Ocean engineering material intelligent identification method and system based on two-dimensional code identification technology

Abstract

The invention discloses an ocean engineering material intelligent identification method and system based on a two-dimensional code identification technology. The method comprises the following steps: firstly, constructing a multi-dimensional material feature information model; the method comprises the following steps of: firstly, generating a high-density two-dimensional code based on improved Reed-Solomon (255, 223) error correction coding, and realizing reliable reading and data tampering prevention under high contamination through data block interleaving, dual verification and an ECC seecp256r1 curve digital signature mechanism; and secondly, generating a high-density two-dimensional code based on improved Reed-Solomon (255, 223) error correction coding, and realizing reliable reading and data tampering prevention under high contamination. The label adopts a special PET (Polyethylene Terephthalate) base material compounding process and passes a 96-hour salt spray test. The mobile terminal identification system constructs a three-level framework of preprocessing enhancement, multi-model fusion and dynamic decision, an optimized lightweight MobileNetV3-Small model is adopted, and the identification rate in a complex scene reaches 95% or above. According to the invention, the information density, the environmental adaptability and the recognition robustness are remarkably improved, and the whole-process traceable management of materials can be realized after the system is integrated with an intelligent warehousing system.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of ocean engineering equipment manufacturing, and particularly relates to a marine engineering material two-dimensional code identification and intelligent identification method and system based on feature information. BACKGROUND

[0002] The construction of an ocean engineering module involves tens of thousands of various materials, including structural parts and process parts of different specifications, materials, corrosion levels, etc. At present, warehouse management mainly relies on manual account books or traditional barcode technology, which has the following significant defects:

[0003] Limited information carrying capacity: Traditional two-dimensional codes only store item IDs and cannot directly carry key information such as structural features (size, weight, material), process requirements (welding level, detection standard), and protection characteristics (corrosion level, storage environment) of the materials. On-site workers need to frequently query the system to obtain detailed information, which is inefficient.

[0004] Poor environmental adaptability: Ocean engineering sites generally have temperature differences of-20℃ to 60℃, 95% relative humidity, salt spray corrosion, oil pollution, and equipment vibration. Ordinary paper labels rust and fall off after 96 hours of salt spray testing, and the two-dimensional code pollution rate is more than 30%, resulting in an identification success rate of less than 85%, which seriously affects the account consistency rate.

[0005] Weak anti-interference ability: Existing mobile identification technology uses the traditional ZBar algorithm, which has a sharp drop in recognition rate to less than 65% in strong light, two-dimensional code bending and deformation (curvature radius <5cm), and motion blur (caused by ship rolling), and the recognition time is more than 100ms, which cannot meet the real-time operation requirements.

[0006] Insufficient data security and tamper-proofing ability: Traditional identification lacks encryption mechanisms, and material information is easy to be forged or tampered with, making it difficult to meet the strict requirements of ocean engineering for quality traceability.

[0007] Although international leading enterprises have adopted RFID technology, its cost is high and its reliability decreases in high-metal interference environments, making it difficult to be widely promoted in large-scale ocean engineering module construction. Therefore, there is an urgent need for a new type of identification and recognition technology with high information density, strong environmental adaptability, good recognition robustness, and security authentication to improve the digital level of ocean engineering intelligent warehouse management. SUMMARY

[0008] The purpose of the present application is to overcome the shortcomings of the prior art, provide a marine engineering material intelligent identification method and system based on two-dimensional code recognition technology, based on the marine engineering material two-dimensional code identification and intelligent identification method with large feature information capacity, excellent environmental adaptability, high identification success rate and tamper-proof ability, realizing embedded management of material feature information, supporting whole-process digitalization of warehouse operation.

[0009] In order to achieve the above-mentioned purpose of the application, the present application proposes the following technical scheme:

[0010] In a first aspect, the present application proposes a marine engineering material intelligent identification method based on two-dimensional code recognition technology, comprising the following steps:

[0011] Step S101, constructing a multi-dimensional material feature information model: establishing a feature database covering three categories of structural features, process features and protection features, a total of 28 key parameters, defining data structure using ASN.1 abstract syntax notation standard, and realizing cross-platform data compatibility through basic encoding rule BER encoding and distinguishable encoding rule DER decoding;

[0012] Step S102, generating high-density two-dimensional code resistant to harsh environment: encoding the structured data obtained in step S101, using a three-way collaborative mechanism of block interleaving-dual verification-digital signature for encoding, including: dividing the original data into logical blocks, and interleaving and storing multiple logical blocks in matrix form to disperse continuous physical damage logically; converting the interleaved data into an encoding block containing redundant data, and adding a check code to each data block to realize fast integrity checking; generating a digital signature using an elliptic curve digital signature algorithm, the signature private key is stored in a hardware security module; the final data block after the above processing is used to generate a two-dimensional code pattern, which is printed on a physical carrier by using an anti-corrosion material;

[0013] Step S103, mobile intelligent identification: constructing a three-level identification architecture of preprocessing enhancement-multi-model fusion-dynamic decision, preprocessing and enhancing the collected two-dimensional code image, using a lightweight MobileNetV3-Small deep learning model for identification, and fusing image recognition results, radio frequency identification RFID signal strength and ultra-wideband UWB positioning information for multi-modal dynamic decision to output the final identification result;

[0014] Step S104, system integration and business linkage: integrating the identification result obtained in step S103 with the intelligent warehouse management system, realizing automatic triggering of business processes by scanning codes, driving a storage location optimization algorithm based on the parsed material feature information, and updating the digital twin state within 500 milliseconds.

[0015] In some embodiments, in step S102, the two-dimensional code is printed on a label made of polyethylene terephthalate (PET) substrate, resin carbon tape and film coating process, which passes the 96-hour neutral salt spray test, and the recognition rate is not less than 99.5% in the temperature range of-20℃ to 60℃ and the relative humidity range of 20% to 95%.

[0016] In some embodiments, the improved Reed-Solomon encoding scheme supports up to 4 logical block data loss for complete recovery; under the conditions of 30% random contamination, 50% oil pollution coverage, 20% edge damage and bending deformation with a curvature radius of 5 cm, the recognition success rate of the two-dimensional code is not less than 98.7%, 85.2%, 98.7% and 96.3% respectively.

[0017] In some embodiments, in step S103, the preprocessing enhancement includes distortion correction based on Zhang Zhengyou calibration method with an error of less than 0.5 pixels, noise removal by bilateral filtering with a 5x5 kernel, adaptive dynamic threshold binarization based on local image mean μ and standard deviation σ, threshold T = μ + k x σ, where k value is adaptively adjusted in the range of 0.5 to 2.0 according to environmental light intensity, contrast enhancement by limited contrast adaptive histogram equalization (CLAHE) algorithm, and anti-fuzzy processing of motion blur with displacement not exceeding 5 pixels and defocus blur with radius not exceeding 3 pixels by Wiener filter with recovery rate greater than 95%.

[0018] In some embodiments, after the MobileNetV3-Small model is optimized by INT8 quantization, knowledge distillation based on teacher model ResNet50 and 40% channel pruning, the model size is compressed to 3.2MB, the inference time on mobile platform is 18ms, and the recognition accuracy reaches 99.5%.

[0019] In some embodiments, in step S103, the multi-modal fusion decision dynamically adjusts the recognition confidence threshold by a weighted voting mechanism, so that the recognition rate in strong light and shadow scenes is improved to 95.7%, the recognition rate in oil pollution scenes is 89.3%, and the recognition rate in motion blur scenes caused by ship heaving is greater than 96%.

[0020] In a second aspect, the present application provides a two-dimensional code recognition technology-based marine engineering material intelligent identification system for implementing the two-dimensional code recognition technology-based marine engineering material intelligent identification method, comprising:

[0021] A feature information modeling module is configured to perform step S101 and construct the multi-dimensional material feature information model.

[0022] A QR code generation and label making module is configured to generate the high-density QR code and control an industrial printer to complete printing on an anti-corrosion label.

[0023] A mobile terminal intelligent recognition module is integrated in the mobile terminal and configured to implement the three-level recognition architecture.

[0024] An intelligent warehouse integration module is configured to implement data interaction with a background management system and business process driving.

[0025] In some embodiments, the mobile terminal intelligent recognition module comprises:

[0026] A preprocessing unit is configured to perform distortion correction, noise filtering, dynamic threshold binarization, contrast enhancement and anti-blurring processing.

[0027] A lightweight model recognition unit is configured to load and run the optimized MobileNetV3-Small model to perform QR code positioning and decoding.

[0028] A multi-modal decision unit is configured to receive and fuse the output of the lightweight model recognition unit (2032), RFID reader signals and UWB positioning module information, and output a final recognition result through a weighted voting mechanism.

[0029] In some embodiments, the intelligent warehouse integration module comprises a digital twin synchronization mechanism based on state hash verification, which ensures that the state synchronization delay between physical inventory and digital twin is less than 500 milliseconds, and the data consistency reaches 99.99%.

[0030] In some embodiments, a security audit module is further included, which is configured to implement two-factor authentication based on OAuth 2.0 and OpenID Connect (OIDC) protocol, attribute-based access control (ABAC), transport layer security (TLS) 1.3 and advanced encryption standard (AES-256) encryption, and store audit logs for 365 days in a write-once-read-many (WORM) mode.

[0031] Compared with the prior art, the technical scheme of the present application has the following beneficial effects:

[0032] 1) The improved RS encoding, the optimized AI model MobileNetV3, the physical process of preparing the PET+film-coated QR code, and the business logic of feature-driven storage location allocation are combined in the manner described in the present application to solve specific problems in ocean engineering, and the deep integration of multiple cross-technology fields produces a synergistic effect.

[0033] 2) The method first establishes a multi-dimensional material characteristic information model covering 28 parameters in three categories of structure, process and protection, and adopts ASN.1 standard for data structured definition; then generates a high-density two-dimensional code based on improved Reed-Solomon (255,223) error correction coding, realizes reliable reading and data tamper-proofing under high pollution through data block interleaving, double checking and ECC secp256r1 curve digital signature mechanism; the label adopts special PET substrate composite process and passes 96-hour salt spray test. The mobile terminal recognition system constructs a three-level architecture of preprocessing enhancement-multi-model fusion-dynamic decision, adopts an optimized lightweight MobileNetV3-Small model, and the recognition rate is above 95% in complex scenes such as strong light, oil stain and bending.

[0034] 3) The application is suitable for high-salt spray, strong vibration and oil pollution in marine engineering environment, and is based on two-dimensional code identification generation technology of multi-dimensional material characteristic information and mobile terminal intelligent identification method combining deep learning and image enhancement algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0035] Figure 1 It is a marine engineering material intelligent identification method flowchart based on two-dimensional code identification technology of the application.

[0036] Figure 2 It is a preprocessing enhancement-multi-model fusion-dynamic decision three-level identification architecture workflow diagram.

[0037] Figure 3 It is a marine engineering material intelligent identification system module diagram based on two-dimensional code identification technology for realizing the marine engineering material intelligent identification method based on two-dimensional code identification technology of any one of the above embodiments 1 of the application.

[0038] Figure 4 It is a marine engineering material intelligent identification system module diagram based on two-dimensional code identification technology for realizing the marine engineering material intelligent identification method based on two-dimensional code identification technology of any one of the above embodiments 1 of the application.

[0039] Figure 5 It is a structured data example diagram. DETAILED DESCRIPTION

[0040] The technical scheme of the application will be described in detail below with reference to the drawings and specific embodiments.

[0041] Embodiment 1: As shown in the figure, the marine engineering material intelligent identification method based on two-dimensional code identification technology, characterized in that it comprises the following steps: Figure 1

[0042] ​Step S101, constructing a multi-dimensional material feature information model: a feature database covering 28 key parameters in three categories of structural features, process features and protection features is established, ASN.1 abstract syntax notation standard is used to define the data structure, and cross-platform data compatibility is realized through basic encoding rule BER encoding and distinguishable encoding rule DER decoding;

[0043] The feature database covering 28 key parameters in three categories of structural features, process features and protection features includes:

[0044] Structural features: length, width, height, weight, material code, surface treatment grade, anticorrosion coating thickness, packaging grade; process features: welding requirement, heat treatment state, non-destructive testing type, pressure test parameter, cleanliness grade, traceability requirement; protection features: rust prevention grade, storage temperature and humidity range, validity period, dangerous goods category, anti-static requirement. ASN.1 standard is used to define the data structure, and cross-platform compatibility is ensured through BER encoding and DER decoding.

[0045] Step S102, generating a high-density two-dimensional code resistant to harsh environment: the structured data obtained in step S101 is encoded, and a three-way collaborative mechanism of block interleaving-dual verification-digital signature is used for encoding, which is an improved Reed-Solomon (255, 223) error correction coding scheme.

[0046] The improved Reed-Solomon (255,223) error correction coding scheme is processed as follows: first, the input data is the 32-byte data block obtained in step S101, which represents the characteristic information of the material. Then, the original data block is divided into an 8x32-byte matrix for interleaved storage to ensure that damaged data can be dispersed in different blocks, enhancing the recovery capability. Then, the Reed-Solomon (255,223) coding scheme is applied to convert the original data into an encoded block containing redundant data, which can recover up to 32 bytes of data loss. To further enhance the integrity of the data, a CRC32 check code is added to check each data block. At the same time, a 64-byte digital signature is generated using the ECDSA algorithm to ensure the security and tamper resistance of the data. Finally, the output data block will be used to generate a two-dimensional code label, ensuring that the two-dimensional code still has high fault tolerance and data recovery capability even in harsh environments. The three-layer fault tolerance mechanism is constructed in cooperation, including: dividing the original data into 32-byte logical blocks and recombining them into an 8x32-byte matrix for interleaved storage; adding an 8-byte RS check code to each logical block and fusing a cyclic redundancy check CRC32 for double checking; generating a 64-byte digital signature using the Elliptic Curve Digital Signature Algorithm ECDSA based on the secp256r1 curve of elliptic curve cryptography, and storing the signature private key in the hardware security module HSM; wherein the Reed-Solomon (255,223) uses forward error correction to detect and correct errors in the transmission or storage process by introducing redundancy into the data; therefore, the core improvement of the improved Reed-Solomon (255,223) error correction coding scheme lies in the introduction of the block interleaving-dual checking-digital signature triple cooperation mechanism; wherein:

[0047] Data block interleaving storage is to arrange data blocks in a matrix to disperse continuous contamination into different error correction blocks.

[0048] The original data is divided into 32-byte logical blocks and recombined into an 8x32-byte matrix to disperse continuous contamination into different encoded blocks.

[0049] The RS check code combines with the CRC32 double check to form a redundancy check mechanism, supporting complete recovery after up to 4 blocks of data loss.

[0050] The private key is stored in the HSM hardware security module certified by the National Cryptography Bureau through the encryption signature layer to ensure data tamper resistance and identity authentication.

[0051] The two-dimensional code label adopts the PET substrate + resin carbon tape + film coating process: first, polyethylene terephthalate (PET) is selected as the substrate of the label to ensure its strong high temperature resistance and corrosion resistance. Then, print using resin carbon tape, firmly print the two-dimensional code pattern on the PET surface through industrial-grade thermal transfer printing technology, control the printing temperature between 120-150°C, and ensure that the printed two-dimensional code has good adhesion. Next, a layer of protective film is added to the printed two-dimensional code using a hot pressing film coating process, with a film thickness of about 0.05mm, a hot pressing temperature of 100°C-120°C, and a pressure of 3-5bar, to ensure that the protective film is evenly and firmly attached to the surface of the two-dimensional code, improving the label's scratch and stain resistance. Finally, after 96 hours of salt spray testing, the label's durability in harsh environments such as salt spray and humidity is ensured, and the two-dimensional code's recognition rate must be no less than 99.5% after testing, ensuring long-term readability in harsh conditions such as marine engineering. The recognition rate is ≥99.5% under the conditions of -20°C~60°C and relative humidity of 20%~95% after 96 hours of salt spray testing.

[0052] In step S103, the mobile terminal intelligently identifies: a pre-processing enhancement-multi-model fusion-dynamic decision three-level identification architecture is constructed, the collected two-dimensional code image is pre-processed and enhanced, a lightweight MobileNetV3-Small deep learning model is used for identification, and the image recognition result, radio frequency identification RFID signal strength and ultra-wide band UWB positioning information are fused for multi-modal dynamic decision to output the final identification result; wherein:

[0053] The pre-processing enhancement-multi-model fusion-dynamic decision three-level identification architecture is adapted to the harsh environment of the offshore platform:

[0054] As shown in Figure 2 , it is the working process of the pre-processing enhancement-multi-model fusion-dynamic decision three-level identification architecture, including its composition, input, output and processing process. The architecture is a progressive image recognition and decision system designed for complex industrial scenarios, and its core idea is to gradually purify, identify and fuse multi-party information through multi-level processing, and finally output high-reliability results. The working process is described in detail as follows:

[0055] First level: pre-processing enhancement:

[0056] The input is the original two-dimensional code image captured from the camera of the mobile device. This image often has quality problems unique to the marine environment, such as geometric distortion caused by shooting angle, noise caused by salt spray / oil stains, low contrast caused by uneven lighting, and motion blur caused by ship heaving.

[0057] The processing process performs the following operations in turn:

[0058] Distortion correction: Based on Zhang Zhengyou's calibration method, correct lens and perspective distortion to ensure the regular shape of the QR code, error <0.5 pixels.

[0059] Noise filtering: Use a 5x5 bilateral filter to remove salt and pepper noise and Gaussian noise while preserving the edge information of the QR code.

[0060] Dynamic threshold binarization: Use adaptive threshold algorithm T=μ+k×σ to dynamically calculate the binarization threshold according to the local mean and standard deviation of the image, solving the problem of uneven lighting.

[0061] Contrast enhancement: Use CLAHE algorithm to enhance the overall contrast of the image, making the QR code module more distinct from the background.

[0062] Anti-blurring processing: Use Wiener filter to specifically repair motion blur and defocus blur, restoring image clarity.

[0063] Output is an enhanced, high-quality, clear QR code image, laying the foundation for accurate recognition.

[0064] Second level: multi-model fusion:

[0065] Input is the enhanced image output by the first level.

[0066] Multi-source information from other sensors, such as: signal strength read by RFID reader (as an auxiliary verification of material proximity), accurate position information provided by UWB positioning system (accuracy ±10cm, used to verify whether the material is in the expected work area).

[0067] The processing process performs the following operations in turn:

[0068] Lightweight AI model recognition: input the enhanced image into a lightweight deep learning model (such as MobileNetV3-Small) that has been deeply optimized (quantization, distillation, pruning) for inference, output the preliminary recognition result of the QR code (i.e. the decoded data) and a confidence score (indicating the model's degree of confidence in this recognition result).

[0069] Multi-source information fusion and weighted voting: the system does not simply rely on image recognition results. It will associate the AI model's confidence with RFID signal strength and UWB position information. For example, when the UWB device is in front of the target material and the RFID signal is very strong, but the image recognition is obscured by oil stains, resulting in low confidence, the system will dynamically improve the final comprehensive confidence through weighted algorithms.

[0070] Output is the preliminary recognition result and its comprehensive confidence after weighted by multi-source information.

[0071] Level 3: Dynamic Decision-Making

[0072] The input consists of the preliminary identification results and overall confidence level of the second-level output.

[0073] Processing procedure: This is a threshold-based intelligent decision-making logic:

[0074] The system sets a dynamic confidence threshold, which can be fine-tuned based on historical success rates and current environmental parameters (such as light intensity).

[0075] Compare the overall confidence level of the second-level output with this threshold:

[0076] If the overall confidence level is greater than the threshold: the recognition is considered successful, and the recognition result is output.

[0077] If the overall confidence level is less than or equal to the threshold, the system determines that the reliability of this identification is insufficient. Instead of directly reporting an error, the system triggers a feedback mechanism. For example, it prompts the user to adjust the angle, clean the label, or rescan, and then returns to the first level to restart the processing flow. This design avoids job interruptions caused by a single accidental failure.

[0078] The output is as follows:

[0079] Success, ultimately resulting in accurate and error-free material characteristic information data.

[0080] Failure / Retry: Provide clear reasons for failure or prompts for rescanning to guide user interaction.

[0081] The innovation of this three-tier architecture lies in its formation of a self-optimizing, fault-tolerant closed-loop system. The preprocessing stage enables recognition; multi-model fusion overcomes the limitations of single visual recognition by utilizing information redundancy; and the dynamic decision-making mechanism provides the system with decision-making capabilities in complex environments, ultimately achieving outstanding performance with high recognition rate, high robustness, and high availability in the harsh environment of marine engineering.

[0082] Furthermore, the image preprocessing includes five layers of enhancement processing performed sequentially: First, distortion correction is performed based on Zhang's calibration method, with a correction error of less than 0.5 pixels; second, a 5×5 bilateral filter is used to remove salt-and-pepper noise and Gaussian noise; next, adaptive dynamic threshold binarization is performed based on the local image mean μ and standard deviation σ, with a threshold T = μ + k × σ, where the value of k is dynamically adjusted from 0.5 to 2.0; then, the contrast is enhanced using the contrast-limited adaptive histogram equalization (CLAHE) algorithm, with parameters clipLimit = 2.0 and tileGridSize = 8×8; finally, Wiener filtering is used to repair motion blur with a displacement of no more than 5 pixels and defocus blur with a radius of no more than 3 pixels, achieving an image restoration rate of greater than 95%.

[0083] The lightweight deep learning model adopts a MobileNetV3-Small architecture, and is optimized by model quantization to INT8 precision, knowledge distillation with a ResNet50 as a teacher model, and 40% channel pruning, so that the final model size is compressed to 3.2MB, the inference time on a mobile terminal is 18 milliseconds, and the recognition accuracy is 99.5%.

[0084] The multi-modal fusion decision dynamically fuses image recognition results, RFID signal strength, and UWB positioning information with an accuracy of ±10cm through a weighted voting mechanism, and adjusts the recognition confidence threshold accordingly, so that the recognition rate is improved to 95.7% in a strong light and shadow scene, and the recognition rate is 89.3% in an oil pollution scene.

[0085] Multi-modal fusion decision: combine image recognition, RFID signal strength, and UWB positioning information (accuracy ±10cm), dynamically adjust the recognition confidence threshold through a weighted voting mechanism, and the recognition rate in a strong light and shadow scene is improved to 95.7%, and the recognition rate in an oil pollution scene is 89.3%.

[0086] Step S104, system integration and business linkage: integrate the recognition results obtained in step S103 with the intelligent warehouse management system to realize automatic triggering of business processes by scanning codes, driving the bin optimization algorithm based on the parsed material feature information, and updating the digital twin state within 500 milliseconds; this step realizes traceable batching and inventory management in the whole process, and realizes business process automation and visualization; wherein:

[0087] Embodiment 2: as shown in Figure 3 and Figure 4 The two-dimensional code recognition technology-based marine engineering material intelligent identification system for implementing the two-dimensional code recognition technology-based marine engineering material intelligent identification method of any one of the above embodiments of the application comprises:

[0088] The feature information modeling module 201 is used to execute the step S101 and construct the multi-dimensional material feature information model;

[0089] The two-dimensional code generation and label making module 202 is used to execute the step S102 to generate the high-density two-dimensional code and control an industrial printer to complete printing on an anti-corrosion label;

[0090] The mobile terminal intelligent identification module 203 is integrated in a mobile terminal and is used to execute the step S103 to implement the three-level recognition architecture;

[0091] The intelligent warehouse integration module 204 is used to execute the step S104 to realize data interaction and business process driving with a background management system.

[0092] Further, the mobile terminal intelligent recognition module 203 comprises:

[0093] The preprocessing unit 2031 is configured to perform distortion correction, noise filtering, dynamic threshold binarization, contrast enhancement and anti-blurring processing.

[0094] The lightweight model recognition unit 2032 is configured to load and run the optimized MobileNetV3-Small model to perform two-dimensional code positioning and decoding.

[0095] The multi-modal decision unit 2033 is configured to receive and fuse the output of the lightweight model recognition unit (2032), the RFID reader signal and the UWB positioning module information, and output the final recognition result through a weighted voting mechanism.

[0096] Further, the intelligent warehouse integration module 20) comprises a digital twin synchronization mechanism based on state hash verification, which ensures that the state synchronization delay between the physical inventory and the digital twin is less than 500 milliseconds, and the data consistency reaches 99.99%.

[0097] Further, the system further comprises a security audit module 205, which is configured to implement two-factor authentication based on OAuth 2.0 and OpenID Connect OIDC protocol, attribute-based access control ABAC, transport layer security TLS 1.3 and advanced encryption standard AES-256 encryption, and store audit logs in write-once-read-many WORM mode for 365 days.

[0098] The application will be described in detail below in combination with the actual application of a certain marine engineering module construction project.

[0099] 1. Structural pipe fitting identification and recognition embodiment:

[0100] Step one, feature information modeling and coding For the pipe fitting with material API 5L X70, specification Φ610x22.2mm, length 12.5m, corrosion grade Sa2.5, and welding requirement AWS D1.1, the system extracts 28 parameters to generate ASN.1 structured data, as shown in Figure 5 .

[0101] After the data is encoded by DER, a total of 468 bytes are obtained, and a 64-byte ECDSA signature is added to form a 532-byte payload.

[0102] Step 2: QR Code Generation and Label Production uses QRCODE mode, version V25 (129×129 module), error correction level H (30% redundancy). After encoding, a QR code image is generated and output via a Zebra ZT610 industrial printer. The label substrate is 3M 7847 polyester material with a surface coating thickness of 0.1mm. The label size is 50×50mm, and it is affixed to the pipe fitting 0.5m from the end to ensure visibility during hoisting.

[0103] Step 3: Mobile Terminal Recognition and Business Linkage. On-site personnel use handheld terminals equipped with Honeywell CT60 scanners to scan tags within a 10-meter range. Recognition process: After the camera captures the image, the system automatically performs distortion correction and CLAHE enhancement, taking 28ms; the MobileNetV3 model locates and decodes the QR code, achieving a recognition accuracy of 99.2%, with a total processing time of 65ms; the parsed feature information is displayed on the APP interface, and the system automatically verifies the BOM list, generating an inbound task after confirmation; the intelligent storage location recommendation algorithm, based on weight (1.2 tons) and size, allocates the storage location to a load-bearing location on the 3rd floor of Zone B in the indoor warehouse, reducing the handling distance by 58% compared to random allocation.

[0104] 2. Reliability verification under extreme environments:

[0105] Environmental adaptability testing was conducted at the prefabrication site of the offshore platform.

[0106] Salt spray test: The label was placed in a salt spray test chamber (5% NaCl solution, 35℃) for 96 hours. No rust was found on the label surface and the QR code recognition rate remained at 100% (100 test samples).

[0107] Oil contamination test: Crude oil was applied to cover 50% of the label surface. After CLAHE enhancement, the recognition rate reached 89.3%, which is 24.1 percentage points higher than the traditional ZBar algorithm.

[0108] Vibration and bending test: The label was attached to a steel pipe on a vibration table (frequency 10-50Hz, amplitude 2mm). After 2 hours of vibration, the QR code was deformed with a radius of curvature of 5cm due to the bending of the steel pipe. The recognition rate of this solution was 96.3%, while the traditional solution could not recognize it.

[0109] In summary, verification has shown that this invention improves upon the following performance aspects:

[0110] Information density increased by 180%: A single QR code can store 12,000 characters and directly carry 28 material characteristic parameters, improving on-site query efficiency by 70% and reducing the frequency of system interactions;

[0111] Excellent environmental adaptability: The recognition success rate is ≥98% under the conditions of 96-hour salt spray, 30% pollution, and 20% bending deformation, which is 13 percentage points higher than traditional technologies;

[0112] Significant enhancement of recognition robustness: The average recognition time of mobile terminals is 65 ms, the recognition rate under complex lighting conditions is >95%, and the recognition rate under motion blur caused by ship pitching is >96%, fully meeting the real-time operation requirements of marine engineering;

[0113] Safety and compliance meet the standards of classification societies: ECC digital signature and HSM key management ensure data tamper resistance, comply with ISO 27001 and ship industry data protection specifications, and support complete quality traceability;

[0114] Significant economic benefits: In a certain FPSO module construction project pilot, material search time is reduced by 70%, inventory time is reduced by 80%, inventory accuracy is improved to 99.9%, and annual labor cost is saved by about 4.5 million yuan.

[0115] The above describes specific embodiments of the present specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or can be advantageous.

[0116] The present application is not limited to the above-described embodiments. The above description of specific embodiments is intended to describe and illustrate the technical solutions of the present application, and the specific embodiments described above are merely illustrative and not limiting. Without departing from the purpose of the present application and the scope protected by the claims, those skilled in the art can make many forms of specific changes under the inspiration of the present application, which are all within the scope of protection of the present application.

Claims

1. A method for intelligent identification of marine engineering materials based on QR code recognition technology, characterized in that, Includes the following steps: Step S101: Construct a multi-dimensional material feature information model: Establish a feature database covering 28 key parameters in three major categories: structural features, process features, and protection features. Use the ASN.1 abstract syntax markup standard to define the data structure and achieve cross-platform data compatibility through the basic encoding rule BER encoding and the discriminative encoding rule DER decoding. Step S102, generating a high-density QR code resistant to harsh environments: The structured data obtained in step S101 is encoded using a triple collaborative mechanism of block interleaving-double verification-digital signature, including: dividing the original data into logical blocks and interleaving multiple logical blocks in a matrix form to logically disperse continuous physical damage; converting the interleaved data into encoded blocks containing redundant data and adding a check code to each data block to achieve fast integrity verification; generating a digital signature using an elliptic curve digital signature algorithm, with the signature private key stored in a hardware security module; the final data block after the above processing is used to generate a QR code graphic, which is printed on a physical carrier using an anti-corrosion material; Step S103, Mobile terminal intelligent recognition: Construct a three-level recognition architecture of preprocessing enhancement - multi-model fusion - dynamic decision-making. The collected QR code image is preprocessed and enhanced, and a lightweight MobileNetV3-Small deep learning model is used for recognition. The image recognition results, radio frequency identification RFID signal strength and ultra-wideband UWB positioning information are fused to make multimodal dynamic decisions and output the final recognition result. Step S104, System Integration and Business Linkage: Integrate the identification results obtained in step S103 with the intelligent warehouse management system to realize automatic business process triggering by scanning, drive the storage location optimization algorithm based on the parsed material feature information, and update the status of the digital twin at a set time.

2. The intelligent identification method for marine engineering materials based on QR code recognition technology according to claim 1, characterized in that, In step S102, the QR code is printed on a label made of polyethylene terephthalate (PET) substrate, resin ribbon and coating process. The label passes a 96-hour neutral salt spray test and has a readability of not less than 99.5% in an ambient temperature range of -20°C to 60°C and a relative humidity range of 20% to 95%.

3. The intelligent identification method for marine engineering materials based on QR code recognition technology according to claim 1 or 2, characterized in that, The improved Reed-Solomon encoding scheme supports complete recovery of up to four logical blocks of data after loss; under conditions of 30% random contamination, 50% oil coverage, 20% edge damage, and bending deformation with a radius of curvature of 5 cm, the recognition success rate of the QR code is no less than 98.7%, 85.2%, 98.7%, and 96.3%, respectively.

4. The intelligent identification method for marine engineering materials based on QR code recognition technology according to claim 1, characterized in that, In step S103, the preprocessing enhancement includes distortion correction based on Zhang Zhengyou calibration method, with an error of less than 0.5 pixels; noise removal using bilateral filtering with a 5×5 kernel; and adaptive dynamic threshold binarization based on the local image mean μ and standard deviation σ, where the threshold T = μ + k × σ, and the value of k is adaptively adjusted within the range of 0.5 to 2.0 according to the ambient light intensity. Contrast enhancement is achieved using the CLAHE algorithm, which employs contrast-limited adaptive histogram equalization. Wiener filtering is used to perform anti-blurring on motion blur with a displacement of no more than 5 pixels and defocus blur with a radius of no more than 3 pixels, with a recovery rate of more than 95%.

5. The intelligent identification method for marine engineering materials based on QR code recognition technology according to claim 1 or 4, characterized in that, After INT8 quantization, knowledge distillation based on the teacher model ResNet50, and 40% channel pruning optimization, the MobileNetV3-Small model is compressed to 3.2MB, with an inference time of 18 milliseconds on mobile platforms and a recognition accuracy of 99.5%.

6. The intelligent identification method for marine engineering materials based on QR code recognition technology according to claim 1, characterized in that, In step S103, the multimodal fusion decision dynamically adjusts the recognition confidence threshold through a weighted voting mechanism, thereby increasing the recognition rate to 95.7% in strong light and shadow scenarios, reaching 89.3% in oil pollution scenarios, and exceeding 96% in motion blur scenarios caused by ship turbulence.

7. A marine engineering material intelligent identification system based on QR code recognition technology for implementing the intelligent identification method for marine engineering materials based on QR code recognition technology according to any one of claims 1 to 6, characterized in that, include: The feature information modeling module is used to execute step S101 and construct the multi-dimensional material feature information model. The QR code generation and label production module is used to perform step S102, generate the high-density QR code and control the industrial printer to complete the printing on the anti-corrosion label. The mobile intelligent recognition module is integrated into the mobile terminal and is used to execute the step S103 to realize the three-level recognition architecture; The intelligent warehouse integration module is used to execute step S104 to realize data interaction and business process driving with the back-end management system.

8. The intelligent identification system for marine engineering materials based on QR code recognition technology according to claim 7, characterized in that, The mobile terminal intelligent recognition module includes: The preprocessing unit is used to perform the distortion correction, noise filtering, dynamic threshold binarization, contrast enhancement, and anti-blurring processing. A lightweight model recognition unit is used to load and run the optimized MobileNetV3-Small model for QR code localization and decoding. The multimodal decision unit is used to receive and fuse the output of the lightweight model recognition unit (2032), the RFID reader signal and the UWB positioning module information, and output the final recognition result through a weighted voting mechanism.

9. The intelligent identification system for marine engineering materials based on QR code recognition technology according to claim 7 or 8, characterized in that, The intelligent warehousing integration module includes a digital twin synchronization mechanism based on state hash verification, ensuring that the state synchronization delay between physical inventory and digital twin is less than 500 milliseconds, and the data consistency reaches 99.99%.

10. The intelligent identification system for marine engineering materials based on QR code recognition technology according to claim 7, characterized in that, It also includes a security audit module, which implements two-factor authentication based on OAuth 2.0 and the Open ID Connectivity (OIDC) protocol, attribute-based access control (ABAC), Transport Layer Security (TLS) 1.3 and Advanced Encryption Standard (AES-256) encryption, and stores audit logs for 365 days using a Write-once-Read-Many (WORM) mode.

Images