Container body surface defect detection method, system, equipment, medium and product

Abstract

The invention provides a container body surface defect detection method, system and device, a medium and a product, and relates to the technical field of container detection.The method comprises the steps that container multi-source data collected by a portable handheld three-dimensional scanner, an unmanned aerial vehicle and a multispectral device are obtained; generating a high-precision three-dimensional digital model through a feature adaptive weighted point cloud fusion algorithm; based on the model and the multispectral texture, identifying and quantifying defect three-dimensional parameters through a multi-scale convolutional neural network; and finally, automatically judging a defect level through a digital rule engine and generating a report. According to the method, the problem that automatic and high-precision quantitative detection of the surface defects of the container is difficult in a scattered network point environment is solved, operation by non-professionals is realized, a quantitative detection report meeting the international standard is automatically output in the whole process, and the detection precision, efficiency and standardization level are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of container inspection technology, and in particular to a method, system, equipment, medium and product for detecting defects on the surface of a container. Background Technology

[0002] As a core carrier of global logistics, the structural safety of containers is of paramount importance. Traditional container surface defect inspection mainly relies on inspectors' visual inspection and simple tools (such as calipers and depth gauges), which is not only inefficient and highly subjective, but also heavily dependent on the inspectors' professional experience. In recent years, with the development of computer vision technology, automated inspection solutions based on two-dimensional images or videos have emerged. However, these methods generally suffer from fundamental shortcomings such as limited data dimensions and the inability to accurately quantify the three-dimensional geometric parameters of defects (such as dent depth and volume). Especially for scattered locations such as docks and railway stations that lack fixed high-precision inspection facilities, it is impossible to deploy large-scale automated inspection equipment, and it is difficult to guarantee the professional level of on-site operators. This makes it difficult to guarantee inspection accuracy, efficiency, and standard consistency, becoming a prominent technical bottleneck restricting the intelligent upgrading of the industry. Summary of the Invention

[0003] To address the core technical problem of the inability to perform high-precision, automated three-dimensional quantitative detection of container surface defects in a dispersed network environment, as described in the background, this invention provides a method, system, equipment, medium, and product for detecting container surface defects by introducing a collaborative technical solution of mobile multimodal data acquisition, cloud-based adaptive weighted fusion of features, and intelligent analysis.

[0004] To achieve the above objectives, a first aspect of the present invention provides a method for detecting defects on the surface of a container body, comprising: Acquire first point cloud data of the container side collected by a portable handheld 3D scanner, second point cloud data of the container top collected and processed by a drone, and surface texture information collected by a multispectral imaging device; Based on the first point cloud data and the second point cloud data, a feature-adaptive weighted point cloud fusion algorithm is used to perform registration and fusion to generate a complete three-dimensional digital model of the container. Based on the complete three-dimensional digital model and the surface texture information, a multi-scale convolutional neural network model is used for defect identification and segmentation, and the identified defects are quantized in three-dimensional geometric parameters to obtain defect parameters. The quantified defect parameters are input into a digital rule engine with built-in container inspection standards, which automatically outputs defect level judgment and maintenance suggestions, and generates a digital inspection report.

[0005] Furthermore, the registration and fusion using the feature-adaptive weighted point cloud fusion algorithm includes an improved iterative nearest-point algorithm with the objective function being: ; in, Let be the transformation matrix to be solved. and To match point pairs, For adaptive weights, and satisfying: ; in, The accuracy weights are determined based on the nominal accuracy of the scanning equipment. The feature weights are determined based on the point cloud feature type. This is a consistency weight that is dynamically adjusted based on the matching residuals.

[0006] Furthermore, before performing defect identification and segmentation using a multi-scale convolutional neural network model, a data preprocessing step is also included: The three-dimensional point cloud of the complete three-dimensional digital model is projected onto multiple orthogonal planes to generate depth maps for each plane; The multispectral image in the surface texture information is registered to the corresponding orthogonal plane to obtain a multichannel texture map; The depth map and the corresponding multi-channel texture map are stitched together along the channel dimension to form a multi-channel fused image, which is then used as the input to the multi-scale convolutional neural network model.

[0007] Furthermore, the quantization of three-dimensional geometric parameters for the identified defects specifically includes: For dent defects, the maximum distance from points within the region to the fitted reference plane is calculated in three-dimensional space as the depth, and the sum of the signed volumes of each triangular facet in the region to the reference plane is calculated as the volume. For defects such as rust or paint peeling, the sum of the surface areas of the corresponding triangular meshes on the surface of the 3D model is calculated as the area. For crack defects, the defect region is skeletonized in three-dimensional space to extract the center line, and the total length of the center line is calculated as the length.

[0008] Furthermore, the decision logic of the digital rule engine is configured to execute the following rules: If the dent depth in the defect parameters is greater than the first threshold, the corresponding defect is determined to be a major defect and a suggestion for cold repair or panel replacement is output. If the dent depth in the defect parameters is less than or equal to the first threshold and greater than the second threshold, the corresponding defect is determined to be a moderate defect and a cold repair suggestion is output. If the dent depth in the defect parameters is less than or equal to the second threshold, the corresponding defect is determined to be a minor defect. If the defect location information indicates that the defect is located within a preset range near the corner fitting of the container, the level of the defect is upgraded.

[0009] Furthermore, the step of generating a complete 3D digital model also includes global scale correction: The standard dimensions of the container are used as constraints and added to the pose graph containing all scan poses for global optimization to correct the absolute scale of the complete 3D digital model.

[0010] A second aspect of the present invention provides a container body surface defect detection system, comprising: The data acquisition module is used to acquire the first point cloud data of the side of the container collected by a portable handheld 3D scanner, the second point cloud data of the top of the container collected and processed by a drone, and the surface texture information collected by a multispectral imaging device. The 3D reconstruction module is used to register and fuse the first point cloud data and the second point cloud data using a feature-adaptive weighted point cloud fusion algorithm to generate a complete 3D digital model of the container. The defect quantization module is used to identify and segment defects based on the complete three-dimensional digital model and the surface texture information through a multi-scale convolutional neural network model, and to quantify the three-dimensional geometric parameters of the identified defects to obtain defect parameters. The decision report module is used to input the quantified defect parameters into a digital rule engine with built-in container inspection standards, automatically output defect level judgment and maintenance suggestions, and generate a digital inspection report.

[0011] Embodiment 3 of the present invention provides a container body surface defect detection device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, it implements the steps of the container body surface defect detection method as described in the first aspect of the present invention. The device also includes: a portable handheld 3D scanner, which is communicatively connected to the processor and used to collect 3D point cloud data from the side of the container; The drone, which is connected in communication with the processor, is used to collect image data of the top of the container; A multispectral imaging device, which is communicatively connected to the processor, is used to acquire multispectral texture information of the container surface.

[0012] A fourth aspect of the present invention provides a computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the steps of the container body surface defect detection method as described in the first aspect of the present invention.

[0013] A fifth aspect of the present invention provides a computer program product including software code, wherein the program in the software code performs the steps of the container body surface defect detection method as described in the first aspect of the present invention.

[0014] Compared with existing technologies, the present invention provides a method, system, equipment, medium, and product for detecting defects on the surface of a container body, which has the following beneficial effects: (1) This invention constructs a highly integrated and portable container surface defect detection system, which encodes the professional inspection process into an augmented reality (AR) guidance interface and a fully automatic drone flight control program on a mobile terminal, enabling ordinary staff to quickly complete data collection in complex sites without professional training. This breaks the dependence of traditional inspection on fixed facilities and professional personnel, and realizes efficient and flexible deployment and application in dispersed network sites.

[0015] (2) This invention achieves high-precision 3D reconstruction of large-scale containers. By introducing adaptive weights determined by accuracy, features, and consistency for point cloud registration, the robustness of fusion between handheld fine point clouds and UAV sparse point clouds on low-texture surfaces is significantly improved. Furthermore, by incorporating the known standard dimensions of the container as hard constraints into global optimization, the scale distortion caused by cumulative errors is corrected. This leads to the construction of a digital twin model capable of generating high geometric accuracy and absolute scale accuracy, providing a reliable digital foundation for subsequent sub-millimeter-level defect quantification.

[0016] (3) This invention achieves accurate identification of various defects and automated measurement of three-dimensional parameters. It integrates depth information characterizing geometric deformation with multispectral information sensitive to material anomalies at the channel level and inputs it into a multi-scale CNN, enabling the network to comprehensively utilize shape and texture features. After identification, it directly performs accurate geometric calculations in three-dimensional space (such as calculating signed volume and surface area). Thus, it can automatically and objectively output key three-dimensional quantitative parameters such as depth, area, and volume of defects such as dents, corrosion, and cracks, completely replacing subjective and inefficient manual visual inspection and tool measurement.

[0017] (4) This invention achieves end-to-end automation from detection to decision reporting. The system automatically inputs the quantified defect parameters into the rule engine. The decision tree built into the engine simulates expert knowledge, compares the parameters with standards (such as IICL), automatically determines the defect level, matches the repair plan, and generates a structured report. The output is a standardized digital report that can be directly used for commercial damage assessment and repair decisions, which greatly improves the efficiency, consistency, and reliability of the inspection process and forms a complete intelligent inspection closed loop. Attached Figure Description

[0018] The accompanying drawings, which form part of this disclosure, are used to provide a further understanding of this disclosure. The illustrative embodiments of this disclosure and their descriptions are used to explain this disclosure and do not constitute an undue limitation of this disclosure.

[0019] Figure 1 This is a flowchart of a method for detecting defects on the surface of a container body provided in Embodiment 1 of the present invention; Figure 2 This is an architecture diagram of the container body surface defect detection system provided in Embodiment 2 of the present invention. Detailed Implementation

[0020] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0021] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, unless the context clearly indicates otherwise, the singular form is intended to include the plural form as well. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0022] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0023] All data acquisition in this embodiment is carried out in accordance with laws and regulations and with user consent, and the data is used legally.

[0024] Example 1 like Figure 1 This embodiment provides a method for detecting defects on the surface of a container body, including: Acquire first point cloud data of the container side collected by a portable handheld 3D scanner, second point cloud data of the container top collected and processed by a drone, and surface texture information collected by a multispectral imaging device; Based on the first point cloud data and the second point cloud data, a feature-adaptive weighted point cloud fusion algorithm is used to perform registration and fusion to generate a complete three-dimensional digital model of the container. Based on the complete three-dimensional digital model and the surface texture information, a multi-scale convolutional neural network model is used for defect identification and segmentation, and the identified defects are quantized in three-dimensional geometric parameters to obtain defect parameters. The quantified defect parameters are input into a digital rule engine with built-in container inspection standards, which automatically outputs defect level judgment and maintenance suggestions, and generates a digital inspection report.

[0025] The solution provided by this invention eliminates the reliance on fixed large-scale equipment and professional operators. It encodes professional inspection knowledge into an augmented reality (AR) guided interface and automated flight control program on a mobile terminal, allowing ordinary personnel to complete the entire data collection process simply by following visual prompts. This achieves the technical effect of enabling non-professionals to quickly deploy and complete high-integrity data collection on the entire surface (including the top) of a container in complex field environments, fundamentally breaking through the limitations of inspection scenarios.

[0026] Specifically, the registration and fusion using the feature-adaptive weighted point cloud fusion algorithm includes an improved iterative nearest-point algorithm with the following objective function: ; in, Let be the transformation matrix to be solved. and To match point pairs, For adaptive weights, and satisfying: ; in, The accuracy weights are determined based on the nominal accuracy of the scanning equipment. The feature weights are determined based on the point cloud feature type. This is a consistency weight that is dynamically adjusted based on the matching residuals.

[0027] More specifically, the point cloud fusion algorithm includes the following steps: 1) Feature extraction and description: Calculate the handheld point cloud separately. With drone point clouds For each point in the graph, the normal vector and curvature are used to extract edge points (those with curvature greater than a threshold). (points) and plane points (curvature less than) And points with high consistency of neighborhood normal vectors), and calculate fast point feature histogram descriptors for edge points, where, ; drone point cloud .

[0028] 2) Initial matching and outlier removal: Nearest neighbor matching is performed based on descriptors to obtain an initial set of corresponding point pairs. The initial rigid body transformation was estimated using the random sample consensus algorithm. To eliminate abnormal matches.

[0029] 3) Improved weighted ICP fine registration: Adaptive weights are introduced into the objective function. ,in: Precision weight For the scanner at the point The reciprocal of the nominal precision at that location; Feature weights For: If If it is an edge point, the value is 2.0; if it is a plane point, the value is 0.5; otherwise, the value is 1.0. Consistency weight Dynamic adjustment based on the matching residuals from the previous iteration.

[0030] The optimal transformation is solved iteratively using the Levberg-Marquardt algorithm. .

[0031] 4) Global optimization and scale correction: The known standard dimensions of the container are added as hard constraints to the pose graph containing the poses of all scanning stations. Global optimization is performed using a graph optimization library to correct the absolute scale of the complete 3D digital model.

[0032] To address the technical challenge of achieving high-precision registration and fusion of multi-source heterogeneous point clouds (such as high-precision handheld point clouds and sparse point clouds from UAV photogrammetry) on large-scale, low-texture container surfaces, this invention substantially improves the traditional Iterative Closest Point (ICP) algorithm. During the registration process, a precision weight is introduced for each matched point pair. Feature weights Consistency weight Adaptive weights Specifically, the precision weight adjusts for inherent precision differences between different devices; the feature weight strengthens the role of significant feature points such as edges in registration, suppressing interference from large planar low-texture areas; and the consistency weight dynamically eliminates abnormal matches during iteration. This mechanism effectively achieves differentiated treatment and optimized fusion of data with different reliability levels, significantly improving the robustness of multi-source point cloud registration in challenging environments and the overall reconstruction accuracy of the final 3D digital model, providing a foundation for subsequent sub-millimeter-level quantization.

[0033] Specifically, before performing defect identification and segmentation using a multi-scale convolutional neural network model, a data preprocessing step is also included: The three-dimensional point cloud of the complete three-dimensional digital model is projected onto multiple orthogonal planes to generate depth maps for each plane; The multispectral image in the surface texture information is registered to the corresponding orthogonal plane to obtain a multichannel texture map; The depth map and the corresponding multi-channel texture map are stitched together along the channel dimension to form a multi-channel fused image, which is then used as the input to the multi-scale convolutional neural network model.

[0034] Furthermore, the multi-scale convolutional neural network model adopts the U-Net++ architecture, and its input is a depth map. Red light texture map Green light texture map Blue light texture map and near-infrared texture map Seven-channel image stitched together The network is trained using a loss function that combines weighted cross-entropy loss and Dice loss. ,in, Give higher weight to rare defect categories.

[0035] This invention constructs a multimodal perception and analysis pipeline: First, 3D geometric information is transformed into a depth map through projection, preserving key deformation features; simultaneously, multispectral images (such as near-infrared bands, which are sensitive to early corrosion) are registered to the same viewpoint, providing rich surface material and spectral information. After being stitched together along the channel dimension, a fused data source containing geometric and texture information is formed and input into a multi-scale CNN. The network can simultaneously capture defect features at different scales.

[0036] Specifically, the quantization of three-dimensional geometric parameters of the identified defects includes: For dent defects, the maximum distance from points within the region to the fitted reference plane is calculated in three-dimensional space as the depth, and the sum of the signed volumes of each triangular facet in the region to the reference plane is calculated as the volume. Here, triangular facets refer to the tiny triangular network units that make up the surface of the three-dimensional model, which are the basis for all surface geometry calculations. The signed volume refers to assigning a positive or negative sign to the tiny prism volume formed by each triangular facet to the reference plane when calculating the dent volume, in order to distinguish between dents and protrusions. The sum of their absolute values ​​is the precise volume of the dent. For defects such as rust or paint peeling, the sum of the surface areas of the corresponding triangular meshes on the surface of the 3D model is calculated as the area. For crack defects, the defect region is skeletonized in three-dimensional space to extract the center line, and the total length of the center line is calculated as the length.

[0037] Based on the specific algorithms described above for directly calculating defect depth, signed volume, surface area, and length in three-dimensional space, the entire process of identifying, classifying, and accurately calculating key three-dimensional geometric parameters for various defects such as dents, corrosion, and cracks is automated, outputting objective and quantifiable defect parameters, replacing subjective visual inspection.

[0038] Specifically, the decision logic of the digital rule engine is configured to execute the following rules: If the dent depth in the defect parameters is greater than the first threshold (in one specific embodiment, the first threshold is 35mm), the corresponding defect is determined to be a major defect and a suggestion for cold repair or panel replacement is output. If the dent depth in the defect parameters is less than or equal to the first threshold and greater than the second threshold (in one specific embodiment, the second threshold is 25mm), the corresponding defect is determined to be a moderate defect and a cold repair suggestion is output. If the dent depth in the defect parameters is less than or equal to the second threshold, the corresponding defect is determined to be a minor defect. If the defect location information indicates that the defect is located within a preset range near the corner fitting of the container, the level of the defect is upgraded.

[0039] Furthermore, the judgment logic of the rule engine also includes composite judgment rules based on defect area and volume. For example, when the dent area exceeds a preset area threshold, the defect level will be upgraded to intermediate even if the depth does not meet the intermediate standard.

[0040] This invention transforms industry standards (such as IICL) into executable computer logic. Defect parameters quantified in the aforementioned steps (e.g., dent depth of 38.2mm) are input into a rule engine. The engine's built-in decision tree automatically executes predefined rules; for example, if the depth is greater than 35mm, it is considered a major defect, and cold repair or panel replacement is recommended. This process achieves automatic mapping from data to internationally standardized decisions and repair recommendations, generating structured digital reports. This ensures the objectivity, consistency, and direct applicability of inspection results, significantly improving the standardization level and decision-making efficiency of the inspection process.

[0041] Specifically, the step of generating a complete three-dimensional digital model also includes global scale correction: The standard dimensions of the container are used as constraints and added to the pose graph containing all scan poses for global optimization to correct the absolute scale of the complete 3D digital model.

[0042] After completing multi-station scan registration, a pose graph containing the spatial attitudes of all scan stations is constructed, with the known standard length of the container (e.g., 6.058 meters for a 20-foot container) added as a strong constraint. Graph optimization techniques (such as using the g2o library) are then used to globally adjust the attitudes and possible scale factors of all nodes. This technique achieves precise correction of the overall scale of the reconstructed model, eliminates scale distortion caused by accumulated errors, and ensures that the final output of quantitative parameters such as defect depth and area are accurate values ​​at a real-world scale, meeting the absolute accuracy requirements of industrial inspection.

[0043] In addition, considering the potential poor network conditions at dispersed sites, the method also includes network adaptive design: using octrees to compress point cloud data and supporting breakpoint resume and differential upload; supporting local lightweight processing on edge devices to generate preliminary reports when the network is interrupted; and automatically reducing the resolution of texture images before transmission while preserving the complete accuracy of depth maps and multispectral feature maps.

[0044] Example 2 like Figure 2 This embodiment provides a container body surface defect detection system, including: The data acquisition module is used to acquire the first point cloud data of the side of the container collected by a portable handheld 3D scanner, the second point cloud data of the top of the container collected and processed by a drone, and the surface texture information collected by a multispectral imaging device. The 3D reconstruction module is used to register and fuse the first point cloud data and the second point cloud data using a feature-adaptive weighted point cloud fusion algorithm to generate a complete 3D digital model of the container. The defect quantization module is used to identify and segment defects based on the complete three-dimensional digital model and the surface texture information through a multi-scale convolutional neural network model, and to quantify the three-dimensional geometric parameters of the identified defects to obtain defect parameters. The decision report module is used to input the quantified defect parameters into a digital rule engine with built-in container inspection standards, automatically output defect level judgment and maintenance suggestions, and generate a digital inspection report.

[0045] In one specific embodiment, the system adopts a three-layer architecture: a mobile acquisition terminal (end), an edge quality controller (edge), and a cloud-based intelligent processing center (cloud). The mobile acquisition terminal is a modular, portable acquisition kit, weighing no more than 30kg, and includes a handheld structured light 3D scanner (such as a blue LED structured light solution with an accuracy of 0.1mm), a miniature UAV oblique photography system (generating dense point clouds through five-lens synchronous oblique photography), and imaging accessories with integrated multispectral filter wheels. The edge quality controller is responsible for real-time monitoring of data quality and basic stitching. The cloud-based intelligent processing center integrates a dedicated algorithm library optimized for container inspection.

[0046] The mobile data acquisition terminal also includes an operation guidance system for non-professionals, including: superimposing a virtual scanning frame on an augmented reality guidance interface on a tablet computer and dynamically prompting the scanning area; providing prompts for scanning speed and distance through voice and haptic feedback; and after the drone takes off, visually recognizing the outline of the container and automatically planning the optimal flight path to achieve fully autonomous operation.

[0047] Example 3 Embodiment 3 of the present invention provides a container body surface defect detection device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the program, it implements the steps of the container body surface defect detection method as described in Embodiment 1 of the present invention.

[0048] The detailed steps are the same as those for the container surface defect detection method provided in Example 1, and will not be repeated here.

[0049] The device also includes: a portable handheld 3D scanner, which is communicatively connected to the processor and used to acquire 3D point cloud data of the side of the container; a drone, which is communicatively connected to the processor and used to acquire image data of the top of the container; and a multispectral imaging device, which is communicatively connected to the processor and used to acquire multispectral texture information of the container surface.

[0050] Furthermore, the device is integrated into a portable equipment case, weighing no more than 30 kg. A built-in battery powers all components and supports solar charging. The equipment case also includes a wireless positioning beacon for establishing a local coordinate system.

[0051] Example 4 Embodiment 4 of the present invention provides a computer-readable storage medium.

[0052] A computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps in the container body surface defect detection method as described in Embodiment 1 of the present invention.

[0053] The detailed steps are the same as those for the container surface defect detection method provided in Example 1, and will not be repeated here.

[0054] Example 5 Embodiment 5 of the present invention provides a computer program product.

[0055] A computer program product includes software code, wherein the program in the software code performs the steps of the container body surface defect detection method as described in Embodiment 1 of the present invention.

[0056] The detailed steps are the same as those for the container surface defect detection method provided in Example 1, and will not be repeated here.

[0057] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The solutions in the embodiments of the present invention can be implemented using various computer languages, such as the object-oriented programming language Java and the interpreted scripting language JavaScript.

[0058] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0059] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0060] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0061] The above description is merely a preferred embodiment of this practice and is not intended to limit the scope of this practice. Various modifications and variations can be made to this practice by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this practice should be included within the protection scope of this practice.

Claims

1. A method for detecting defects on the surface of a container body, characterized in that, include: Acquire first point cloud data of the container side collected by a portable handheld 3D scanner, second point cloud data of the container top collected and processed by a drone, and surface texture information collected by a multispectral imaging device; Based on the first point cloud data and the second point cloud data, a feature-adaptive weighted point cloud fusion algorithm is used to perform registration and fusion to generate a complete three-dimensional digital model of the container. Based on the complete three-dimensional digital model and the surface texture information, a multi-scale convolutional neural network model is used for defect identification and segmentation, and the identified defects are quantized in three-dimensional geometric parameters to obtain defect parameters. The quantified defect parameters are input into a digital rule engine with built-in container inspection standards, which automatically outputs defect level judgment and maintenance suggestions, and generates a digital inspection report.

2. The method as described in claim 1, characterized in that, The point cloud fusion algorithm using adaptive feature weighting for registration and fusion includes an improved iterative nearest-point algorithm with the following objective function: ； in, Let be the transformation matrix to be solved. and To match point pairs, For adaptive weights, and satisfying: ； in, The accuracy weights are determined based on the nominal accuracy of the scanning equipment. The feature weights are determined based on the point cloud feature type. This is a consistency weight that is dynamically adjusted based on the matching residuals.

3. The method as described in claim 1, characterized in that, Before performing defect identification and segmentation using a multi-scale convolutional neural network model, a data preprocessing step is also included: The three-dimensional point cloud of the complete three-dimensional digital model is projected onto multiple orthogonal planes to generate depth maps for each plane; The multispectral image in the surface texture information is registered to the corresponding orthogonal plane to obtain a multichannel texture map; The depth map and the corresponding multi-channel texture map are stitched together along the channel dimension to form a multi-channel fused image, which is then used as the input to the multi-scale convolutional neural network model.

4. The method as described in claim 1, characterized in that, The process of quantifying the three-dimensional geometric parameters of the identified defects specifically includes: For dent defects, the maximum distance from points within the region to the fitted reference plane is calculated in three-dimensional space as the depth, and the sum of the signed volumes of each triangular facet in the region to the reference plane is calculated as the volume. For defects such as rust or paint peeling, the sum of the surface areas of the corresponding triangular meshes on the surface of the 3D model is calculated as the area. For crack defects, the defect region is skeletonized in three-dimensional space to extract the center line, and the total length of the center line is calculated as the length.

5. The method as described in claim 1, characterized in that, The decision logic of the digital rule engine is configured to execute the following rules: If the dent depth in the defect parameters is greater than the first threshold, the corresponding defect is determined to be a major defect and a suggestion for cold repair or panel replacement is output. If the dent depth in the defect parameters is less than or equal to the first threshold and greater than the second threshold, the corresponding defect is determined to be a moderate defect and a cold repair suggestion is output. If the dent depth in the defect parameters is less than or equal to the second threshold, the corresponding defect is determined to be a minor defect. If the defect location information indicates that the defect is located within a preset range near the corner fitting of the container, the level of the defect is upgraded.

6. The method as described in claim 1, characterized in that, The step of generating a complete 3D digital model also includes global scale correction: The standard dimensions of the container are used as constraints and added to the pose graph containing all scan poses for global optimization to correct the absolute scale of the complete 3D digital model.

7. A container body surface defect detection system, characterized in that, include: The data acquisition module is used to acquire the first point cloud data of the side of the container collected by a portable handheld 3D scanner, the second point cloud data of the top of the container collected and processed by a drone, and the surface texture information collected by a multispectral imaging device. The 3D reconstruction module is used to register and fuse the first point cloud data and the second point cloud data using a feature-adaptive weighted point cloud fusion algorithm to generate a complete 3D digital model of the container. The defect quantization module is used to identify and segment defects based on the complete three-dimensional digital model and the surface texture information through a multi-scale convolutional neural network model, and to quantify the three-dimensional geometric parameters of the identified defects to obtain defect parameters. The decision report module is used to input the quantified defect parameters into a digital rule engine with built-in container inspection standards, automatically output defect level judgment and maintenance suggestions, and generate a digital inspection report.

8. A container body surface defect detection device, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the program, it implements the steps of the container body surface defect detection method as described in any one of claims 1 to 6; The device also includes: a portable handheld 3D scanner, which is communicatively connected to the processor and used to acquire 3D point cloud data of the side of the container; a drone, which is communicatively connected to the processor and used to acquire image data of the top of the container; and a multispectral imaging device, which is communicatively connected to the processor and used to acquire multispectral texture information of the container surface.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the container body surface defect detection method as described in any one of claims 1 to 6.

10. A computer program product, comprising software code, characterized in that, The program in the software code performs the steps of the container body surface defect detection method as described in any one of claims 1 to 6.

Images