CN-121981981-A - Method, system, equipment, medium and product for detecting surface defects of container body

CN121981981ACN 121981981 ACN121981981 ACN 121981981ACN-121981981-A

Abstract

The invention provides a method, a system, equipment, a medium and a product for detecting surface defects of a container body, and relates to the technical field of container detection, wherein the method comprises the steps of acquiring container multi-source data acquired by a portable handheld three-dimensional scanner, an unmanned aerial vehicle and multi-spectrum equipment; the method comprises the steps of generating a high-precision three-dimensional digital model through a point cloud fusion algorithm with characteristic self-adaptive weighting, identifying and quantifying three-dimensional parameters of defects through a multi-scale convolutional neural network based on the model and multispectral texture, and finally automatically judging the defect level through a digital rule engine and generating a report. The invention solves the difficult problems of difficult automation and high-precision quantitative detection of the surface defects of the container in the scattered network environment, realizes the operation by non-professional personnel, automatically outputs the quantitative inspection report meeting the international standard in the whole process, and remarkably improves the detection precision, efficiency and standardization level.

Inventors

KONG DEZHI
LIU BO
HAN CHAO
LIU JUAN

Assignees

鑫三利集装箱服务有限公司
鑫三利智慧集装箱服务有限公司

Dates

Publication Date: 20260505
Application Date: 20260116

Claims (10)

1. A method for detecting surface defects of a container body, comprising: acquiring first point cloud data of the side surface of a container acquired by a portable handheld three-dimensional scanner, second point cloud data of the top of the container acquired and processed by an unmanned aerial vehicle and surface texture information acquired by multispectral imaging equipment; registering and fusing by a characteristic self-adaptive weighted point cloud fusion algorithm based on the first point cloud data and the second point cloud data to generate a complete three-dimensional digital model of the container; performing defect identification and segmentation through a multi-scale convolutional neural network model based on the complete three-dimensional digital model and the surface texture information, and performing three-dimensional geometric parameter quantification on the identified defects to obtain defect parameters; inputting the quantized defect parameters into a digital rule engine with built-in container inspection standards, automatically outputting defect grade judgment and maintenance suggestions, and generating a digital inspection report.
2. The method of claim 1, wherein the registering and fusing by the feature-adaptive weighted point cloud fusion algorithm comprises employing a modified iterative closest point algorithm with an objective function of: ; Wherein, the For the transformation matrix to be solved, And In order to match the pairs of points, Is an adaptive weight and satisfies: ; Wherein, the For a precision weight determined based on the nominal precision of the scanning device, For feature weights determined based on the point cloud feature types, Is a consistency weight dynamically adjusted based on the matching residual.
3. The method of claim 1, further comprising the step of preprocessing data prior to defect identification and segmentation by the multi-scale convolutional neural network model: Projecting the three-dimensional point cloud of the complete three-dimensional digital model to a plurality of orthogonal planes, and generating a depth map of each plane; registering the multispectral image in the surface texture information to a corresponding orthogonal plane to obtain a multichannel texture map; and splicing the depth map and the multi-channel texture map of the corresponding plane in the channel dimension to form a multi-channel fusion image which is used as the input of the multi-scale convolutional neural network model.
4. The method according to claim 1, wherein said quantifying the three-dimensional geometrical parameters of the identified defects comprises: For dent defects, calculating the maximum distance from points in its region to a fitted reference plane in three-dimensional space as depth, and calculating the sum of the signed volumes of each triangular patch of the region to the reference plane as volume; For rust or paint layer falling defects, calculating the sum of the surface areas of the triangular meshes corresponding to the defects on the surface of the three-dimensional model as the area; for a crack defect, the defective region is skeletonized in a three-dimensional space to extract a center line, and the total length of the center line is calculated as a length.
5. The method of claim 1, wherein the decision logic of the digitizing rule engine is configured to execute the following rules: If the dent depth in the defect parameters is larger than a first threshold value, judging that the corresponding defect is a heavy defect and outputting a cold repair or panel replacement suggestion; If the dent depth in the defect parameters is smaller than or equal to a first threshold value and larger than a second threshold value, judging the corresponding defect as a moderate defect and outputting a cold repair proposal; if the dent depth in the defect parameters is smaller than or equal to a second threshold value, judging that the corresponding defect is a mild defect; and if the defect position information indicates that the defect is positioned in a preset range near the container corner fitting, the grade of the determined defect is improved.
6. The method of claim 1, wherein in the step of generating a complete three-dimensional digital model, further comprising global scale correction: and adding the standard size of the container as a constraint condition into a pose graph containing all scanning poses to perform global optimization so as to correct the absolute scale of the complete three-dimensional digital model.
7. A container surface defect detection system, comprising: The data acquisition module is used for acquiring first point cloud data of the side face of the container acquired by the portable handheld three-dimensional scanner, second point cloud data of the top of the container acquired and processed by the unmanned aerial vehicle and surface texture information acquired by the multispectral imaging equipment; The three-dimensional reconstruction module is used for registering and fusing the first point cloud data and the second point cloud data through a point cloud fusion algorithm with characteristic self-adaptive weighting to generate a complete three-dimensional digital model of the container; the defect quantification module is used for carrying out defect identification and segmentation through a multi-scale convolutional neural network model based on the complete three-dimensional digital model and the surface texture information, and carrying out three-dimensional geometric parameter quantification on the identified defects to obtain defect parameters; and the decision report module is used for inputting the quantized defect parameters into a digital rule engine with a container inspection standard, automatically outputting defect grade judgment and maintenance suggestions, and generating a digital inspection report.
8. A container body surface defect detection device comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that the processor implements the steps of the container body surface defect detection method according to any one of claims 1 to 6 when executing the program; the device further comprises a portable handheld three-dimensional scanner, an unmanned aerial vehicle and a multispectral imaging device, wherein the portable handheld three-dimensional scanner is in communication connection with the processor and is used for collecting three-dimensional point cloud data of the side face of the container, the unmanned aerial vehicle is in communication connection with the processor and is used for collecting image data of the top of the container, and the multispectral imaging device is in communication connection with the processor and is used for collecting multispectral texture information of the surface of the container.
9. A computer-readable storage medium, on which a computer program is stored, characterized in that the program, when being executed by a processor, implements the steps of the container body surface defect detection method according to any one of claims 1 to 6.
10. A computer program product comprising software code, characterized in that a program in said software code performs the steps of the container body surface defect detection method according to any one of claims 1 to 6.

Description

Method, system, equipment, medium and product for detecting surface defects of container body Technical Field The invention relates to the technical field of container detection, in particular to a method, a system, equipment, a medium and a product for detecting surface defects of a container body of a container. Background The container is used as a core carrier of global logistics, and the structural safety of the container is important. The traditional container surface defect detection mainly relies on manual visual observation of inspectors and measurement by simple tools (such as calipers and depth gauges), and is low in efficiency, high in subjectivity and severely depends on professional experience of inspectors. In recent years, with the development of computer vision technology, automatic detection schemes based on two-dimensional images or videos are developed, but these methods generally have the fundamental defects of single data dimension, incapacity of realizing accurate quantification of three-dimensional geometric parameters (such as dent depth and volume) of defects, and the like. Especially for the scattered network sites such as wharfs, railway stations and the like which do not have fixed high-precision detection facilities, large-scale automatic detection equipment cannot be deployed, the professional level of field operators is difficult to ensure, the detection precision, efficiency and standard consistency are difficult to ensure, and the method becomes an outstanding technical bottleneck for restricting the intelligent upgrading of industries. Disclosure of Invention In order to solve the core technical problem that the surface defects of the container cannot be detected in a high-precision and automatic three-dimensional quantification mode in a scattered network environment in the background technology, the invention provides a method, a system, equipment, a medium and a product for detecting the surface defects of the container body of the container by introducing a collaborative technical scheme of mobile terminal multi-mode data acquisition, cloud characteristic self-adaptive weighted fusion and intelligent analysis. To achieve the above object, a first aspect of the present invention provides a container surface defect detection method, including: acquiring first point cloud data of the side surface of a container acquired by a portable handheld three-dimensional scanner, second point cloud data of the top of the container acquired and processed by an unmanned aerial vehicle and surface texture information acquired by multispectral imaging equipment; registering and fusing by a characteristic self-adaptive weighted point cloud fusion algorithm based on the first point cloud data and the second point cloud data to generate a complete three-dimensional digital model of the container; performing defect identification and segmentation through a multi-scale convolutional neural network model based on the complete three-dimensional digital model and the surface texture information, and performing three-dimensional geometric parameter quantification on the identified defects to obtain defect parameters; inputting the quantized defect parameters into a digital rule engine with built-in container inspection standards, automatically outputting defect grade judgment and maintenance suggestions, and generating a digital inspection report. Further, the registering and fusing are performed by a point cloud fusion algorithm with characteristic self-adaptive weighting, which comprises the following steps of adopting an improved iterative nearest point algorithm, wherein the objective function is as follows: ; Wherein, the For the transformation matrix to be solved,AndIn order to match the pairs of points,Is an adaptive weight and satisfies: ; Wherein, the For a precision weight determined based on the nominal precision of the scanning device,For feature weights determined based on the point cloud feature types,Is a consistency weight dynamically adjusted based on the matching residual. Further, before the defect identification and segmentation are performed through the multi-scale convolutional neural network model, the method further comprises the step of data preprocessing: Projecting the three-dimensional point cloud of the complete three-dimensional digital model to a plurality of orthogonal planes, and generating a depth map of each plane; registering the multispectral image in the surface texture information to a corresponding orthogonal plane to obtain a multichannel texture map; and splicing the depth map and the multi-channel texture map of the corresponding plane in the channel dimension to form a multi-channel fusion image which is used as the input of the multi-scale convolutional neural network model. Further, the three-dimensional geometric parameter quantification of the identified defect specifically includes: For dent defects, calculating the maximum distance from points