CN117095294A - Precast floor slab construction quality diagnosis method, medium and system - Google Patents

Precast floor slab construction quality diagnosis method, medium and system Download PDF

Info

Publication number
CN117095294A
CN117095294A CN202311078117.7A CN202311078117A CN117095294A CN 117095294 A CN117095294 A CN 117095294A CN 202311078117 A CN202311078117 A CN 202311078117A CN 117095294 A CN117095294 A CN 117095294A
Authority
CN
China
Prior art keywords
image
floor slab
precast floor
prefabricated
aerial vehicle
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202311078117.7A
Other languages
Chinese (zh)
Other versions
CN117095294B (en
Inventor
王书洞
苏占宏
曹勇
张琳
孔文龙
李亚伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Construction Installation Group Huanghe Construction Co ltd
China Construction Industrial and Energy Engineering Group Co Ltd
Original Assignee
China Construction Installation Group Huanghe Construction Co ltd
China Construction Industrial and Energy Engineering Group Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Construction Installation Group Huanghe Construction Co ltd, China Construction Industrial and Energy Engineering Group Co Ltd filed Critical China Construction Installation Group Huanghe Construction Co ltd
Priority to CN202311078117.7A priority Critical patent/CN117095294B/en
Publication of CN117095294A publication Critical patent/CN117095294A/en
Application granted granted Critical
Publication of CN117095294B publication Critical patent/CN117095294B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/176—Urban or other man-made structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/51—Indexing; Data structures therefor; Storage structures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
    • G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
    • G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G06Q10/06395—Quality analysis or management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/08—Construction
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/17—Terrestrial scenes taken from planes or by drones
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
    • Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
    • Y02P90/30—Computing systems specially adapted for manufacturing

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Educational Administration (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Development Economics (AREA)
  • Library & Information Science (AREA)
  • Tourism & Hospitality (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Game Theory and Decision Science (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Remote Sensing (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Primary Health Care (AREA)
  • Image Processing (AREA)

Abstract

The invention provides a method, medium and system for diagnosing the construction quality of precast floor slabs, which belong to the technical field of building construction quality diagnosis, and comprise the steps of acquiring a first image before construction and a second image set after construction of each precast floor slab, and establishing a precast floor slab image database by utilizing the first images of all precast floor slabs; preprocessing each image in the first image and the second image set; for each prefabricated floor slab, carrying out left-right combination and merging on each preprocessed image in the corresponding preprocessed first image and second image set to form a merged image set containing a plurality of merged images; and for each precast floor slab, calculating each combined image in the combined image set by using a pre-trained precast floor slab construction quality diagnosis model to obtain flaw index of the corresponding precast floor slab as construction quality diagnosis data, and sending the flaw index to constructors.

Description

Precast floor slab construction quality diagnosis method, medium and system
Technical Field
The invention belongs to the technical field of building construction quality diagnosis, and particularly relates to a method, medium and system for diagnosing construction quality of a precast floor slab.
Background
The prefabricated floor slab is a building component commonly used in building engineering and has the characteristics of high production efficiency, stable quality, high construction speed and the like. In modern construction engineering, prefabricated floors are widely used. However, in the course of construction of the precast floor slabs, flaws such as cracks, pitting, bubbles, surface tilting, etc. may occur for various reasons, and these flaws may affect the workability of the precast floor slabs and the overall quality of the construction work. The construction steps of the prefabricated floor slab are generally as follows: leveling layer or hard frame formwork, drawing board position line, hoisting prefabricated floor slab, regulating floor slab position and welding anchoring rib; in the work progress, need detect the position of precast floor slab and just can adjust, simultaneously, in hoist precast floor slab in-process, can make precast floor slab take place crackle or flaw because of collision etc. at this moment, need change the precast floor slab that produces crackle or flaw. Therefore, it is very important to effectively diagnose and monitor the construction quality of the precast floor slabs.
At present, the diagnosis method for the construction quality of the precast floor slabs mainly comprises the following steps: manual inspection, non-destructive inspection techniques (e.g., ultrasonic inspection, magnetic particle inspection, etc.), and image processing techniques. The manual inspection method is the most traditional diagnosis method, but has the defects of strong subjectivity, low efficiency, great influence on detection accuracy by the skill level of operators, incomplete construction, lack of protective measures, danger caused by manual inspection and the like. The nondestructive testing technology can detect internal flaws of the precast floor slabs, but has higher equipment cost, complex operation and incapability of monitoring the construction quality of the precast floor slabs in real time.
In recent years, with the development of computer vision, image processing and machine learning technologies, a precast floor slab construction quality diagnosis method based on the image processing technology is attracting attention. The method is used for identifying the flaws on the precast floor slabs by processing and analyzing the images acquired in the precast floor slab construction process, so that the diagnosis of the precast floor slab construction quality is realized. In practical use, the method can only identify the flaws on the precast floor slabs, can not quantify the flaws of the piled precast floor slabs, and is difficult to diagnose the construction quality of the precast floor slabs in a quantified mode.
Disclosure of Invention
In view of the above, the invention provides a method, medium and system for diagnosing the construction quality of a precast floor slab, which can realize the diagnosis of the construction quality of the precast floor slab in a quantitative manner.
The invention is realized in the following way:
the first aspect of the invention provides a method for diagnosing construction quality of a precast floor slab, which comprises the following steps:
s10, acquiring a first image before construction and a second image set after construction of each precast floor slab, and establishing a precast floor slab image database by using the first images of all precast floor slabs; each image in the first image and the second image set is a surface image of the precast floor slab, the acquisition time of the first image is the time when the precast floor slab is to be constructed when being taken out of a warehouse, and the second image set is an image group formed by acquisition according to a specified time interval after the precast floor slab is installed at a set position;
s20, preprocessing each image in the first image set and the second image set;
s30, for each prefabricated floor slab, carrying out left-right combination and merging on each preprocessed image in the corresponding preprocessed first image and the corresponding preprocessed second image set to form a merged image set containing a plurality of merged images, wherein the left part after merging is the corresponding preprocessed first image, and the right part is the preprocessed image in the second image set;
s40, for each prefabricated floor slab, calculating each combined image in the combined image set by using a pre-trained prefabricated floor slab construction quality diagnosis model to obtain flaw index of the corresponding prefabricated floor slab as construction quality diagnosis data, and sending the flaw index to constructors;
wherein the flaws include cracks, pitting, bubbles and surface tilting, the flaw index comprises 0-5 levels, a numerical value indicates the severity of the flaw, a larger numerical value indicates a greater severity of the flaw, and 0 indicates no flaws of this type; the flaw index is a vector representing the level of each flaw.
The first image acquisition mode is that constructors shoot the surface of the precast floor slab by adopting a portable camera when the precast floor slab is ready to be constructed; the operation of preprocessing each image in the first and second image sets includes graying, noise removal, and image enhancement of the image.
The flaw index of the corresponding precast floor slab is obtained and used as construction quality diagnosis data, and when the flaw index is sent to constructors, the flaw index can be output in a spider-web diagram mode, so that the constructors can conveniently solve the construction quality of the threshold floor slab.
On the basis of the technical scheme, the method for diagnosing the construction quality of the precast floor slabs can be improved as follows:
the method for collecting each image in the second image set is to collect the surface image of each precast floor slab by using an unmanned aerial vehicle carrying a camera, and specifically comprises the following steps:
acquiring a three-dimensional structure of a large-scale building for construction;
setting a movement path of the unmanned aerial vehicle according to the three-dimensional structure of the constructed large building;
controlling the unmanned aerial vehicle to enter the interior of a large-scale building for construction according to a motion path to shoot;
and acquiring the surface image of each prefabricated floor slab shot by the unmanned aerial vehicle.
The beneficial effects of adopting above-mentioned improvement scheme are: when the unmanned aerial vehicle with the camera is used for collecting the second image set, the number of prefabricated floors used in a large building is large, and the real-time connection between a control end and the unmanned aerial vehicle cannot be guaranteed because the wireless communication signal strength is insufficient in the large building, so that the unmanned aerial vehicle cannot be controlled in a real-time control mode, and therefore, the movement path of the unmanned aerial vehicle needs to be set according to the three-dimensional structure of the large building. The three-dimensional structure of the large building constructed in this step also includes the area of the large building where the precast floor slabs are used or the specific location of the precast floor slabs.
Further, the step of controlling the unmanned aerial vehicle to enter the interior of the constructed large building for shooting according to the motion path specifically comprises the following steps of;
controlling the unmanned aerial vehicle to enter a large-scale building for construction and fly according to a motion path;
the unmanned aerial vehicle is controlled to identify the prefabricated floors in the way, wherein the identification method is to train a neural network by using a first image of each prefabricated floor, and identify the prefabricated floors by using the trained neural network;
the unmanned aerial vehicle is controlled to carry out image acquisition to the prefabricated floor of discernment, if the prefabricated floor image of shooting has to shelter from, then the unmanned aerial vehicle change position of control carries out many picture acquisition to the prefabricated floor that has to shelter from to it merges to have a picture to shelter from many pictures of prefabricated floor.
The beneficial effects of adopting above-mentioned improvement scheme are: because the wireless communication inside the large-scale building is not smooth in the construction, the unmanned aerial vehicle is required to automatically identify the prefabricated floor slab and shoot the surface of the prefabricated floor slab. In the steps, the trained neural network is utilized to identify the precast floor slabs, whether the unmanned aerial vehicle approaches the precast floor slabs can be judged, and if the unmanned aerial vehicle approaches the precast floor slabs, the surface images of the identified precast floor slabs are acquired.
Further, the step of controlling unmanned aerial vehicle to carry out image acquisition to the prefabricated floor of discernment, if the prefabricated floor image of shooing has sheltered from, then controlling unmanned aerial vehicle change position carries out many pictures to the prefabricated floor that has sheltered from to have, and will have a plurality of pictures of the prefabricated floor that shelters from to merge into a picture specifically includes:
basic collection: the unmanned aerial vehicle is controlled to collect images of the identified prefabricated floor slab, specifically, during collection, firstly, the vertical determination device arranged on a camera carried by the unmanned aerial vehicle determines that the camera is vertical, and then, the unmanned aerial vehicle is controlled to shoot the prefabricated floor slab;
and (5) merging and collecting: if the photographed image of the precast floor slab is shielded, the change position of the unmanned aerial vehicle is controlled to collect multiple pictures of the shielded precast floor slab, and the multiple pictures of the shielded precast floor slab are combined into one picture, specifically:
step 1, acquiring a prefabricated floor image acquired basically as a first basic image;
step 2, searching a first image of the prefabricated floor slab corresponding to the basic image in the prefabricated floor slab image database according to the surface characteristics of the first basic image, and recording the first image as a reference image;
step 3, calculating the surface area difference of the first basic image and the reference image;
step 4, if the surface area difference is not more than 10%, taking the first basic image as an acquired image of the corresponding precast floor slab; if the surface area difference is greater than 10%, adjusting the position of the unmanned aerial vehicle and acquiring the surface of the corresponding prefabricated floor again to obtain a second basic image;
and 5, taking the image obtained by combining the second basic image and the first basic image as the first basic image, repeating the steps 3-5, and stopping repeating the steps if the step 3 is repeatedly executed for more than 6 times.
Further, the specific step of calculating the surface area difference between the first basic image and the reference image includes:
preprocessing the first basic image and the reference image;
calculating the surface areas of the prefabricated floors of the first basic image and the reference image after pretreatment;
and calculating the difference between the surface areas of the prefabricated floors of the first basic image and the reference image after pretreatment.
The method specifically comprises the steps of building and training the precast floor slab construction quality diagnosis model, and specifically comprises the following steps:
building a training sample;
establishing and training a model prototype;
the step of establishing a training sample specifically includes:
acquiring a first historical merged image set comprising a plurality of merged images;
setting flaw indexes for each combined image by adopting a MiniGPT-4 model;
deleting the combined image without any flaw in the training sample;
the step of building the training model embryonic form specifically comprises the following steps:
establishing a precast floor slab construction quality diagnosis model prototype by utilizing a convolutional neural network;
and training the prototype of the precast floor slab construction quality diagnosis model by taking each combined image in the training sample as training input and the flaw index corresponding to the combined image as training output to obtain the precast floor slab construction quality diagnosis model.
The beneficial effects of adopting above-mentioned improvement scheme are: considering the problem of unsmooth wireless communication in a large building in construction, an unmanned aerial vehicle flies in the construction building, a precast floor slab is shot, and the diagnosis of the construction quality of the precast floor slab is realized, and the current large model needs a high-memory display card, for example, a display card with more than 23G of display memory is needed by a MiniGPT-4 model, so that the unmanned aerial vehicle cannot carry the display card; therefore, the MiniGPT-4 model is adopted to set the flaw index on the combined image, the combined image with the flaw index set is adopted as a training set, a neural network is trained, the operation amount required by the neural network is small, and the neural network can be directly operated in a chip carried on the unmanned aerial vehicle.
Furthermore, the MiniGPT-4 model is a targeted fine-tuned model,
the step of targeted fine tuning of the MiniGPT-4 model specifically comprises the following steps:
acquiring a second historical merged image set comprising a plurality of merged images;
extracting features from the left part of each combined image to form a left feature matrix;
extracting features from the right part of each combined image to form a right feature matrix;
calculating a right characteristic matrix and subtracting the left characteristic matrix to obtain a difference matrix;
adopting ViT of a MiniGPT-4 model to encode the difference matrix to obtain ViT codes;
the method comprises the steps of obtaining a flaw manual description of the right side part of each combined image, wherein the flaw manual description is performed on a corresponding prefabricated floor slab by constructors after actual observation and measurement are performed on the corresponding prefabricated floor slab according to the combined image;
summarizing the manual description of flaws on the right side part of each combined image by using a Vicuna large language model of a MiniGPT-4 model, extracting flaw indexes, and recording the flaw indexes as description flaw indexes;
combining said ViT code and said description blemish index into one construction data;
fine-tuning the MiniGPT-4 model by adopting the construction data to obtain a Lora fine-tuning weight;
combining the obtained Lora fine tuning weight with the weight of the Vicuna large language model in the MiniGPT-4 model to obtain a comprehensive weight, and using the obtained comprehensive weight as the weight of the MiniGPT-4 model to finish fine tuning of the MiniGPT-4 model.
The beneficial effects of adopting above-mentioned improvement scheme are: the flaw index obtained by directly utilizing the MiniGPT-4 model is not accurate enough, so that the flaw index of the combined image is required to be extracted by adopting the MiniGPT-4 model subjected to targeted fine adjustment, and the flaw index is set in order to solve the problems that in actual use, construction quality of a prefabricated floor is detected by constructors, the detection results are all dictated and quantification is difficult; the flaw index of the combined image can be directly extracted by adopting the finely-adjusted MiniGPT-4 model, so that the construction quality of the prefabricated floor slab is quantized.
In the process of targeted fine tuning, the left part and the right part of each combined image are required to be subjected to feature extraction respectively to obtain a left feature matrix and a right feature matrix, and the aim of the processing is that before and after the processing is used, some image features similar to flaws, which do not affect the construction quality, exist on the surface of a part of prefabricated floor slab, so that the right feature matrix is calculated to subtract the left feature matrix to obtain a difference matrix, and the difference matrix is encoded by ViT to obtain ViT codes, so that the image features similar to flaws can be prevented from interfering a MiniGPT-4 model, and the flaw index outputted by the MiniGPT-4 model is prevented from being influenced.
Further, the step of extracting features from the left portion of each combined image to form a left feature matrix specifically includes:
normalizing the left part of the combined image;
edge detection is carried out on the normalized left part image;
extracting LBP characteristics of the left partial image;
and fusing the image obtained by edge detection with the LBP feature, and flattening the fused feature image into a vector serving as a left feature matrix.
A second aspect of the present invention provides a computer readable storage medium having stored therein program instructions which when executed are adapted to carry out a precast floor deck construction quality diagnostic method as described above.
A third aspect of the present invention provides a precast floor plank construction quality diagnostic system, comprising an unmanned aerial vehicle carrying a camera, wherein a control chip in the unmanned aerial vehicle comprises the computer readable storage medium.
Compared with the prior art, the method, medium and system for diagnosing the construction quality of the precast floor plank have the beneficial effects that: the method establishes a precast floor slab image database by acquiring a first image before construction and a second image set after construction of each precast floor slab. The images are preprocessed and combined to form a combined image set, the combined image set is calculated by using the pre-trained precast floor slab construction quality diagnosis model, and the flaw index corresponding to the precast floor slab is obtained as construction quality diagnosis data, so that diagnosis of precast floor slab construction quality is realized in a quantitative mode. The method can effectively monitor the quality problem in the construction process of the precast floor slabs in real time, provide timely quantized quality feedback for constructors, and improve construction quality and efficiency.
Firstly, a prefabricated floor image database is built by acquiring a first image before the construction of the prefabricated floor and a second image set after the construction of the prefabricated floor, and basic data is provided for subsequent image processing and quality diagnosis. The method for acquiring the image can enable the diagnosis method to reflect the actual construction condition of the precast floor slab more intuitively and accurately, and provides powerful data support for quality diagnosis.
Secondly, the invention improves the quality of the images by preprocessing the first image and the second image set, and provides a better data basis for subsequent image processing and quality diagnosis. The preprocessing comprises operations such as denoising and enhancing of the image, so that the definition and recognition rate of the image can be effectively improved, and a better data base is provided for subsequent image recognition and quality diagnosis.
And thirdly, the invention combines the preprocessed first image with each preprocessed image in the second image set to form a combined image set containing a plurality of combined images. The combining mode can integrate the image information before and after the construction of the precast floor slab, and is convenient for subsequent image processing and quality diagnosis.
In addition, the combined image set is calculated by using the pre-trained precast floor slab construction quality diagnosis model, and the flaw index of the corresponding precast floor slab is obtained and used as construction quality diagnosis data. The method can effectively identify various flaws such as cracks, pitting surfaces, bubbles, surface tilting and the like in the construction process of the precast floor slabs, and quantify the severity of the flaws as flaw indexes. The flaw index can intuitively reflect the severity of flaws, and is convenient for constructors to carry out corresponding quality adjustment and improvement according to the flaw index.
Finally, the obtained flaw index is used as construction quality diagnosis data to be sent to constructors, so that the constructors can monitor the construction process of the precast floor slabs in real time and adjust the quality according to the diagnosis result. The method can greatly improve the construction quality and efficiency, reduce the construction cost, and improve the service life and the safety of the precast floor slab.
In a word, the invention provides a method for diagnosing the construction quality of the precast floor slab, and the real-time monitoring and quality adjustment of the construction process of the precast floor slab are realized through image processing and quality diagnosis technologies. The method has good practicability and wide application prospect, and has important significance for improving the construction quality and efficiency of the precast floor slabs.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the description of the embodiments of the present invention will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for diagnosing the construction quality of a precast floor slab.
Detailed Description
As shown in fig. 1, a flow chart of a method for diagnosing construction quality of a precast floor slab according to a first aspect of the present invention is provided, the method comprising the steps of:
s10, acquiring a first image before construction and a second image set after construction of each precast floor slab, and establishing a precast floor slab image database by using the first images of all precast floor slabs; each image in the first image and the second image set is a surface image of the precast floor slab, the acquisition time of the first image is the moment when the precast floor slab is to be constructed when being taken out of the warehouse, and the second image set is an image group formed by acquisition according to a specified time interval after the precast floor slab is arranged at a set position;
s20, preprocessing each image in the first image and the second image set;
s30, for each prefabricated floor slab, carrying out left-right combination and merging on each preprocessed image in the corresponding preprocessed first image and the corresponding preprocessed second image set to form a merged image set containing a plurality of merged images, wherein the left part after merging is the corresponding preprocessed first image, and the right part is the preprocessed image in the second image set;
s40, for each prefabricated floor slab, calculating each combined image in the combined image set by using a pre-trained prefabricated floor slab construction quality diagnosis model to obtain flaw index of the corresponding prefabricated floor slab as construction quality diagnosis data, and sending the flaw index to constructors;
wherein the flaws include cracks, pitting, bubbles and surface tilting, the flaw index comprises 0-5 levels, a numerical value indicates the severity of the flaw, a larger numerical value indicates the severity of the flaw, and a 0 indicates the absence of this type of flaw; the flaw index is a vector representing the level of each flaw.
In the above technical solution, the collecting mode of each image in the second image set is that an unmanned aerial vehicle carrying a camera is utilized to collect the surface image of each prefabricated floor slab, and the specific steps include:
acquiring a three-dimensional structure of a large-scale building for construction;
setting a movement path of the unmanned aerial vehicle according to the three-dimensional structure of the constructed large building;
controlling the unmanned aerial vehicle to enter the interior of a large-scale building for construction according to the motion path to shoot;
and acquiring the surface image of each prefabricated floor slab shot by the unmanned aerial vehicle.
Further, in the above technical solution, the step of controlling the unmanned aerial vehicle to enter the interior of the constructed large building to shoot according to the motion path specifically includes;
controlling the unmanned aerial vehicle to enter a large-scale building for construction and fly according to a motion path;
the unmanned aerial vehicle is controlled to identify the prefabricated floors in the way, wherein the identification method is to train a neural network by using a first image of each prefabricated floor, and identify the prefabricated floors by using the trained neural network;
and controlling the unmanned aerial vehicle to acquire images of the identified precast floor slabs, if the photographed precast floor slab images are shielded, controlling the unmanned aerial vehicle to change positions to acquire multiple pictures of the shielded precast floor slabs, and combining the multiple pictures of the shielded precast floor slabs into one picture.
Further, in the above technical scheme, the step of controlling unmanned aerial vehicle to carry out image acquisition to the prefabricated floor of discernment, if the prefabricated floor image of shooting has to shelter from, then controlling unmanned aerial vehicle change position to carry out many pictures to the prefabricated floor that has to shelter from to merge into a picture with having many pictures of prefabricated floor that shelter from, specifically include:
basic collection: the unmanned aerial vehicle is controlled to collect images of the identified prefabricated floor slab, specifically, during collection, firstly, the vertical determination device arranged on a camera carried by the unmanned aerial vehicle determines that the camera is vertical, and then, the unmanned aerial vehicle is controlled to shoot the prefabricated floor slab; the general vertical determination device adopts GPS or gyro;
and (5) merging and collecting: if the photographed image of the precast floor slab is shielded, the change position of the unmanned aerial vehicle is controlled to collect multiple pictures of the shielded precast floor slab, and the multiple pictures of the shielded precast floor slab are combined into one picture, specifically:
step 1, acquiring a prefabricated floor image acquired basically as a first basic image;
step 2, searching a first image of the precast floor slab corresponding to the basic image in a precast floor slab image database according to the surface characteristics of the first basic image, and recording the first image as a reference image;
step 3, calculating the surface area difference of the first basic image and the reference image;
step 4, if the surface area difference is not more than 10%, taking the first basic image as an acquired image of the corresponding precast floor slab; if the surface area difference is greater than 10%, adjusting the position of the unmanned aerial vehicle and acquiring the surface of the corresponding prefabricated floor slab again to obtain a second basic image;
and 5, taking the image obtained by combining the second basic image and the first basic image as the first basic image, repeating the steps 3-5, and stopping repeating the steps if the step 3 is repeatedly executed for more than 6 times.
Further, in the above technical solution, the specific step of calculating the surface area difference between the first base image and the reference image includes:
preprocessing a first basic image and a reference image;
calculating the surface areas of the prefabricated floors of the first basic image and the reference image after pretreatment;
the difference between the surface areas of the prefabricated floors of the first basic image and the reference image after pretreatment is calculated.
In this step we need to calculate the surface area difference of the first base image and the reference image. To achieve this goal, we can employ the following algorithm or formula:
image preprocessing
First, we need to pre-process the first base image and the reference image in order to extract the contour of the prefabricated floor slab. We can use the following method:
converting into a gray image: the first base image and the reference image are converted into grayscale images to reduce computational complexity.
Gaussian filtering: the gray-scale image is smoothed using a gaussian filter to eliminate noise.
Edge detection: and extracting the profile of the precast floor slab by using a Canny edge detection algorithm.
Calculation of surface area
Next, we need to calculate the pre-floor surface area of the first base image and the reference image.
To this end, we can use the following method:
contour extraction: an image processing library (such as OpenCV) is used to extract contours in the edge-detected image.
Polygonal approximation: the extracted contour is approximated as a polygon to calculate an area thereof. Approximation can be performed using the Ramer-Douglas-Peucker algorithm.
Area calculation: the area of the polygon is calculated. The calculation may be performed using a sheelace formula.
Calculating the surface area difference
Finally, we need to calculate the surface area difference of the first base image and the reference image. The calculation can be performed using the following formula:
wherein A1 and A2 represent the precast floor surface areas of the first base image and the reference image, respectively, and D represents the difference in surface area.
By the above steps we can calculate the surface area difference of the first base image and the reference image.
In the above technical solution, the steps of building and training the diagnostic model of the construction quality of the precast floor slab specifically include:
building a training sample;
establishing and training a model prototype;
the step of establishing a training sample specifically comprises the following steps:
acquiring a first historical merged image set comprising a plurality of merged images;
setting flaw indexes for each combined image by adopting a MiniGPT-4 model;
deleting the combined image without any flaw in the training sample;
the method specifically comprises the following steps of:
establishing a precast floor slab construction quality diagnosis model prototype by utilizing a convolutional neural network;
and training the prototype of the precast floor slab construction quality diagnosis model by taking each combined image in the training sample as training input and the flaw index corresponding to the combined image as training output to obtain the precast floor slab construction quality diagnosis model.
In the step of building a training model prototype, we will build and train a precast floor slab construction quality diagnostic model prototype based on convolutional neural networks (Convolutional Neural Network, CNN). The specific implementation mode is as follows:
(1) Design convolutional neural network model structure
We devised a convolutional neural network model comprising a plurality of convolutional layers, a pooling layer, and a fully-connected layer. The input of the model is the preprocessed combined image, and the output is the corresponding flaw index. The specific structure is as follows:
constructing a convolutional neural network model structure: first, we construct a convolutional neural network model that consists of multiple convolutional layers, an active layer, a pooling layer, and a fully-connected layer. The convolution layer is used for extracting local features of the image, the activation layer is used for increasing nonlinearity of the model, the pooling layer is used for reducing the space dimension of the image, and the full-connection layer is used for integrating the previous features and outputting a prediction result.
Initializing model parameters: before training the model, we need to initialize the model parameters. Here, we initialize the weights of the convolutional layer and the full-connection layer using Xavier initialization method, initializing the bias term to zero.
Setting a loss function: to measure the predictive effect of the model, we need to set a loss function. In this method, we use the idea of multi-task learning to set a separate loss function for each flaw, and then weight-sum the loss functions to obtain the total loss function. Specifically, we use the mean square error (Mean Squared Error, MSE) as a loss function for each flaw, namely:
wherein L is i Loss function representing the ith flaw, i.e. [1,4 ]]N represents the number of samples, y i,j True index of the ith flaw representing the jth sample, j.epsilon.1, N]，A predictive index indicating the ith blemish of the jth sample. The loss functions of various flaws are weighted and summed to obtain a total loss function:
wherein w is i Indicating the weight of the ith flaw.
Setting an optimizer: to optimize the model parameters, we need to select an optimizer. In the method, an Adam optimizer is adopted for parameter optimization. The Adam optimizer combines the advantages of a momentum method and an RMSProp algorithm, has the characteristic of self-adaptively adjusting the learning rate, and is suitable for training a deep neural network.
Model training is carried out: inputting the training sample into a convolutional neural network model, calculating a prediction result through forward propagation, calculating a gradient according to a loss function, and updating model parameters through an optimizer. This process is repeated for a number of cycles (epoch) until the model converges.
Verification of model effect: and evaluating the prediction effect of the model on the verification set, and observing whether the loss function value is stable or not and the prediction effect of the model on various flaws. If the verification is not good, attempts may be made to adjust the model structure, loss function weights, optimizer parameters, etc., until a satisfactory result is achieved. Furthermore, in the technical proposal, the MiniGPT-4 model is a model subjected to targeted fine tuning,
the MiniGPT-4 model targeted fine tuning step specifically comprises the following steps:
acquiring a second historical merged image set comprising a plurality of merged images;
extracting features from the left part of each combined image to form a left feature matrix;
extracting features from the right part of each combined image to form a right feature matrix;
calculating a right characteristic matrix and subtracting the left characteristic matrix to obtain a difference matrix;
ViT of a MiniGPT-4 model is adopted to encode a difference matrix, so as to obtain ViT encoding;
the method comprises the steps of obtaining a flaw manual description of the right side part of each combined image, wherein the flaw manual description is performed on a corresponding prefabricated floor slab by constructors after actual observation and measurement are performed on the prefabricated floor slab corresponding to the combined image;
summarizing the manual description of flaws on the right side part of each combined image by using a Vicuna large language model of a MiniGPT-4 model, extracting flaw indexes, and recording the flaw indexes as description flaw indexes;
combining ViT coding and description blemish indices into one instruction data;
fine-tuning the MiniGPT-4 model by adopting construction data to obtain a Lora fine-tuning weight;
combining the obtained Lora fine tuning weight with the weight of the Vicuna large language model in the MiniGPT-4 model to obtain a comprehensive weight, and using the obtained comprehensive weight as the weight of the MiniGPT-4 model to finish fine tuning of the MiniGPT-4 model.
Further, in the above technical solution, the step of extracting features from the left portion of each combined image to form a left feature matrix specifically includes:
normalizing the left part of the combined image;
edge detection is carried out on the normalized left part image;
extracting LBP characteristics of the left partial image;
and fusing the image obtained by edge detection with the LBP feature, and flattening the fused feature image into a vector serving as a left feature matrix.
In the precast floor slab construction quality diagnosis method, the specific embodiment of extracting the features for the left portion of each combined image is as follows:
first, the left part of the combined image is normalized, i.e., the pixel value is mapped to [0,1 ]]Interval to eliminate the influence of illumination, exposure, etc. The normalization process can be expressed as:
wherein I represents the original image, I norm Representing normalized image, I min And I max Representing the minimum and maximum pixel values in the original image, respectively.
For normalized left part image I norm Edge detection is performed to highlight the shape and imperfections of the precast floor slabs. A Canny edge detection algorithm may be used here, the calculation process of which comprises the following steps:
a) Pair I norm And (5) performing Gaussian filtering to eliminate noise. Gaussian filtering can be expressed as: i smooth ＝I norm ×G σ ；
Wherein I is smooth Representing the filtered image, G σ A gaussian filter with standard deviation σ is shown.
b) Calculation I smooth Gradient magnitude and direction of (c). The Sobel operator can be used to calculate the gradients in the horizontal and vertical directions, respectively:
G x ＝I smooth ×S x ；G y ＝I smooth ×S y ；
wherein G is x And G y Gradient images respectively representing the horizontal direction and the vertical direction S x And S is y The Sobel operator in the horizontal direction and the vertical direction are represented respectively. The gradient magnitude G and direction θ can be calculated as:
c) Non-maximum suppression is performed on the gradient direction θ to obtain refined edges. Specifically, checking whether the gradient amplitude G of each pixel point is larger than the gradient amplitude G of other points in the neighborhood of the pixel point in the gradient direction theta, and if so, reserving the point; otherwise, the gradient magnitude for that point is set to zero.
d) And applying a double-threshold algorithm to binarize the gradient image after non-maximum suppression. Setting a high threshold T high And a low threshold T low If the gradient amplitude of a pixel point is greater than T high It is marked as a strong edge; if the gradient amplitude is at T low And T high And then mark it as a weak edge; if the gradient amplitude is smaller than T low It is set to zero. Finally, for each weak edge pixel point, if a strong edge point exists in the neighborhood of the weak edge pixel point, reserving the point; otherwise, it is set to zero.
Texture features of the left partial image are extracted. A Local Binary Pattern (LBP) algorithm may be employed, the calculation process of which is as follows:
a) For I norm Consider other pixels within its 3x3 neighborhood. Comparing the pixel value in the adjacent area with the central pixel value, and marking the pixel value in the adjacent area as 1 if the pixel value in the adjacent area is more than or equal to the central pixel value; otherwise, it is marked as 0.
b) The marks in the neighborhood are connected in a clockwise direction to obtain a binary number. The binary number is converted into a decimal number and serves as the LBP value for the center pixel.
c) An LBP histogram of the entire image is calculated for representing the texture features of the image.
And fusing the image obtained by edge detection with the LBP features to form a left feature matrix. Specifically, the LBP histogram may be multiplied by the binary image obtained by Canny edge detection element by element to obtain a fused feature image. The fused feature image is then flattened into a vector as the left feature matrix.
In summary, we can represent the extracted features of the left part of the combined image as a left feature matrix, and the steps of extracting the features of the right part of the combined image are the same.
A second aspect of the present invention provides a computer readable storage medium having stored therein program instructions which when executed are adapted to carry out a precast floor deck construction quality diagnostic method as described above.
A third aspect of the present invention provides a precast floor plank construction quality diagnostic system, comprising an unmanned aerial vehicle carrying a camera, wherein a control chip in the unmanned aerial vehicle comprises the computer readable storage medium.
The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any person skilled in the art will readily recognize that variations or substitutions are within the scope of the present invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.

Claims (10)

1. The method for diagnosing the construction quality of the precast floor slab is characterized by comprising the following steps of:
s10, acquiring a first image before construction and a second image set after construction of each precast floor slab, and establishing a precast floor slab image database by using the first images of all precast floor slabs; each image in the first image and the second image set is a surface image of the precast floor slab, the acquisition time of the first image is the time when the precast floor slab is to be constructed when being taken out of a warehouse, and the second image set is an image group formed by acquisition according to a specified time interval after the precast floor slab is installed at a set position;
s20, preprocessing each image in the first image set and the second image set;
s30, for each prefabricated floor slab, carrying out left-right combination and merging on each preprocessed image in the corresponding preprocessed first image and the corresponding preprocessed second image set to form a merged image set containing a plurality of merged images, wherein the left part after merging is the corresponding preprocessed first image, and the right part is the preprocessed image in the second image set;
s40, for each prefabricated floor slab, calculating each combined image in the combined image set by using a pre-trained prefabricated floor slab construction quality diagnosis model to obtain flaw index of the corresponding prefabricated floor slab as construction quality diagnosis data, and sending the flaw index to constructors;
wherein the flaws include cracks, pitting, bubbles and surface tilting, the flaw index comprises 0-5 levels, a numerical value indicates the severity of the flaw, a larger numerical value indicates a greater severity of the flaw, and 0 indicates no flaws of this type; the flaw index is a vector representing the level of each flaw.
2. The method for diagnosing construction quality of precast floor slabs according to claim 1, wherein the collecting mode of each image in the second image set is to collect the surface image of each precast floor slab by using an unmanned aerial vehicle carrying a camera, and the specific steps include:
acquiring a three-dimensional structure of a large-scale building for construction;
setting a movement path of the unmanned aerial vehicle according to the three-dimensional structure of the constructed large building;
controlling the unmanned aerial vehicle to enter the interior of a large-scale building for construction according to a motion path to shoot;
and acquiring the surface image of each prefabricated floor slab shot by the unmanned aerial vehicle.
3. The method for diagnosing the construction quality of the prefabricated floor according to claim 2, wherein the step of controlling the unmanned aerial vehicle to enter the interior of the large-scale building to be constructed according to the movement path for shooting comprises the following steps of;
controlling the unmanned aerial vehicle to enter a large-scale building for construction and fly according to a motion path;
the unmanned aerial vehicle is controlled to identify the prefabricated floors in the way, wherein the identification method is to train a neural network by using a first image of each prefabricated floor, and identify the prefabricated floors by using the trained neural network;
the unmanned aerial vehicle is controlled to carry out image acquisition to the prefabricated floor of discernment, if the prefabricated floor image of shooting has to shelter from, then the unmanned aerial vehicle change position of control carries out many picture acquisition to the prefabricated floor that has to shelter from to it merges to have a picture to shelter from many pictures of prefabricated floor.
4. The method for diagnosing construction quality of a prefabricated slab according to claim 3, wherein the step of controlling the unmanned aerial vehicle to perform image acquisition on the identified prefabricated slab, if the photographed image of the prefabricated slab is blocked, controlling the unmanned aerial vehicle to change the position to perform multi-picture acquisition on the blocked prefabricated slab, and combining the pictures of the blocked prefabricated slab into one picture specifically comprises the following steps:
basic collection: the unmanned aerial vehicle is controlled to collect images of the identified prefabricated floor slab, specifically, during collection, firstly, the vertical determination device arranged on a camera carried by the unmanned aerial vehicle determines that the camera is vertical, and then, the unmanned aerial vehicle is controlled to shoot the prefabricated floor slab;
and (5) merging and collecting: if the photographed image of the precast floor slab is shielded, the change position of the unmanned aerial vehicle is controlled to collect multiple pictures of the shielded precast floor slab, and the multiple pictures of the shielded precast floor slab are combined into one picture, specifically:
step 1, acquiring a prefabricated floor image acquired basically as a first basic image;
step 2, searching a first image of the prefabricated floor slab corresponding to the basic image in the prefabricated floor slab image database according to the surface characteristics of the first basic image, and recording the first image as a reference image;
step 3, calculating the surface area difference of the first basic image and the reference image;
step 4, if the surface area difference is not more than 10%, taking the first basic image as an acquired image of the corresponding precast floor slab; if the surface area difference is greater than 10%, adjusting the position of the unmanned aerial vehicle and acquiring the surface of the corresponding prefabricated floor again to obtain a second basic image;
and 5, taking the image obtained by combining the second basic image and the first basic image as the first basic image, repeating the steps 3-5, and stopping repeating the steps if the step 3 is repeatedly executed for more than 6 times.
5. The method for diagnosing construction quality of a precast floor plank according to claim 4, wherein the step of calculating the difference between the surface areas of the first basic image and the reference image comprises:
preprocessing the first basic image and the reference image;
calculating the surface areas of the prefabricated floors of the first basic image and the reference image after pretreatment;
and calculating the difference between the surface areas of the prefabricated floors of the first basic image and the reference image after pretreatment.
6. The method for diagnosing the construction quality of the precast floor plank according to claim 1, wherein the steps of establishing and training the diagnosis model for the construction quality of the precast floor plank specifically comprise the following steps:
building a training sample;
establishing and training a model prototype;
the step of establishing a training sample specifically includes:
acquiring a first historical merged image set comprising a plurality of merged images;
setting flaw indexes for each combined image by adopting a MiniGPT-4 model;
deleting the combined image without any flaw in the training sample;
the step of building the training model embryonic form specifically comprises the following steps:
establishing a precast floor slab construction quality diagnosis model prototype by utilizing a convolutional neural network;
and training the prototype of the precast floor slab construction quality diagnosis model by taking each combined image in the training sample as training input and the flaw index corresponding to the combined image as training output to obtain the precast floor slab construction quality diagnosis model.
7. The method for diagnosing the construction quality of the precast floor plank according to claim 6, wherein the MiniGPT-4 model is a targeted fine-tuned model,
the step of targeted fine tuning of the MiniGPT-4 model specifically comprises the following steps:
acquiring a second historical merged image set comprising a plurality of merged images;
extracting features from the left part of each combined image to form a left feature matrix;
extracting features from the right part of each combined image to form a right feature matrix;
calculating a right characteristic matrix and subtracting the left characteristic matrix to obtain a difference matrix;
adopting ViT of a MiniGPT-4 model to encode the difference matrix to obtain ViT codes;
the method comprises the steps of obtaining a flaw manual description of the right side part of each combined image, wherein the flaw manual description is performed on a corresponding prefabricated floor slab by constructors after actual observation and measurement are performed on the corresponding prefabricated floor slab according to the combined image;
summarizing the manual description of flaws on the right side part of each combined image by using a Vicuna large language model of a MiniGPT-4 model, extracting flaw indexes, and recording the flaw indexes as description flaw indexes;
combining said ViT code and said description blemish index into one construction data;
fine-tuning the MiniGPT-4 model by adopting the construction data to obtain a Lora fine-tuning weight;
combining the obtained Lora fine tuning weight with the weight of the Vicuna large language model in the MiniGPT-4 model to obtain a comprehensive weight, and using the obtained comprehensive weight as the weight of the MiniGPT-4 model to finish fine tuning of the MiniGPT-4 model.
8. The method for diagnosing construction quality of a prefabricated floor according to claim 7, wherein the step of extracting features from the left portion of each of the combined images to form a left feature matrix comprises the steps of:
normalizing the left part of the combined image;
edge detection is carried out on the normalized left part image;
extracting LBP characteristics of the left partial image;
and fusing the image obtained by edge detection with the LBP feature, and flattening the fused feature image into a vector serving as a left feature matrix.
9. A computer readable storage medium having stored therein program instructions which, when executed, are adapted to carry out a method of diagnosing the quality of a precast floor plank construction according to any one of claims 1 to 8.
10. A precast floor plank construction quality diagnostic system comprising an unmanned aerial vehicle carrying a camera, the unmanned aerial vehicle having a control chip comprising a computer readable storage medium of claim 9.
CN202311078117.7A 2023-08-24 2023-08-24 Precast floor slab construction quality diagnosis method, medium and system Active CN117095294B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311078117.7A CN117095294B (en) 2023-08-24 2023-08-24 Precast floor slab construction quality diagnosis method, medium and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311078117.7A CN117095294B (en) 2023-08-24 2023-08-24 Precast floor slab construction quality diagnosis method, medium and system

Publications (2)

Publication Number Publication Date
CN117095294A true CN117095294A (en) 2023-11-21
CN117095294B CN117095294B (en) 2024-06-25

Family

ID=88773099

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311078117.7A Active CN117095294B (en) 2023-08-24 2023-08-24 Precast floor slab construction quality diagnosis method, medium and system

Country Status (1)

Country Link
CN (1) CN117095294B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118313051A (en) * 2024-06-11 2024-07-09 中建安装集团黄河建设有限公司 Construction method of large-span steel structure

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110599481A (en) * 2019-09-18 2019-12-20 成都数之联科技有限公司 Cement prefabricated plate detection system based on image recognition
CN112465748A (en) * 2020-11-10 2021-03-09 西南科技大学 Neural network based crack identification method, device, equipment and storage medium
KR20210053509A (en) * 2019-11-04 2021-05-12 강원대학교산학협력단 Structure defect detecting system using drone
CN113340240A (en) * 2021-06-07 2021-09-03 高速铁路建造技术国家工程实验室 Detection system and detection method for flatness of concrete precast slab
US20230003663A1 (en) * 2020-03-31 2023-01-05 Fujifilm Corporation Inspection support device, inspection support method, and inspection support program
CN115578693A (en) * 2022-11-09 2023-01-06 广东电网有限责任公司 Construction safety early warning method and device based on significance neural network model
US20230075504A1 (en) * 2020-05-29 2023-03-09 Fujifilm Corporation Damage diagram creation support method and damage diagram creation support device

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110599481A (en) * 2019-09-18 2019-12-20 成都数之联科技有限公司 Cement prefabricated plate detection system based on image recognition
KR20210053509A (en) * 2019-11-04 2021-05-12 강원대학교산학협력단 Structure defect detecting system using drone
US20230003663A1 (en) * 2020-03-31 2023-01-05 Fujifilm Corporation Inspection support device, inspection support method, and inspection support program
US20230075504A1 (en) * 2020-05-29 2023-03-09 Fujifilm Corporation Damage diagram creation support method and damage diagram creation support device
CN112465748A (en) * 2020-11-10 2021-03-09 西南科技大学 Neural network based crack identification method, device, equipment and storage medium
CN113340240A (en) * 2021-06-07 2021-09-03 高速铁路建造技术国家工程实验室 Detection system and detection method for flatness of concrete precast slab
CN115578693A (en) * 2022-11-09 2023-01-06 广东电网有限责任公司 Construction safety early warning method and device based on significance neural network model

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118313051A (en) * 2024-06-11 2024-07-09 中建安装集团黄河建设有限公司 Construction method of large-span steel structure

Also Published As

Publication number Publication date
CN117095294B (en) 2024-06-25

Similar Documents

Publication Publication Date Title
KR102166458B1 (en) Defect inspection method and apparatus using image segmentation based on artificial neural network
CN112700444B (en) Bridge bolt detection method based on self-attention and central point regression model
CN104599280B (en) A kind of crack on road crosspoint extracting method and system
CN113284107B (en) Attention mechanism-introduced improved U-net concrete crack real-time detection method
CN117095294B (en) Precast floor slab construction quality diagnosis method, medium and system
CN115482195A (en) Train part deformation detection method based on three-dimensional point cloud
CN110909657A (en) Method for identifying apparent tunnel disease image
Santos et al. Detection of exposed steel rebars based on deep-learning techniques and unmanned aerial vehicles
CN116823800A (en) Bridge concrete crack detection method based on deep learning under complex background
CN115995056A (en) Automatic bridge disease identification method based on deep learning
CN114548253A (en) Digital twin model construction system based on image recognition and dynamic matching
CN118225803B (en) Visual detection method for appearance of blade surface of bulldozer
CN115861226A (en) Method for intelligently identifying surface defects by using deep neural network based on characteristic value gradient change
CN117422699A (en) Highway detection method, highway detection device, computer equipment and storage medium
CN117237902A (en) Robot character recognition system based on deep learning
CN115439763A (en) Water engineering surface crack detection method based on YOLOv5 model
AVENDAÑO Identification and quantification of concrete cracks using image analysis and machine learning
Shajahan et al. Automated inspection of monopole tower using drones and computer vision
CN118469933A (en) Identification and quantification method for apparent defects and diseases of concrete structure
CN116434087B (en) Concrete crack identification method and device based on GOA-SVM cooperative algorithm and unmanned aerial vehicle
CN118072193A (en) Dam crack detection method based on unmanned aerial vehicle image and deep learning
CN115797914A (en) Metallurgical crane trolley track surface defect detection system
CN112164024B (en) Concrete surface crack detection method and system based on field self-adaption
CN115620121A (en) Photoelectric target high-precision detection method based on digital twinning
CN117726239B (en) Engineering quality acceptance actual measurement method and system

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant