CN114923413A - Automatic discrimination method for point cloud steel structure quality based on three-dimensional laser scanner - Google Patents

Abstract

The invention relates to a point cloud steel structure quality automatic judging method based on a three-dimensional laser scanner, which solves the problems that the existing tunnel steel structure construction quality inspection technology has larger subjective intention and can not completely express the tunnel steel structure construction quality through point cloud analysis, pretreatment, feature extraction, point cloud segmentation/single-point feature analysis, target identification, point cloud classification and judgment result acquisition.

Description

A method for automatic identification of steel structure quality based on point cloud of 3D laser scanner

technical field

The invention belongs to the technical field of tunnel engineering construction, in particular to a method for automatically judging the quality of steel bars of a steel frame based on a point cloud of a three-dimensional laser scanner.

Background technique

In the tunnel construction process, the initial supporting steel arch spacing, the horizontal and vertical deviation of the steel arch, the verticality of the steel arch, the spacing of the connecting bars of the steel arch, the spacing of the lining steel bars, the layer spacing of the lining steel bars, and the thickness of the lining steel bar protection layer are An important item of construction quality inspection. The traditional operation method is measured by the on-site quality inspector using a steel tape measure or using a total station, which is a point-to-surface inspection method.

Limited by the field operation platform, operation efficiency and internal data processing ability, the interval between sections and points on the same section are often large, and the measurement points are very random, and it is easy to occur that the measurement data cannot be truly and completely It reflects the actual situation of the site, such as the deviation data of the key arch spacing appears between the two points measured by the quality inspector. At the same time, this traditional operation method is also greatly affected by the experience and subjective wishes of the operators, which is not conducive to the standardization, informatization and intelligent construction of the safety and quality management of the tunnel construction process.

SUMMARY OF THE INVENTION

Due to the traditional prior art using a manual tape measure to measure the spacing between steel bars and steel arches, in tunnel construction projects, high-altitude operations are required in many places, resulting in greater safety hazards and longer time consuming, affecting on-site construction. Moreover, manual measurement is one-sided and cannot comprehensively display quality problems, which is a rough statistical method based on points and areas. After development and application of the present invention, manual measurement is not required, and a 3D scanner is used to scan the entire point cloud, which takes less time, has a high safety factor, and provides comprehensive results.

In view of the above problems, the present invention provides a method for automatically judging the quality of steel bars of a steel frame based on a point cloud of a three-dimensional laser scanner. This method solves the problem that the existing tunnel steel structure construction quality inspection technology has a large subjective will and cannot fully reflect the tunnel steel structure construction. quality issue.

The present invention is realized by the following measures:

A method for automatically judging the quality of a steel structure based on a point cloud of a three-dimensional laser scanner, the method comprises the following steps:

Step 1, point cloud analysis: use a 3D laser scanner to scan the target to obtain 3D point cloud data of the steel structure;

Step 2, preprocessing: import the 3D point cloud data into the software and preprocess the 3D point cloud data to obtain a steel structure point cloud model;

Step 3, feature extraction: plane extraction, edge detection and calculation of feature descriptors, which are mainly realized by artificial design and deep learning;

Step 4, point cloud segmentation: grouping points into a part or an object based on low-level attributes;

Step 5, target identification: identify the steel structure in the model;

Step 6, obtain the judgment result: analyze the steel structure to obtain the construction parameter data of the steel structure, so as to make an accurate judgment on the construction quality of the steel structure.

Further, after step 5, it also includes point cloud classification, classifying the point clouds into different point cloud sets, and dividing similar or identical attributes into the same point cloud set.

Further, in step 1, the target is a tunnel section.

Further, in step 2, the method for preprocessing the 3D point cloud data is: automatic filtering and downsampling according to the quality and scale of the point cloud data.

Further, in step 3, feature extraction is the key to characterize the morphological structure of the point cloud, and the basis and premise of semantic information extraction.

Further, in step 4, compared with processing or analyzing each point individually, the segmentation process further processes and analyzes each object, so that it has more abundant information.

Further, in step 5, target recognition is usually performed by performing analysis according to the results of feature extraction and segmentation, and is performed under given constraints and rules based on prior knowledge.

Further, in step 6, the construction parameter data are spacing, quantity, shape, diameter, length, and volume.

Further, the steel structure is a steel bar or a steel arch.

Further, the point cloud model in step 2 is obtained after noise reduction processing.

The method steps may be implemented or implemented by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer readable memory.

Further, the method also adds a hole filling method. Using this method, after triangulating the point cloud, a preset rule can be used to quickly fill the hole, and the normal phase constraint can be used to adjust the filling added in the initial filling process. The position of the point makes the shape of the filling part consistent with the surrounding mesh, so as to realize intelligent and rapid repair of holes, which not only ensures the naturalness of filling, but also ensures that the shape of the object to be measured will not change.

Beneficial effects of the present invention:

1. After the technology is applied to the actual construction on site, it can effectively improve the standardization and informatization of safety and quality management in the tunnel construction process, and reduce the risk of loss of quality control caused by the quality inspection staff;

2. The use of integrated and standardized management procedures can effectively reduce the demand for quality inspection personnel, simplify the personnel structure, and improve project benefits;

3. Visualize the measurement results, which can effectively record various problems in the construction process, effectively improve the detection rate of hidden dangers in the construction process, and avoid rework caused by quality problems in the later stage;

4. Perform noise reduction processing on the point cloud data to obtain a more effective point cloud data set.

5. The point cloud hole filling point can be used to quickly fill the hole with the preset rule after triangulating the point cloud, and use the normal phase constraint to adjust the position of the filling point added in the initial filling process, so that the filling part is the same as the filling point. The surrounding meshes are consistent in shape characteristics to achieve intelligent and rapid repair of holes, which not only ensures the naturalness of filling, but also ensures that the shape of the object to be measured will not change.

Description of drawings

Fig. 1 is a flow chart of point cloud processing of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, but not all of the embodiments.

The components of the embodiments of the invention generally described and illustrated in the drawings herein may be arranged and designed in a variety of different configurations. Thus, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but is merely representative of selected embodiments of the invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work fall within the protection scope of the present invention.

Example 1

Based on a method for automatically judging the quality of a steel structure from a point cloud of a 3D laser scanner, the steel structure is a steel bar or a steel arch, and the method includes the following steps:

Step 1, point cloud analysis: use a 3D laser scanner to scan the target, this target can be a tunnel section, and obtain the 3D point cloud data of the steel structure;

Step 2, preprocessing: import the 3D point cloud data into the software and preprocess the 3D point cloud data, automatically filter and downsample according to the quality and scale of the point cloud data, and obtain the point cloud model of the steel structure; It is closely connected to ensure that the point cloud data of the steel structure is accurately retained, and other interfering point clouds are removed. It has high requirements for the automatic filtering algorithm of the point cloud. It can truly achieve fully intelligent automatic filtering, which is not something that can be achieved by a single algorithm. The best solution is to use machine learning of a large amount of data. Before that, use the computer to initially filter, and then finely remove it manually;

Step 3: Feature extraction: plane extraction, edge detection and feature descriptor calculation are mainly realized by artificial design and deep learning; feature extraction is the key to characterize the morphological structure of point clouds, and the basis and premise of semantic information extraction;

Step 4, point cloud segmentation and/or single point feature analysis: grouping points into a part or an object based on low-level attributes; point cloud segmentation is a further step in the segmentation process for each object than processing or analyzing each point individually. processing and analysis to make it more informative;

Step 5, target identification: identify the steel structure in the model; target identification is usually performed by performing analysis according to the results of feature extraction and segmentation, and based on prior knowledge under given constraints and rules; after steel structure feature identification, For the occluded part, it is necessary to make up points linearly, so as to obtain the result data of the hierarchical steel structure mesh structure. The point-filling algorithm includes the rules for parameter calculation, including point cloud point selection methods, calculation rules, and weighting conditions. Theoretical calculation rules can be quickly formed through mathematical models, and then need to be adjusted in combination with practice until the algorithm results meet the actual expectations and the data has a high degree of confidence.

Then perform point cloud classification, classify the point cloud into different point cloud sets, and divide similar or identical attributes into the same point cloud set;

Step 6: Obtain the judgment result: analyze the steel structure to obtain the construction parameter data (spacing, quantity, shape, diameter, length, volume) of the steel structure, so as to accurately judge the construction quality of the steel structure.

Example 2

The difference between this embodiment and the specific embodiment 1 is that the point cloud model in step 2 is obtained after noise reduction processing. The specific noise reduction processing method is to first read the point cloud, use the straight-through filtering method to remove useless information, and then pass through statistical filtering or radius filtering to remove large-scale noise.

For the statistical filtering method, calculate the distance value of the adjacent area of the data point, count the number within the distance mean value and set the distance threshold value, when the distance mean value is not less than the distance threshold value, remove the changed point, if it is less than the distance threshold value, keep the data point;

For the radius filtering method, set the radius, set the number of adjacent regions of a certain data point, and set whether the number of neighborhoods within the radius is greater than the number of neighborhood points set, if not less than the set number of neighborhoods, then Remove the data point, and keep the data point if it is less than the number of neighbor points;

Then use the moving least squares method to remove the small size noise, and estimate the smoothness and detect the denoising effect to determine whether the normal direction is consistent, if not, return to statistical filtering or radius filtering to perform noise reduction processing again until the normal direction is consistent So far, the denoised point cloud model is obtained.

Example 3

Compared with the second embodiment, this specific embodiment adds the implementation carrier of each step. The method steps in this embodiment can be implemented by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer-readable memory. or implement. The method can use standard programming techniques, and each program can be implemented in a high-level procedural or object-oriented programming language to communicate with a computer system. If desired, the program can be implemented in assembly or machine language. In any case, the language can be a compiled or interpreted language. Furthermore, the program can be run on a programmed application specific integrated circuit for this purpose.

Furthermore, the operations of the processes described in the method may be performed in any suitable order unless otherwise indicated by the method or otherwise clearly contradicted by context. The processes described in this method (or variations and/or combinations thereof) can be performed under the control of one or more computer systems configured with executable instructions, and as code that executes collectively on one or more processors ( For example, executable instructions, one or more computer programs or one or more applications), implemented by hardware, or a combination thereof, a computer program comprising a plurality of instructions executable by one or more processors.

Example 4

This embodiment is different from the above-mentioned specific embodiments 1-3 by adding a hole filling method. When the three-dimensional scanning device scans the object to be measured to generate a point cloud, due to the scanning device itself or the interference of other external factors, the generated point cloud often There is a lack of holes. These holes must be filled algorithmically for subsequent applications. For point cloud holes, the following hole filling methods can be adopted. The first step is to read the point cloud that needs to be repaired and triangulate the read point cloud; the second step is to use the characteristics of the holes to find the holes that need to be filled. ; The third step is to use the preset rule to initially fill each hole; the fourth step is to use the normal phase constraint to adjust the position of the filling point added in the initial filling process, so that the filling part and the surrounding mesh are consistent in shape characteristics ; The fifth step is to smooth the filled part and its surrounding meshes. Using this method, after triangulating the point cloud, the holes can be quickly filled with the preset rules, and the positions of the filling points added in the initial filling process can be adjusted by using the normal phase constraints, so that the filling part and the surrounding mesh are in the same position. The shape features are consistent to achieve intelligent and rapid repair of holes, which not only ensures the naturalness of filling, but also ensures that the shape of the object to be measured will not change.

Example 5

Using another point cloud replenishment method based on Embodiment 4, the pre-acquired point cloud is divided into multiple cubes with overlapping parts; the target block in the multiple cubes and the target source corresponding to the target block are determined The block includes: determining other cubes except the target block among the multiple cubes as candidate blocks; obtaining the similarity between all the candidate blocks and the target block, and obtaining the DC component gaps and directions of all the candidate blocks and the target block. anisotropy map total variation gap; determine the similarity between all candidate blocks and the target block according to the DC component gap and the anisotropy map total variation gap; determine the source block corresponding to the target block according to the similarity; The block and the source block determine the target source block, and determine the target source block corresponding to the target block in the multiple cubes. The target block contains missing data; repair the missing area in the target block according to the information in the target source block, and obtain the repaired Result block; replace the target block in the point cloud with the repair result block to obtain the repaired point cloud.

The above content is a further detailed description of the present application in conjunction with the specific preferred embodiments, and it cannot be considered that the specific embodiments of the present application are limited to these descriptions. For those of ordinary skill in the technical field of the present application, without departing from the concept of the present application, some simple deductions or substitutions can be made, which should be regarded as belonging to the protection scope of the present application.

Claims (10)

1. a method for automatically judging the quality of a steel structure based on a point cloud of a three-dimensional laser scanner, is characterized in that, the method comprises the steps: Step 1, point cloud analysis: use a 3D laser scanner to scan the target to obtain 3D point cloud data of the steel structure; Step 2, preprocessing: import the 3D point cloud data into the software and preprocess the 3D point cloud data to obtain a steel structure point cloud model; Step 3, feature extraction: plane extraction and edge detection and calculation of feature descriptors; Step 4, point cloud segmentation: grouping points into a part or an object based on low-level attributes; Step 5, target identification: identify the steel structure in the model; Step 6, obtain the judgment result: analyze the steel structure to obtain the construction parameter data of the steel structure, so as to make an accurate judgment on the construction quality of the steel structure. 2 . The method according to claim 1 , wherein: after the step 5, the method further comprises point cloud classification, classifying the point clouds into different point cloud sets, and dividing similar or identical attributes into the same point cloud set. 3 . 3 . The method according to claim 1 , wherein in the step 1, the target is a tunnel section. 4 . 4 . The method according to claim 1 , wherein in the second step, the method for preprocessing the three-dimensional point cloud data is: automatic filtering and downsampling according to the quality and scale of the point cloud data. 5 . 5 . The method according to claim 1 , wherein in the third step, the feature extraction is the key to describe the morphological structure of the point cloud, and the basis and premise of semantic information extraction. 6 . 6 . The method according to claim 1 , wherein in the step 4, the point cloud segmentation further processes and analyzes each object in the segmentation process compared to processing or analyzing each point individually. 7 . 7. The method according to claim 1, characterized in that: in the step 5, the target recognition is usually performed by performing analysis according to the results of feature extraction and segmentation, and based on a priori knowledge in given constraints and rules proceed below. 8. The method according to claim 1 or 2, wherein in the step 6, the construction parameter data are spacing, quantity, shape, diameter, length, and volume. 9 . The method according to claim 1 , wherein the steel structure is a steel bar or a steel arch. 10 . 10 . The method according to claim 1 , wherein the fourth step can also be point cloud single-point feature analysis or single-point feature analysis and point cloud segmentation performed simultaneously. 11 .

Images