle pixel block, the greater the retrieval weight assigned to it. For example, compared with section A1306, section A2307 is assigned a smaller retrieval weight, and compared with section A2307, section A3308 is assigned a smaller retrieval weight, that is, section A3308 is assigned the smallest retrieval weight. Processing scenes or targets within a spatial range based on the weight information of the base point pixel block and its surrounding pixel blocks can enhance the accuracy of spatial processing and intelligent recognition.

[0083] A better method for embedding redundant checksum information in the scene multi-dimensional code information to ensure data integrity and data damage resistance includes:

[0084] First, redundant check information is embedded in the scene multi-dimensional code information using a redundant coding method for data verification and recovery. The redundant coding method includes: parity check code or hash check code; then, during the decoding process, a check method is used to detect error information in the scene multi-dimensional code information to ensure that it is not damaged during transmission or storage. The check method includes: cyclic redundancy check; finally, if there is error information, the detected error information is corrected using redundant check information and error correction technology to achieve automatic correction of erroneous data and ensure data integrity and reliability. The error correction technology includes: Hamming code and Reed-Solomon code.

[0085] For the storage of scene multi-dimensional codes, a better method includes: first, storing and retrieving the scene multi-dimensional code information by optimizing the database structure, and the optimized database structure includes index optimization, data partitioning and parallel processing technology; specifically, using B-tree and hash index technology to accelerate the query speed, optimizing the data storage layout through horizontal and vertical partitioning technology, and using parallel processing technology to improve data access performance, ensuring a rapid response to the scene multi-dimensional code information; and, in the storage process, dividing the multi-dimensional code encoding information into spatial ranges, giving the data close to the base point pixel block a higher retrieval weight coefficient for rapid retrieval and processing.

[0086] Subsequently, the scene multi-dimensional code information is stored in a dispersed manner through multiple storage nodes; specifically, when writing data, the scene multi-dimensional code information is distributed to different storage nodes using data sharding and consistent hashing algorithms to improve storage efficiency; when reading data, the target data is quickly located and retrieved through intelligent routing and load balancing strategies, and nodes are allocated based on base point pixel blocks, making spatial range-based retrieval more efficient.

[0087] Finally, local processing and rapid response of scene multi-dimensional code information are achieved through edge computing nodes. Specifically, scene multi-dimensional code information is stored locally on edge devices, and lightweight database or distributed storage technology is used to reduce the load of the central server and network transmission delay, ensuring storage efficiency and reliability. Incremental synchronization and delayed synchronization technology are used to synchronize data between edge devices and central servers to achieve efficient data transmission and synchronization and ensure seamless collaboration. At the same time, containerization technology is used to dynamically adjust edge computing and storage resources according to data scale and business needs, to achieve elastic expansion and automated management of edge computing nodes, and to ensure efficient storage and computing performance when the amount of data increases significantly.

[0088] Step 102: Compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information to achieve accurate scene recognition and understanding.

[0089] The scene multi-dimensional code information is obtained and stored through step 101. Later, in the actual application process, for example, when the intelligent robot works in a certain scene, it can obtain the visual space-time information pair in real time based on the visual space-time information acquisition device and generate new scene multi-dimensional code information, and then directly compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information, so as to achieve accurate scene recognition and understanding.

[0090] A preferred method for comparing and matching new scene multi-dimensional code information with stored scene multi-dimensional code information includes:

[0091] First, the similarity between the new scene multi-dimensional code information and the stored scene multi-dimensional code information is calculated to improve the overall matching efficiency, and the matching results are verified and evaluated, including:

[0092] A variety of similarity measurement methods are used to comprehensively evaluate the similarity. At the same time, deep learning models based on convolutional neural networks, recurrent neural networks or generative adversarial networks are used to learn feature representation and matching patterns, and transfer learning technology is used to fine-tune them in different application scenarios. The multiple similarity measurement methods include: Euclidean distance, cosine similarity, Hamming distance and Mahalanobis distance.

[0093] Use dynamic programming algorithms to record and utilize historical matching results during multiple matching processes, optimize matching paths and results, and reduce repeated calculations; or use graph embedding algorithms to embed complex scene information into low-dimensional vector space to improve matching efficiency and accuracy. The graph embedding algorithm uses graph theory and embedding technology to convert the node and edge relationships in the scene into low-dimensional vector representations.

[0094] Similarity verification, result verification and consistency verification methods are used respectively to verify the similarity calculation results, the logical consistency of the matching results and the overall matching effect to ensure the accuracy and reliability of data matching.

[0095] The above-mentioned multi-dimensional code encoding and processing method integrates image, depth of field and spatiotemporal features into a unified encoding framework, which improves the expression ability and processing efficiency of scene information. Through the redundancy check technology, the integrity and damage resistance of the encoded data are enhanced to ensure reliable performance in complex environments. This method adopts an efficient database structure and distributed storage system to improve the storage and retrieval efficiency of multi-dimensional code information, and provides a strong guarantee for intelligent scene processing in large-scale data environments. In addition, similarity calculation and comparison algorithms are designed, combined with optimization and auxiliary algorithms to ensure the accuracy and real-time performance of matching results, and improve the scene recognition and understanding capabilities of complex environments.

[0096] Based on the above multi-dimensional code intelligent parsing method, the embodiment of the present invention also proposes a multi-dimensional code intelligent parsing system based on visual spatiotemporal information processing results, referring to Figure 4 As shown in the block diagram, the multi-dimensional code intelligent parsing system includes:

[0097] The encoding module 410 is used to efficiently encode the input visual spatiotemporal information to generate scene multi-dimensional code information containing spatiotemporal features;

[0098] The storage module 420 is used to store the scene multi-dimensional code information;

[0099] The matching module 430 is used to compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information, so as to achieve accurate scene recognition and understanding.

[0100] Combination Figure 5 The schematic diagram of a better multi-dimensional code intelligent parsing system architecture is shown, wherein the encoding module includes: a feature extraction unit, a multi-dimensional code generation unit and a redundancy check unit;

[0101] A feature extraction unit, used to extract key visual features and depth of field information features from the processing results of the visual spatiotemporal information;

[0102] A multi-dimensional code generating unit, used for generating scene multi-dimensional code information according to key visual features and depth of field information features;

[0103] The redundant check unit is used to embed redundant check information in the scene multi-dimensional code information to ensure data integrity and data damage resistance.

[0104] The feature extraction unit includes:

[0105] The visual feature extraction subunit is used to extract key visual feature data of edge, texture, color, grayscale and shape representing the image content from the processing results of visual spatiotemporal information using visual feature extraction technology. The visual feature extraction technology includes: Canny edge detection algorithm, Gabor filter, color histogram or shape descriptor;

[0106] The depth feature extraction subunit is used to extract the depth information feature data related to the depth information, such as distance, object contour and spatial distribution relationship, from the processing results of the visual spatiotemporal information using the depth feature extraction technology. The depth feature extraction technology includes: stereo ranging technology.

[0107] The multi-dimensional code generation unit is specifically used for:

[0108] According to the key visual feature data and the depth information feature data, the scene multi-dimensional code information including the multi-dimensional code based on the visual and depth information is constructed, including:

[0109] The multi-dimensional encoding sub-unit is used to map the key visual feature data and the depth of field feature data into a multi-dimensional space coordinate system, and construct scene multi-dimensional code information of different dimensions according to different visual features and different depth of field features;

[0110] The embedding subunit is used to embed the weight information of the base point pixel block and its surrounding pixel blocks in the scene multi-dimensional code information. The base point pixel block is the intersection of the visual spatiotemporal information and the visual intersection line. If the surrounding pixel blocks are closer to the base point pixel block in the same section, the greater the retrieval weight coefficient assigned to them. If the distance to the base point pixel block is closer to the base point pixel block in different sections, the greater the retrieval weight coefficient assigned to them. In general, the retrieval weight coefficient increases as the distance from the base point pixel block decreases until it reaches the maximum at the base point pixel block. Processing scenes or targets within a spatial range based on the weight information of the base point pixel block and its surrounding pixel blocks can enhance the accuracy of spatial processing and intelligent recognition;

[0111] The coding optimization subunit is used to optimize the structure and data layout of the scene multi-dimensional code to improve the coding efficiency and data compression rate; compress the feature data using the compression algorithm to reduce the size of the scene multi-dimensional code, while controlling the proportion of redundant information in the scene multi-dimensional code to balance data integrity and coding efficiency; dynamically adjust the coding method according to the complexity of each feature data to achieve efficient coding;

[0112] The multi-dimensional code verification sub-unit is used to perform real-time verification during the scene multi-dimensional code generation process to ensure the correctness and stability of the encoding; it performs real-time data verification, integrity verification and anti-damage testing respectively to ensure that the generated scene multi-dimensional code can still be correctly decoded in the case of partial damage.

[0113] The redundancy check unit includes:

[0114] A redundant information embedding subunit is used to embed redundant check information in the scene multi-dimensional code information by using a redundant coding method for data verification and recovery, wherein the redundant coding method includes: a parity check code or a hash check code;

[0115] The error detection subunit is used to detect error information in the scene multi-dimensional code information by using a verification method during the decoding process to ensure that it is not damaged during transmission or storage, and the verification method includes: cyclic redundancy check;

[0116] The error correction subunit is used to correct the detected erroneous information using redundant check information and error correction technology if there is erroneous information, so as to realize automatic correction of erroneous data and ensure data integrity and reliability. The error correction technology includes: Hamming code and Reed-Solomon code.

[0117] In order to store the scene multi-dimensional code information, a storage module is provided. In the storage module, the database structure (i.e. Figure 5An efficient database structure is used to store and retrieve scene multi-dimensional code information. The optimized database structure includes index optimization, data partitioning and parallel processing technology. Specifically, B-tree and hash index technology are used to accelerate query speed, horizontal and vertical partitioning technology is used to optimize data storage layout, and parallel processing technology is used to improve data access performance to ensure rapid response to scene multi-dimensional code information. In addition, the multi-dimensional code encoding information is divided into spatial ranges during the storage process, and data close to the base point pixel block is given a higher retrieval weight coefficient for rapid retrieval and processing.

[0118] The storage module uses multiple storage nodes (i.e. Figure 5 The distributed storage system in the system stores the scene multi-dimensional code information in a dispersed manner. Specifically, when writing data, the scene multi-dimensional code information is distributed to different storage nodes using data sharding and consistent hashing algorithms to improve storage efficiency. When reading data, the target data is quickly located and retrieved through intelligent routing and load balancing strategies, and nodes are allocated based on base point pixel blocks, making spatial range-based retrieval more efficient.

[0119] The storage module is connected through the edge computing node (i.e. Figure 5 The edge storage computing unit in the middle realizes local processing and rapid response to the scene multi-dimensional code information; specifically, the scene multi-dimensional code information is stored locally on the edge device, and a lightweight database or distributed storage technology is used to reduce the load of the central server and network transmission delay, ensuring the efficiency and reliability of storage; incremental synchronization and delayed synchronization technology are used to synchronize data between the edge device and the central server, to achieve efficient data transmission and synchronization, and ensure seamless collaboration; at the same time, containerization technology is used to dynamically adjust edge computing and storage resources according to data scale and business needs, to achieve elastic expansion and automated management of edge computing nodes, and to ensure efficient storage and computing performance when the amount of data increases significantly.

[0120] In order to compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information, a matching module is provided. The matching module uses a variety of similarity measurement methods (i.e., using Figure 5 In the similarity calculation and comparison unit, the similarity is comprehensively evaluated. At the same time, a deep learning model based on a convolutional neural network, a recurrent neural network or a generative adversarial network is used to learn feature representation and matching patterns, and transfer learning technology is used to perform fine-tuning in different application scenarios. The multiple similarity measurement methods include: Euclidean distance, cosine similarity, Hamming distance and Mahalanobis distance.

[0121] Use dynamic programming algorithms to record and utilize historical matching results during multiple matching processes, optimize matching paths and results, and reduce repeated calculations; or use graph embedding algorithms to embed complex scene information into low-dimensional vector space to improve matching efficiency and accuracy (that is, using Figure 5 The optimization and auxiliary algorithm units optimize the matching paths and results, or use Figure 5 The optimization and auxiliary algorithm unit embeds complex scene information into a low-dimensional vector space), and the graph embedding algorithm uses graph theory and embedding technology to convert the node and edge relationships in the scene into a low-dimensional vector representation.

[0122] The similarity verification, result verification and consistency verification methods are used to verify the similarity calculation results, the logical consistency of the matching results and the overall matching effect (that is, using Figure 5 The matching result verification unit verifies the similarity calculation results, the logical consistency of the matching results, and the overall matching effect) to ensure the accuracy and reliability of data matching.

[0123] In summary, the multi-dimensional code intelligent parsing method based on the processing results of visual spatiotemporal information pairs proposed in the present invention is intended to simulate the human rapid recognition and understanding process of complex scenes. Unlike the traditional method of image recognition that relies on deep learning neural networks, this method adopts a novel processing algorithm based on visual spatiotemporal information pairs. Visual spatiotemporal information pairs are similar to stereo pairs. By providing multi-angle images of the scene and their corresponding depth of field information, they can capture rich spatiotemporal information, which can be used to accurately describe and analyze the spatial structure and various target objects in the scene.

[0124] The processing results of visual spatiotemporal information are used as input to identify and encode each target in the scene. In the multi-dimensional code encoding process, the visual features and depth information of the scene are combined, and multi-dimensional code encoding technology similar to QR code or 3D code is used to convert complex scene information into an efficient and compressed encoding form. This encoding process can not only retain the key features of the scene, but also ensure the integrity and anti-corruption of the data through the redundant verification mechanism of the multi-dimensional code.

[0125] After the encoding is completed, the encoded information is stored in an efficient database structure. When the same or similar scene is encountered again, the scene can be quickly identified and understood by quickly encoding the current scene and matching and comparing it with the stored multi-dimensional code scene information. The matching process relies on the corresponding algorithm to ensure the speed and accuracy of the matching, while greatly reducing the reliance on complex computing resources.

[0126] The scene multi-dimensional code design and processing method of the present invention realizes rapid recognition and understanding of scenes through innovative coding and matching methods, and is particularly suitable for application scenarios with high real-time requirements, such as intelligent monitoring, autonomous driving, and augmented reality. The advantages of this method are its efficient coding structure, strong fault tolerance, and excellent recognition speed, which provides a new technical framework for intelligent scene processing and significantly improves the efficiency and reliability of scene information processing. This method improves the scene recognition and understanding capabilities of complex environments, and provides strong technical support for humanoid robots, driverless cars, smart glasses, and inspection equipment.

[0127] The innovation of the present invention is that a multi-dimensional code encoding and processing method is proposed, which integrates image, depth of field and spatiotemporal features into a unified encoding framework, thereby improving the expression ability and processing efficiency of scene information. Through the redundant verification technology, the integrity and damage resistance of the encoded data are enhanced, ensuring reliable performance in complex environments. The method adopts an efficient database structure and a distributed storage system to improve the storage and retrieval efficiency of multi-dimensional code information, and provides a strong guarantee for intelligent scene processing in large-scale data environments. In addition, a similarity calculation and comparison algorithm is designed, combined with optimization and auxiliary algorithms to ensure the accuracy and real-time performance of the matching results, and improve the scene recognition and understanding capabilities of complex environments. It has important application value and broad market prospects.

[0128] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0129] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0130] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A multi-dimensional code intelligent analysis method for processing results based on visual information, characterized in that: The multi-dimensional code intelligent parsing method comprises: The input visual spatiotemporal information processing result with depth information is efficiently coded and processed to generate and store scene multi-dimensional code information containing spatiotemporal features, wherein the visual spatiotemporal information processing result with depth information is multi-dimensional scene raster information; wherein generating the scene multi-dimensional code information containing spatiotemporal features comprises: extracting key visual features and depth information features from the visual spatiotemporal information processing result; constructing the scene multi-dimensional code information including multi-dimensional coding based on visual and depth information according to the key visual features and the depth information feature data; and embedding redundant check information in the scene multi-dimensional code information to ensure data integrity and data damage resistance; Compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information to achieve accurate scene recognition and understanding; Among them, according to the key visual features and the depth of field information feature data, the scene multidimensional code information including the multidimensional encoding based on vision and depth of field information is constructed, including: mapping the key visual feature data and the depth of field information feature data into a multidimensional space coordinate system, and constructing scene multidimensional code information of different dimensions according to different visual features and different depth of field features; embedding weight information of the base point pixel block and its surrounding pixel blocks in the scene multidimensional code information, the base point pixel block refers to the intersection of the visual spatiotemporal information and the visual intersection line when observing the target, if in the same section, the closer the surrounding pixel block is to the base point pixel block, the greater the retrieval weight coefficient assigned to it, if in different sections, along the base point pixel block to the midpoint of the dual camera baseline, the closer the base point pixel block is to the base point pixel block, the greater the retrieval weight coefficient assigned to it; the retrieval weight coefficient increases continuously as the distance from the base point pixel block decreases until the base point pixel block. The maximum value is reached at the block; the scene or target within the spatial range is processed based on the weight information of the base point pixel block and its surrounding pixel blocks to enhance the accuracy of spatial processing and intelligent recognition; the structure and data layout of the scene multi-dimensional code are optimized to improve the encoding efficiency and data compression rate; the feature data is compressed using a compression algorithm to reduce the size of the scene multi-dimensional code, while controlling the proportion of redundant information in the scene multi-dimensional code to balance data integrity and encoding efficiency; the encoding method is dynamically adjusted according to the complexity of each feature data to achieve efficient encoding; real-time verification is performed during the scene multi-dimensional code generation process to ensure the correctness and stability of the encoding; real-time data verification, integrity verification and damage resistance testing are performed separately to ensure that the generated scene multi-dimensional code can still be correctly decoded in the case of partial damage.

2. The multi-dimensional code intelligent parsing method according to claim 1, characterized in that: Extracting key visual features and depth information features from the visual spatiotemporal information processing results, including: Extracting key visual feature data of edge, texture, color, grayscale and shape representing the image content from the processing results of the visual spatiotemporal information using visual feature extraction technology, wherein the visual feature extraction technology includes: Canny edge detection algorithm, Gabor filter, color histogram or shape descriptor; Depth feature extraction technology is used to extract depth information feature data related to depth information, such as distance, object contour and spatial distribution relationship, from the processing results of the visual spatiotemporal information. The depth feature extraction technology includes: stereo ranging technology.

3. The multi-dimensional code intelligent parsing method according to claim 1, characterized in that: Redundant check information is embedded in the scenario multi-dimensional code information to ensure data integrity and data damage resistance, including: The redundant check information is embedded in the scene multi-dimensional code information by using a redundant coding method for data verification and recovery, wherein the redundant coding method includes: a parity check code or a hash check code; During the decoding process, a verification method is used to detect error information in the scene multi-dimensional code information to ensure that it is not damaged during transmission or storage, and the verification method includes: a cyclic redundancy check; If there is erroneous information, the redundant check information and error correction technology are used to correct the detected erroneous information to achieve automatic correction of erroneous data and ensure data integrity and reliability. The error correction technology includes: Hamming code and Reed-Solomon code.

4. The multi-dimensional code intelligent parsing method according to claim 1, characterized in that: The method for storing the scene multi-dimensional code information includes: The scene multi-dimensional code information is stored and retrieved by optimizing the database structure, and the optimized database structure includes index optimization, data partitioning and parallel processing technology; specifically, the query speed is accelerated by using B-tree and hash index technology, the data storage layout is optimized by horizontal and vertical partitioning technology, and the data access performance is improved by using parallel processing technology to ensure a fast response to the scene multi-dimensional code information; and, in the storage process, the multi-dimensional code encoding information is spatially divided, and based on the spatial position information of the base point pixel block, a higher retrieval weight coefficient is given to the data of the pixel block close to the base point, so as to facilitate fast retrieval and processing; The scene multi-dimensional code information is stored in a dispersed manner through multiple storage nodes; specifically, when writing data, the scene multi-dimensional code information is distributed to different storage nodes using data sharding and consistent hashing algorithms to improve storage efficiency; when reading data, the target data is quickly located and retrieved through intelligent routing and load balancing strategies, and nodes are allocated based on base point pixel blocks, making retrieval based on spatial range more efficient; Local processing and rapid response of the scene multi-dimensional code information are achieved through edge computing nodes; specifically, the scene multi-dimensional code information is stored locally on the edge device, and lightweight database or distributed storage technology is used to reduce the load of the central server and network transmission delay, thereby ensuring storage efficiency and reliability; incremental synchronization and delayed synchronization technology are used to synchronize data between the edge device and the central server, thereby achieving efficient data transmission and synchronization and ensuring seamless collaboration; at the same time, containerization technology is used to dynamically adjust edge computing and storage resources according to data scale and business needs, thereby achieving elastic expansion and automated management of edge computing nodes, thereby ensuring efficient storage and computing performance when the amount of data increases significantly.

5. The multi-dimensional code intelligent parsing method according to claim 1, characterized in that: Compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information, including: Calculating the similarity between the new scene multi-dimensional code information and the stored scene multi-dimensional code information to improve the overall matching efficiency, and verifying and evaluating the matching results, specifically including: A variety of similarity measurement methods are used to comprehensively evaluate the similarity, and a deep learning model based on a convolutional neural network, a recurrent neural network, or a generative adversarial network is used to learn feature representation and matching patterns, and transfer learning technology is used to perform fine-tuning in different application scenarios. The multiple similarity measurement methods include: Euclidean distance, cosine similarity, Hamming distance, and Mahalanobis distance; Use dynamic programming algorithms to record and utilize historical matching results during multiple matching processes, optimize matching paths and results, and reduce repeated calculations; or use graph embedding algorithms to embed complex scene information into low-dimensional vector space to improve matching efficiency and accuracy. The graph embedding algorithms use graph theory and embedding technology to convert node and edge relationships in the scene into low-dimensional vector representations; Similarity verification, result verification and consistency verification methods are used respectively to verify the similarity calculation results, the logical consistency of the matching results and the overall matching effect to ensure the accuracy and reliability of data matching.

6. A multi-dimensional code intelligent analysis system based on visual information processing results, characterized in that: The multi-dimensional code intelligent parsing system comprises: The encoding module includes: a feature extraction unit, a multi-dimensional code generation unit and a redundancy check unit; The feature extraction unit is used to extract key visual features and depth of field information features from the processing results of the visual spatiotemporal information pair; The multi-dimensional code generating unit is used to construct scene multi-dimensional code information including multi-dimensional coding based on vision and depth information according to the key visual features and the depth information feature data; The redundant check unit is used to embed redundant check information in the scene multi-dimensional code information to ensure data integrity and data damage resistance; A storage module, used to store the scene multi-dimensional code information; A matching module is used to compare and match the new scene multi-dimensional code information with the stored scene multi-dimensional code information, so as to achieve accurate scene recognition and understanding; Wherein, the multi-dimensional code generating unit specifically includes: A multi-dimensional encoding subunit, used to map the key visual feature data and the depth information feature data into a multi-dimensional space coordinate system, and construct scene multi-dimensional code information of different dimensions according to different visual features and different depth features; The embedding subunit is used to embed the weight information of the base point pixel block and its surrounding pixel blocks in the scene multi-dimensional code information. The base point pixel block refers to the intersection of the visual spatiotemporal information and the visual intersection line when observing the target. If the surrounding pixel blocks are closer to the base point pixel block in the same section, the larger the retrieval weight coefficient assigned to them is. If the surrounding pixel blocks are closer to the base point pixel block in the direction from the base point pixel block to the midpoint of the dual camera baseline in different sections, the larger the retrieval weight coefficient assigned to them is. The retrieval weight coefficient increases continuously as the distance from the base point pixel block decreases until it reaches the maximum at the base point pixel block. The scene or target within the spatial range is processed based on the weight information of the base point pixel block and its surrounding pixel blocks to enhance the accuracy of spatial processing and intelligent recognition. The coding optimization subunit is used to optimize the structure and data layout of the scene multi-dimensional code to improve the coding efficiency and data compression rate; compress the feature data using the compression algorithm to reduce the size of the scene multi-dimensional code, while controlling the proportion of redundant information in the scene multi-dimensional code to balance data integrity and coding efficiency; dynamically adjust the coding method according to the complexity of each feature data to achieve efficient coding; The multi-dimensional code verification sub-unit is used to perform real-time verification during the scene multi-dimensional code generation process to ensure the correctness and stability of the encoding; it performs real-time data verification, integrity verification and anti-damage testing respectively to ensure that the generated scene multi-dimensional code can still be correctly decoded in the case of partial damage.

7. The multi-dimensional code intelligent parsing system according to claim 6, characterized in that: The feature extraction unit comprises: A visual feature extraction subunit, for extracting key visual feature data of edge, texture, color, grayscale and shape representing the image content from the processing result of the visual spatiotemporal information pair using visual feature extraction technology, wherein the visual feature extraction technology includes: Canny edge detection algorithm, Gabor filter, color histogram or shape descriptor; The depth feature extraction subunit is used to extract the depth information feature data related to the depth information, such as distance, object contour and spatial distribution relationship, from the processing results of the visual spatiotemporal information using the depth feature extraction technology, wherein the depth feature extraction technology includes: stereo ranging technology.

8. The multi-dimensional code intelligent parsing system according to claim 7, characterized in that: The redundancy check unit comprises: A redundant information embedding subunit, used to embed the redundant check information in the scene multi-dimensional code information by using a redundant coding method for data verification and recovery, wherein the redundant coding method includes: a parity check code or a hash check code; The error detection subunit is used to detect error information in the scene multi-dimensional code information by using a verification method during the decoding process to ensure that it is not damaged during transmission or storage, and the verification method includes: cyclic redundancy check; The error correction subunit is used to correct the detected erroneous information using the redundant check information and error correction technology if there is erroneous information, so as to realize automatic correction of erroneous data and ensure data integrity and reliability. The error correction technology includes: Hamming code and Reed-Solomon code.

Images