Intelligent identification and corrosion grade evaluation method for steel structural building steel members and evaluation equipment thereof

Abstract

The invention relates to an intelligent identification and corrosion grade evaluation method and evaluation equipment for a steel structural building steel member, and the evaluation method comprises the steps: collecting an original image of the steel member in a target building construction stage, carrying out the marking and distinguishing of the original image, and constructing an original image data set of the target building steel member; the method comprises the following steps: performing an indoor accelerated corrosion experiment on experimental samples of the same material of a target building steel member, collecting sample corrosion images, training a sample corrosion classification model through image recognition and a deep learning convolutional neural network, and constructing a sample corrosion image database; afterwards, collecting a target building steel member corrosion image in a target building maintenance stage, comparing the steel member corrosion image with the original image data set, identifying the type of the steel member in the steel member corrosion image and the position of the steel member in the target building, and comparing the steel member corrosion image with the sample corrosion image database; and determining the corrosion grade of the steel member in the steel member corrosion image.

Description

Technical Field

[0001] The present invention relates to the technical field of steel structure building maintenance, and in particular to a method for intelligently identifying steel components of a steel structure building and evaluating corrosion levels thereof, and evaluation equipment thereof. Background Art

[0002] Steel structure buildings are buildings that use construction steel to form a load-bearing structure. Usually, the load-bearing structure is composed of steel beams, steel columns, steel trusses and other steel components made of steel and steel plates. Together with building maintenance structures such as roofs, floors and walls, they form a complete building. Steel structure buildings are one of the main types of building structures. Because of their light weight and simple construction, they are widely used in large factories, bridges, stadiums, super high-rise buildings and other fields.

[0003] During the use of steel structure buildings, steel components are easily affected by the use environment and cause corrosion problems. The corrosion of steel components seriously affects the service life and structural safety of the building. Therefore, steel structure buildings need to be maintained regularly during use. At present, the maintenance method for steel structure buildings is: manual regular inspection of steel structure buildings, discovering and evaluating the corrosion of steel components, and timely repairing corroded steel components to ensure the structural safety of steel structure buildings.

[0004] However, this maintenance method is not practical for large-span steel structures such as large stadiums and shopping malls. First, it is difficult for maintenance personnel to get close to observe large-span steel structures, and their daily maintenance and monitoring are difficult to complete through manual close inspections. Secondly, the human judgment of the corrosion level is highly subjective and has a high misjudgment rate, which is difficult to meet the maintenance requirements of steel structures. For this reason, maintenance personnel set up multiple non-destructive monitoring points at key steel components of large-span steel structures to monitor the corrosion of steel components. However, the fixed monitoring points can only monitor fixed points, making it difficult to grasp the overall situation of the steel structure. Adding too many fixed monitoring points will significantly increase the maintenance cost.

[0005] In view of the problems existing in the maintenance stage of large-span steel structure buildings, such as the difficulty of manual approach inspection, high rate of human misjudgment, and difficulty in low-cost comprehensive assessment of corrosion conditions, there is an urgent need for a low-cost corrosion assessment method for steel components in steel structure buildings to achieve comprehensive maintenance of steel components in large-span steel structure buildings. Summary of the invention

[0006] In view of this, in order to overcome the shortcomings of the prior art problems, the present invention provides a method for intelligently identifying and evaluating the corrosion grade of steel components of steel structure buildings and an evaluation device thereof. The evaluation method constructs an original image data set of steel components of a target building during the construction stage of the target building, and constructs a sample corrosion image database through indoor accelerated corrosion experiments. Then, during the maintenance stage of the target building, corrosion images of steel components of the target building are collected, and the corrosion images of the steel components are compared with the original image data set of steel components of the target building to determine the type of steel components in the corrosion images of the steel components and their positions in the target building, and the corrosion images of the steel components are compared with the sample corrosion image database to determine the corrosion grade of the steel components in the corrosion images of the steel components.

[0007] To achieve the above-mentioned purpose, the technical solution of the method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings provided by the present invention is: A method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings, comprising the following process steps: S1. During the construction phase of the target steel structure building, the original images of the steel components of the target building including the background are collected, the types and location information of the steel components in the original images are distinguished through semantic annotation, and the original image dataset of the steel components of the target building is constructed; S2. Conduct indoor accelerated corrosion experiments on experimental samples with the same material as the steel components of the target steel structure building, regularly collect sample corrosion images of the indoor accelerated corrosion experiments, train the sample corrosion grade classification model through image recognition and corrosion grade classification, and build a sample corrosion image database; S3, during the maintenance phase of the target steel structure building, a movable image acquisition device is used to acquire corrosion images of the steel components of the target building including the background; S4, comparing the steel component corrosion image of the target building with the background acquired in step S3 with the original image in the original image data set of the steel components of the target building constructed in step S1 one by one, using the semantic segmentation model to identify the type information of the steel components in the steel component corrosion image acquired in step S3 and their position information in the target building, and extracting the steel component surface image without the background from the steel component corrosion image by sliding window sampling; S5. Perform feature comparison between the steel component surface image obtained in step S4 and the corrosion image of the sample in the sample corrosion image database constructed in step S2 to determine the corrosion grade of the steel component in the steel component surface image.

[0008] Furthermore, in the step S1, an image acquisition device is used to acquire original images including backgrounds of steel components of the target building in at least four directions. The acquired original images are labeled according to the types and position information of the steel components in the original images using the open source image semantic annotation software Labelme. The acquired original images are distinguished according to the types of the steel components in the original images. Original images of the same type are grouped into one data set. Original images in the same data set are distinguished by the annotated position information to form an original image data set of the steel components of the target building.

[0009] Furthermore, when annotating the collected original image, the collected original image is annotated at three levels, wherein the first level annotation is the name of the target building, the second level annotation is the position information of the steel components in the original image in the target building, and the third level annotation is the type of the steel components in the original image.

[0010] Furthermore, in step S2, an image acquisition device is used to regularly acquire sample corrosion images of the indoor accelerated corrosion experiment, and the corrosion area size and corrosion color depth of the sample in the sample corrosion image are extracted through an image recognition algorithm. According to the corrosion area size and corrosion color depth, a deep learning convolutional neural network is used to train a sample corrosion grade classification model, and the collected sample corrosion images are classified according to the corrosion grade to establish a sample corrosion image database.

[0011] Furthermore, the deep learning convolutional neural network is one of AlexNet network, ResNet network, VGGNet network and Inception network.

[0012] Furthermore, the image acquisition device is one of a digital camera, a mobile phone with a camera function, an electronic camera, an unmanned aerial vehicle with a camera function, and a mobile robot with a camera function that can crawl on the surface of a steel structure by adsorption.

[0013] Furthermore, in the step S3, when collecting the corrosion image of the steel structure, the movable image acquisition device moves continuously and smoothly around the steel structure of the target building. During the movement, the distance between the movable image acquisition device and the steel structure of the target building is fixed. During the movement, the movable image acquisition device continuously shoots the steel structure to obtain the corrosion image of the steel structure of the target building including the background.

[0014] Furthermore, the movable image acquisition device is an unmanned aerial vehicle with a photographing function, or a mobile robot with a photographing function that can adsorb and crawl on the surface of a steel component.

[0015] Furthermore, in step S4, the semantic segmentation model is one of an RCNN model, a Yolo model, a SegNet model, and a U-Net model.

[0016] Further, the step S5 comprises: S51, extracting the corrosion area characteristics and corrosion color characteristics of the steel component in the steel component surface image obtained in step S4 using an image recognition algorithm; S52. Using an image recognition algorithm, extract a sample corrosion image that is closest to the corrosion area characteristics and corrosion color characteristics obtained in step S51 from a sample corrosion image database. The corrosion level of the sample corrosion image is the corrosion level of the steel component in the steel component surface image.

[0017] The evaluation equipment used in the above-mentioned intelligent identification and corrosion grade evaluation method of steel structure building steel components includes: A memory for storing a computer program, an original image data set of a target building steel component, a sample corrosion image database, and a corrosion image of a target building steel component, Communication bus, used to realize the communication connection between memory, processor and communication interface, Communication interface, used to communicate with display devices, image acquisition devices, and print output devices, The processor executes the computer program in the memory to implement the method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings as claimed in any one of claims 1 to 9.

[0018] The beneficial effects of the present invention are: 1. During the construction phase of the target steel structure building, the present invention establishes an original image data set of the steel components of the target building. The original image data set can facilitate the management of the steel components of the entire steel structure building. During the maintenance phase of the steel structure building, the type and location information of the corroded steel components can be accurately and conveniently identified, which facilitates the maintenance and repair of the steel components.

[0019] 2. When constructing a sample corrosion image database, the present invention uses experimental samples with the same material as the steel components of the target steel structure building as experimental objects, and conducts indoor accelerated corrosion experiments on the samples. The accelerated corrosion experiment can simulate the corrosion conditions of the samples in a single corrosion environment or multiple corrosion environments such as salt spray corrosion, gas corrosion, and ultraviolet corrosion, and reflect the corrosion environment of the steel components as realistically as possible, thereby improving the accuracy of the sample corrosion images in the sample corrosion image database, improving its reference value, and ensuring the accuracy of the evaluation results of the degree of corrosion of the steel components.

[0020] 3. The present invention identifies corrosion features such as corrosion color and corrosion area in the corrosion image of steel components through image recognition algorithms, and determines the corrosion level according to the identified corrosion features. Compared with the manual method of judging the corrosion level, it can effectively improve the efficiency of corrosion feature recognition and the recognition accuracy, thereby improving the operation and maintenance efficiency of steel structure buildings, reducing the operation and maintenance costs, and realizing digital and information-based intelligent monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 The process flow diagram of the evaluation method of the present invention is shown in FIG. Figure 1 ; Figure 2 The process flow diagram of the evaluation method of the present invention is shown in FIG. Figure 2 ; Figure 3 It is a schematic diagram of the structure of the evaluation device of the present invention. DETAILED DESCRIPTION

[0022] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments: like Figure 1 and Figure 2 As shown, a method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings, which includes the following process steps: S1. During the construction phase of the target steel structure building, the original images of the steel components of the target building including the background are collected, and the types and location information of the steel components in the original images are distinguished through semantic annotation to construct the original image dataset of the steel components of the target building.

[0023] An image acquisition device is used to collect original images including the background of the target building steel components in four directions. The image acquisition device is a mobile phone with a camera function, or a digital camera, or a drone with a camera function. The pixel size of the collected original image is greater than or equal to 1000×1000 pixels, and the area of ​​the steel component in the original image accounts for more than 50% of the area of ​​the original image. The median filtering method or the mean filtering method is used to perform noise reduction on the collected original image.

[0024] The collected original images are annotated according to the type and location information of the steel components in the original images through the open source image semantic annotation software Labelme. During annotation, the collected original images are annotated at three levels, where the first level annotation is the name of the target building, the second level annotation is the location information of the steel components in the original image in the target building, and the third level annotation is the type of steel components in the original image.

[0025] The collected original images are distinguished according to the types of steel components in the original images, and the original images of the same type are classified into one data set. The original images in the same data set are distinguished by the annotated position information to form an original image data set of the steel components of the target building.

[0026] S2. Conduct indoor accelerated corrosion experiments on experimental samples with the same material as the steel components of the target steel structure building, regularly collect sample corrosion images of the indoor accelerated corrosion experiments, train the sample corrosion grade classification model through image recognition and corrosion grade classification, and build a sample corrosion image database.

[0027] Indoor accelerated corrosion tests are conducted in a corrosion test chamber, which can simulate a single corrosion environment such as salt spray corrosion environment, gas corrosion environment, ultraviolet corrosion environment, or a corrosion environment coupled with multiple corrosion conditions. Indoor accelerated corrosion tests must continue to accelerate the corrosion of samples until the sample thickness does not meet the bearing performance requirements of the steel structure project.

[0028] The sample corrosion images of the indoor accelerated corrosion experiment are collected regularly by using an image acquisition device, the image acquisition device is a digital camera or a mobile phone with a camera function or an electronic camera, and the pixel size of the collected sample corrosion images is greater than or equal to 1000×1000 pixels. The collected sample corrosion images are images containing backgrounds, and the sample corrosion images containing backgrounds are intercepted, and the image pixels of the intercepted part are also required to be greater than or equal to 1000×1000 pixels. The collected sample corrosion images are subjected to noise reduction processing by using a median filtering method or a mean filtering method.

[0029] The size of the corrosion area and the depth of the corrosion color of the sample in the image intercepted from the sample corrosion image are extracted by an image recognition algorithm. According to the size of the corrosion area and the depth of the corrosion color, a deep learning convolutional neural network is used to train a sample corrosion grade classification model, and the collected sample corrosion images are classified according to the corrosion grade to establish a sample corrosion image database. The deep learning convolutional neural network is one of the AlexNet network, the ResNet network, the VGGNet network and the Inception network.

[0030] According to the corrosion area and corrosion color depth of the sample in the sample corrosion image, the severity of the sample corrosion is divided into four levels from light to heavy: A, B, C and D. Of course, the classification of corrosion levels can be further subdivided according to the corrosion conditions of steel structure buildings. The corrosion level information contained in the sample corrosion image database should also include sample material information, corrosion condition information, corrosion area, corrosion color depth, mass loss rate, tensile strength, deflection, shear strength and other information.

[0031] S3. During the maintenance phase of the target steel structure building, a movable image acquisition device is used to collect corrosion images of the steel components of the target building including the background.

[0032] The maintenance phase of the target steel structure building refers to the period from the completion of the construction of the target steel structure building to the cessation of use of the steel structure building.

[0033] When collecting steel component corrosion images, the movable image acquisition device moves continuously and smoothly around the target building steel components. During the movement, the distance between the movable image acquisition device and the steel components of the target building is fixed. During the movement, the movable image acquisition device continuously shoots the steel components to obtain the steel component corrosion images of the target building with background. The movable image acquisition device is an unmanned aerial vehicle with a camera function, or a mobile robot with a camera function that can crawl on the surface of the steel component. The pixel size of the collected steel component corrosion image of the target building with background is greater than or equal to 1000×1000 pixels, and the steel component in the steel component corrosion image occupies greater than or equal to 50% of the image area, and the median filtering method or the mean filtering method is used to perform noise reduction processing on the collected steel component corrosion image. .

[0034] S4. Compare the steel component corrosion image of the target building with background collected in step S3 with the original image in the original image data set of the steel components of the target building constructed in step S1 one by one, use the semantic segmentation model to identify the type information of the steel components in the steel component corrosion image obtained in step S3 and their position information in the target building, and extract the steel component surface image without background from the steel component corrosion image by sliding window sampling; the semantic segmentation model is one of RCNN model, Yolo model, SegNet model and U-Net model.

[0035] S5, comparing the steel component surface image obtained in step S4 with the corrosion image of the sample in the sample corrosion image database constructed in step S2, and determining the corrosion level of the steel component in the steel component surface image. Specifically including: S51. Extracting the corrosion area characteristics and corrosion color characteristics of the steel component in the steel component surface image obtained in step S4 using an image recognition algorithm.

[0036] S52. Using an image recognition algorithm, extract a sample corrosion image that is closest to the corrosion area characteristics and corrosion color characteristics obtained in step S51 from a sample corrosion image database. The corrosion level of the sample corrosion image is the corrosion level of the steel component in the steel component surface image.

[0037] After determining the corrosion level of the steel component in the surface image of the steel component, the position information, steel component type information, corrosion level, corrosion condition information, corrosion area, corrosion color depth, mass loss rate, tensile strength, deflection, shear strength and other information corresponding to the steel component in the corrosion image of the steel component are formed into an evaluation table and output. At this point, the intelligent identification and corrosion level evaluation of the steel components of the target steel structure building are completed.

[0038] like Figure 3 As shown, the evaluation equipment used in the above-mentioned steel structure building steel member intelligent identification and corrosion grade evaluation method includes: a memory 101, a communication bus 102, a communication interface 103 and a processor 104. The communication bus 102 is electrically connected to the processor 104, the memory 101 and the communication interface 103 respectively to realize communication. The communication interface 103 can be connected to a display device, an image acquisition device and a print output device through a communication cable. The memory 101 is used to store computer programs, original image data sets of target building steel members, sample corrosion image databases and corrosion images of steel members of the target building. The processor 104 is used to call and execute the computer program on the memory, and realize the intelligent identification of steel members of steel structure buildings and corrosion grade evaluation through the computer program.

[0039] It should be noted that the above embodiments are illustrative of the technical solution of the present invention rather than limiting. Equivalent substitutions made by ordinary technicians in the relevant technical field or other modifications made according to the prior art should be included in the scope of rights required by the present invention as long as they do not exceed the concept and scope of the technical solution of the present invention.

Claims

1. A method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings, characterized by: The process steps include: S1. During the construction phase of the target steel structure building, the original images of the steel components of the target building including the background are collected, the types and location information of the steel components in the original images are distinguished through semantic annotation, and the original image dataset of the steel components of the target building is constructed; S2. Conduct indoor accelerated corrosion experiments on experimental samples with the same material as the steel components of the target steel structure building, regularly collect sample corrosion images of the indoor accelerated corrosion experiments, train the sample corrosion grade classification model through image recognition and corrosion grade classification, and build a sample corrosion image database; S3, during the maintenance phase of the target steel structure building, a movable image acquisition device is used to acquire corrosion images of the steel components of the target building including the background; S4, comparing the steel component corrosion image of the target building including the background acquired in step S3 with the original image in the original image data set of the steel components of the target building constructed in step S1 one by one, using the semantic segmentation model to identify the type information of the steel components in the steel component corrosion image acquired in step S3 and their position information in the target building, and extracting the steel component surface image from the steel component corrosion image by sliding window sampling; S5. Perform feature comparison between the steel component surface image obtained in step S4 and the sample corrosion image in the sample corrosion image database constructed in step S2 to determine the corrosion grade of the steel component in the steel component surface image.

2. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 1 is characterized by: In the step S1, an image acquisition device is used to acquire original images including backgrounds of steel components of the target building in at least four directions. The acquired original images are labeled according to the types and position information of the steel components in the original images using the open source image semantic annotation software Labelme. The acquired original images are distinguished according to the types of the steel components in the original images. Original images of the same type are grouped into one data set. Original images in the same data set are distinguished by the annotated position information to form an original image data set of the steel components of the target building.

3. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 2 is characterized by: When annotating the collected original image, the collected original image is annotated at three levels, wherein the first level annotation is the name of the target building, the second level annotation is the position information of the steel components in the original image in the target building, and the third level annotation is the type of the steel components in the original image.

4. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 1 is characterized by: In the step S2, the sample corrosion images of the indoor accelerated corrosion experiment are regularly collected by using an image acquisition device, and the corrosion area size and corrosion color depth of the sample in the sample corrosion image are extracted by an image recognition algorithm. According to the corrosion area size and corrosion color depth, a deep learning convolutional neural network is used to train a sample corrosion grade classification model, and the collected sample corrosion images are classified according to the corrosion grade to establish a sample corrosion image database.

5. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 4 is characterized by: The deep learning convolutional neural network is one of AlexNet network, ResNet network, VGGNet network and Inception network.

6. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 1 is characterized by: In the step S3, when collecting the corrosion image of the steel structure, the movable image acquisition device moves continuously and smoothly around the steel structure of the target building. During the movement, the distance between the movable image acquisition device and the steel structure of the target building is fixed. During the movement, the movable image acquisition device continuously shoots the steel structure to obtain the corrosion image of the steel structure of the target building including the background.

7. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 6 is characterized by: The movable image acquisition device is an unmanned aerial vehicle with a photographing function, or a mobile robot with a photographing function that can adsorb and crawl on the surface of a steel component.

8. The method for intelligent identification and corrosion grade evaluation of steel structure building steel components according to claim 1 is characterized by: In the step S4, the semantic segmentation model is one of the RCNN model, Yolo model, SegNet model, and U-Net model.

9. The method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to claim 1 is characterized by: The step S5 comprises: S51, extracting the corrosion area characteristics and corrosion color characteristics of the steel component in the steel component surface image obtained in step S4 using an image recognition algorithm; S52. Using an image recognition algorithm, extract a sample corrosion image that is closest to the corrosion area characteristics and corrosion color characteristics obtained in step S51 from a sample corrosion image database. The corrosion level of the sample corrosion image is the corrosion level of the steel component in the steel component surface image.

10. The evaluation device used in the method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings according to any one of claims 1 to 9 is characterized in that: include: A memory for storing a computer program, an original image data set of a target building steel component, a sample corrosion image database, and a corrosion image of a target building steel component, Communication bus, used to realize the communication connection between memory, processor and communication interface, Communication interface, used to communicate with display devices, image acquisition devices, and print output devices, The processor executes the computer program in the memory to implement the method for intelligent identification and corrosion grade evaluation of steel components of steel structure buildings as claimed in any one of claims 1 to 9.

Images