KR102555009B1 - Damage detection method based on automatic image acquisition control - Google Patents

Abstract

The present invention relates to a damage detection method based on automatic image acquisition control.
A damage detection method based on automatic image acquisition control according to the present invention includes the steps of (a) acquiring an image based on automatic control; (b) constructing damaged image data using similarly damaged objects; and (c) detecting damage of at least one of concrete cracks, network cracks, exfoliation, damage, exposure of reinforcing bars, material separation, efflorescence, corrosion of steel materials, and peeling of paint using a deep learning model.

Description

Image Acquisition Automatic Control Based Damage Detection Method {DAMAGE DETECTION METHOD BASED ON AUTOMATIC IMAGE ACQUISITION CONTROL}

The present invention relates to a damage detection method based on automatic image acquisition control.

All structures undergo deterioration or damage over time due to natural and environmental factors.

In order to maintain the functions of aging structures, safety inspections are conducted at regular intervals. Most of the most basic inspection items, exterior inspections, depend on visual inspections by manpower, making accurate quantification difficult and reliability of inspection results due to accumulated damage labeling errors. There is a problem with this fall.

The series of processes of recording the type and shape of the damage investigated during the exterior inspection in the field field, recording the type and shape of the damage on the exterior survey network map based on the results of the investigation in the office, and preparing a damage inventory table are low in productivity. It corresponds to simple repetitive work, and there is an inefficient problem.

According to the prior art, a damage detection technology using machine learning has been proposed, but since the Gray Scale method is used, all parts with color differences are recognized as cracks, so the false detection rate is very high, and the non-damaged part is recognized as damage. Therefore, since the photographed image data must be confirmed by manpower and then the drawing work is performed, there is a problem in that the subjective judgment of the checker is intervened and errors caused by manpower are generated. In addition, cross-section damage occurring to structures such as whitening, peeling, exposure of reinforcing bars, and material separation has a limitation that cannot be detected.

The present invention has been proposed to solve the above problems, and proposes an image processing technology that automatically acquires high-resolution structural images and simultaneously analyzes multiple damages using artificial intelligence learned from selected big data. The purpose of this study is to provide a damage detection method based on automatic image acquisition control that supports automated exterior inspection of structures (bridges, tunnels, retaining walls, ports, power generation facilities, etc.) constructed with concrete and steel.

The present invention relates to a damage detection method based on automatic image acquisition control.

A damage detection method based on automatic image acquisition control according to the present invention includes the steps of (a) acquiring an image based on automatic control; (b) constructing damaged image data using similarly damaged objects; and (c) detecting damage of at least one of concrete cracks, network cracks, exfoliation, damage, exposure of reinforcing bars, material separation, efflorescence, corrosion of steel materials, and peeling of paint using a deep learning model.

According to the present invention, it is effective to establish a preemptive preventive management system for structures by securing objective and quantified data, and to prevent the occurrence of safety accidents by inspectors through automated precise safety diagnosis of facilities.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned will be clearly understood by those skilled in the art from the description below.

1 illustrates operation of a high-resolution automatically controlled image capture unit according to an embodiment of the present invention.
2 shows a structure damage detection process using artificial intelligence.
3 shows an example of a damaged-like object synthesized image.
4 shows a process of detecting 9 types of damage using an artificial intelligence model for each damage and a secondary classification criterion.
5 shows an artificial intelligence-based rapid learning data construction system framework.
6 shows a damage quantification algorithm.
7 illustrates ICT-based safety diagnosis automation data management.
8 illustrates a damage detection method based on automatic image acquisition control according to an embodiment of the present invention.
9A to 9D show a high-resolution image acquisition device according to an embodiment of the present invention.
10 shows a lighting device according to an embodiment of the present invention.
11 shows a high-resolution image acquisition device in which a lighting unit is disposed according to an embodiment of the present invention.
12 shows photographing equipment disposed on a moving cart according to an embodiment of the present invention.
13 shows a system configuration diagram according to an embodiment of the present invention.
14 shows a software configuration diagram according to an embodiment of the present invention.
15 shows a control and measurement device according to an embodiment of the present invention.
16 shows data flow according to an embodiment of the present invention.
17 illustrates an image merging process according to an embodiment of the present invention.

The foregoing and other objects, advantages and characteristics of the present invention, and a method of achieving them will become clear with reference to the detailed embodiments described below in conjunction with the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms, and only the following embodiments provide the purpose of the invention, As only provided to easily inform the configuration and effect, the scope of the present invention is defined by the description of the claims.

Meanwhile, terms used in this specification are for describing the embodiments and are not intended to limit the present invention. In this specification, singular forms also include plural forms unless specifically stated otherwise in a phrase. As used herein, “comprises” and/or “comprising” means the presence of one or more other components, steps, operations, and/or elements in which a stated component, step, operation, and/or element is present. or added.

Hereinafter, referring to FIG. 1 , operation of a high-resolution automatically controlled image capture unit according to an embodiment of the present invention will be described.

Using a smartphone and a DSLR camera, high-resolution images that can be identified from a distance of less than 0.1mm in damage width are taken, and this is to evaluate cracks from 0.1mm in consideration of concrete durability and watertightness.

The photographing unit is operated at a shooting angle of 4 axes ±135° and a rotational speed of up to 45deg/sec to perform photographing, and it is possible to photograph the bottom plate of the bridge and the underside of the girder after installing the photographing unit.

Through lightweight high-resolution image shooting using a smartphone, it is possible to inspect the inside of a structure within a shooting distance of 20 meters (eg, inside a bridge box, inside a repair facility, inside a building, etc.), and precisely control the shooting angle of a smartphone in 0.01 degree increments. and inspect the exterior of large structures within a distance of 50 meters using a DSLR camera. Image capture is automatically controlled by the photographing unit operation program, images are transmitted in real time, and focusing information is transmitted. A photographing app is installed in the smartphone to receive a command generated by a photographing unit operating program and control a camera built in the smartphone.

The captured high-resolution image is used for damage detection, and the captured image and information are transmitted to data management software and captured. In relation to the photographing information, information acquired by the distance measurement sensor with a measurement accuracy of ±1 mm or less at a distance of 20 meters or more from the structure is transmitted. In addition, the photographing distance, angle, and current location information are measured and transmitted. The photographed image, photographing speed, and focusing information are transmitted by the photographing unit operation program. Lighting is arranged in the photographing unit so that a damage width of 0.1 mm can be identified at a distance of 20 meters.

Hereinafter, an artificial intelligence-based damage detection automation module according to an embodiment of the present invention will be described with reference to FIGS. 2 to 6 . Figure 2 shows a structure damage detection process using artificial intelligence, Figure 3 shows an example of a similarly damaged object composite image, Figure 4 shows nine types of damage detection processes using artificial intelligence models for each damage and secondary classification criteria 5 shows an AI-based rapid learning data construction system framework, and FIG. 6 shows a damage quantification algorithm.

In order to automatically generate an exterior survey network map and a damage quantity table, an artificial intelligence (deep learning) model is applied to high-resolution images obtained through the operation of the filming department to determine the deterioration of structures and damage occurring during use (crack/reticular cracks in concrete). /exfoliation/damage/rebar exposure/material separation/whitening, steel corrosion/film peeling) are detected and quantified.

When detecting damage, image augmentation techniques are applied to consider image shooting conditions in various sites, and illumination, blurring, and shooting angles are set as shooting conditions to be considered.

For damage analysis, image stitching optimized for facility inspection using shooting distance and angle information is performed.

The order of image stitching is: (a) Matching the same feature point after analyzing each feature point in the two images to be connected. (b) Distortion between images is determined based on matching information. (c) The image pixels are rearranged by inverting the distortion information of one image. (d) Depending on the degree of distortion of the rearranged images, an empty space is created between the rearranged images, and this is filled with the average value of nearby images. (e) The process of (a) to (d) is repeated until all images become one image. (f) The matched image is corrected with the average value of brightness, and the differences in business cards caused by the continuation of the shooting time are removed. (g) In order for the analysis program to read the image, the completed image is cut and saved to a preset pixel size (20,000Х20,000 pixel size) without overlapping parts, and 1/10 so that the user can visually check the completed image. save by size

Referring to FIG. 3, in relation to building a damaged image data set, a training image for distinguishing similarly damaged objects is built. In general, when detecting damage from images, the accuracy of damage detection is lowered due to objects (similarly damaged objects) having similar characteristics (shape, color, etc.) to the damage. For example, in the case of concrete cracks, formwork used in concrete It tends to be confused by linear marks and cobwebs and contaminants. According to the present invention, after investigating the type of similarly damaged objects for each damage, images of similarly damaged objects are collected, and the damaged objects are synthesized with images in which similarly damaged objects exist to construct synthetic learning data for excluding similarly damaged objects. . Using the damaged object learning data and similarly damaged object synthesis learning data, it is possible to accurately detect only damage while excluding similarly damaged objects.

A segmentation-based artificial intelligence model that automatically detects damage using damage learning data and similar damage object synthesis learning data is used, and each damage detection model is used in parallel.

Referring to FIG. 4, since cracks, network cracks, exfoliation and breakage have high similarities, they are primarily grouped into cracks and breaks to detect, and then a classification criterion is applied to make a detailed determination.

In order to develop an artificial intelligence model with high accuracy, a large number of accurate training data must be secured, but it takes a lot of time to generate training data, and the most time-consuming part of producing training data is the part of searching for damage candidates on the image. and the time to perform direct labeling on it.

Referring to FIG. 5, by using an artificial intelligence model, damage candidates are primarily detected and displayed (labeled), and only the marked damage candidates are provided to experts so that the experts can review and correct them, thereby performing labeling. It has the effect of reducing data construction time by more than 10 times.

The learning data rapid construction framework trains a pre-detection AI model capable of maximizing the detection performance of damage candidates, and minimizes false detections of damage by detecting only a designated area by the pre-detection AI model.

According to the present invention, it collects selected learning images for damage, quickly builds learning data using artificial intelligence, performs damage detection using artificial intelligence (deep learning), and detects damage using artificial intelligence. After quantifying the damage, an external survey network map and damage quantity table are automatically created.

Referring to FIG. 6, damage is classified into linear damage (crack, exposure of reinforcing bars) and area type damage (exfoliation, efflorescence, corrosion, coating peeling, etc.), and a damage quantification method suitable for each type of damage is determined. In the case of linear damage, image processing techniques such as filtering and dilation-erosion are applied to determine the exact damaged pixel, and damage information (width, length) is calculated in units of pixels. In the case of area-type damage, filtering and circumscribed boxes are The damage area is calculated in units of pixels.

By using the Blob Detection (stain detection) function, etc., minute false detection damage is removed, and the damage information calculated in pixel units is used as a reference distance (the shooting distance from the image sensor to the image, or the actual length of the object that can be identified in the image). It is used to quantify damage information in real units by converting it into actual units (mm, etc.).

Hereinafter, with reference to FIG. 7, ICT-based safety diagnosis automation data management will be described.

The captured image and shooting information are transmitted to the head office server or cloud, and managed using an automated data management program.

According to the present invention, the structure inspection and diagnosis results solve the problem that the judgment criteria are different for each inspector, and it is difficult to check the progress of damage due to the accumulation of errors in marking the location of damage in the field, and to check the progress of damage during periodic imaging. It has the effect of establishing a preemptive prevention system for structures.

Using ICT-based safety diagnosis automation data management program, it is possible to check the temporal change of structural damage on the web.

8 illustrates a damage detection method based on automatic image acquisition control according to an embodiment of the present invention.

The damage detection method based on the automatic control of image acquisition according to the present invention includes acquiring an image based on automatic control (S810), constructing damage image data using a similar damage object (S820), and using a deep learning model. and detecting damage of at least one of concrete cracks, network cracks, exfoliation, damage, exposure of reinforcing bars, material separation, efflorescence, corrosion of steel materials, and coating peeling (S830).

Step S810 acquires a high-resolution image of 0.1 mm/pixel from a preset distance in consideration of concrete durability and watertightness.

In step S810, photographing information including focusing information, photographing distance information, photographing angle information, and current location information is acquired.

In step S810, image stitching is performed using the photographing distance information and the photographing angle information, matching feature points in the image to be connected, and determining whether or not distortion between images is warped using the matching information to calculate distortion information, and calculating the distortion information Inverse calculation is performed and image pixels are rearranged to fill the empty space between images with the average value of neighboring images.

In step S820, the image of the similarly damaged object is collected for each damage, and the damaged object is synthesized with the image in which the similarly damaged object exists to construct synthetic learning data.

In step S820, labeling is performed using the damage candidate detection result, and training data is constructed using feedback on the displayed damage candidate.

Step S830 performs damage detection by utilizing a plurality of damage detection models in parallel using a segmentation-based artificial intelligence model.

Step S830 performs detection through grouping based on similarity for each damage.

In step S830 , linear damage and area-type damage are separately detected, a damaged pixel is determined for the linear damage, damage information is calculated in units of pixels, and a damage area for the area-type damage is calculated in units of pixels.

Step S830 quantifies the damage information in real units by converting the damage information calculated in units of pixels into actual units using at least one of shooting distance information and length information of an object identified in the image.

The photographing device according to the embodiment of the present invention may use a smartphone whose angle is adjusted from 0 to 270 degrees or a DSLR camera. When using a smartphone, the shooting screen can be checked remotely through communication, and when using a DSLR camera, the inspector can check the shooting screen.

Through the high-resolution automatically controlled image acquisition system according to an embodiment of the present invention, when a high-resolution image is automatically obtained using a shooting hardware that improves shooting speed, shooting angle, weight, and assembly complexity, and there is no natural light using a lighting device It supports smooth shooting, provides image merging, preview of sequentially merging images, and editing functions, and the effect of easy program handling can be expected.

A high-resolution automatically controlled image acquisition system according to an embodiment of the present invention includes a photographing device for performing photographing and transmitting images and photographing information, a lighting device for irradiating light on a photographing area, and the photographing device. A computing device that performs integrated control of a photographing device that adjusts a photographing angle of the device and the photographing equipment; and a storage unit for storing high-resolution images.

According to another embodiment of the present invention, it is possible to control a rotation operation by using a smartphone or a DSLR camera as a photographing device and transmitting a control signal to the photographing equipment on which the photographing device is mounted.

The computing device transmits a motor control signal to the photographing equipment, and the photographing equipment performs rotation angle control for two orthogonal axes using the motor control signal.

The photographing equipment measures distance information to a subject and position information of a motor in real time using distance sensor information and limit sensor information.

The photographing equipment is supported by a tripod structure, and each leg constituting the tripod is adjusted in a different height to maintain a horizontal level for photographing according to the topography.

The photographing equipment is disposed on a movable carriage, so that it can be easily moved from a photographing location.

The photographing device transmits camera and video state information in real time.

The storage unit includes at least one of a cloud server and a NAS.

The lighting unit includes a communication module, receives a control signal from the computing device, and changes a lighting irradiation area.

The lighting unit is interlocked when the photographing device is driven by controlling the photographing area, and an angle of the lighting irradiation area is changed when the angle of the photographing area is controlled.

According to the present invention, there are advantages in that the shooting speed is increased, the shooting angle of 0 to 270 degrees is secured, the structure is lightened and the assembly is simplified, there are advantages in that focusing information and captured images can be transmitted, and handling is easy. It provides one image merging program and has the advantage of being able to check damage anywhere in real time by utilizing the cloud or NAS.

9A to 9D show a high-resolution image acquisition device according to an embodiment of the present invention, wherein FIG. 9A is a front view, FIG. 9B is a left view, FIG. 9C is a right view, and FIG. 9D is a rear view.

A high-resolution image acquisition device according to an embodiment of the present invention includes a motor generator 101, a base panel 102, a y-axis rotation base 103, a tripod fixing bracket 104, a bearing 105, and an x-axis rotation column 106. ), tripod 107, x-axis rotation pulley 108, fixing bracket 109, distance sensor 110 for measuring the distance between the camera and subject, x-axis rotation bearing 111, y-axis rotation pulley 112 ), camera fixing part 113, y-axis rotation stopper 114, camera fixing bracket 115, lower motor cover 116, smartphone 117, timing belt 118, x-axis motor rotation shaft assembly 119 ).

According to an embodiment of the present invention, a sensor (encoder) for measuring a rotation angle of a rotation transfer device is provided.

The tripod 107 can be adjusted to a preset height, and the adjustable height is manufactured from a minimum of 100 mm to a maximum of 500 to 700 mm. Each leg of the tripod 107 can be adjusted in height so that it can be leveled without affecting the installation terrain, and is manufactured in a prefabricated manner for easy movement, and can be folded and stored when disassembled.

According to an embodiment of the present invention, the camera is rotated in up, down, left and right directions using a servo stepping motor (a driving motor for precisely controlling equipment), and up, down, left and right angle rotation is possible using a plurality of rotation devices.

The rotation angle is set to move in the range of 0 to 270 degrees up and down, left and right, without interference.

According to an embodiment of the present invention, it is possible to perform up, down, left, right, and tilt using a stepping motor, simplify the electric circuit of the driving unit with power consumption DC 24V, and the rotation radius of each axis is set to 270 degrees.

According to an embodiment of the present invention, there are advantages of simplifying the circuit through unification of 24V power consumption, reducing the volume by integrally configuring the motor and the driver, and minimizing the wiring connection.

According to an embodiment of the present invention, using a drive-integrated stepping motor, it is possible to omit an inverter required for driving a motor, and a motor driver (a motor control device that connects a motor and receives a control signal to drive a motor) is integrated through a configuration It is possible to reduce the volume of the control box and simplify wiring.

According to an embodiment of the present invention, by arranging the stepping motor inside the mechanical unit, the center of gravity is aligned to the center to secure stability when installing the equipment and reduce the overall volume.

Referring to FIG. 10, when there is no natural light, a high-resolution image can be acquired using a separate lighting device, and referring to FIG. 11, a high-resolution image can be obtained even when there is no natural light using a lighting device included in the integrated equipment. do.

10 shows a lighting device according to an embodiment of the present invention.

The lighting device according to the embodiment of the present invention is provided separately from the image acquisition device, and a plurality of lighting units connected to the tripod structure are installed. In this case, the lighting unit may include a communication module, and may control the lighting irradiation area by receiving a control signal from the control device. The control signal is determined by comprehensively examining the shooting area of the camera and the area requiring light irradiation, and is transmitted to the communication module of the lighting unit, so that the lighting area of the lighting unit can be automatically controlled.

11 shows a high-resolution image acquisition device in which a lighting unit is disposed according to an embodiment of the present invention.

The lighting units 310a and 310b may be applied in a manner that is disposed in a predetermined area of the high-resolution image capture device and radiates light to a photographing area.

In this case, the irradiation area of the lighting unit may be controlled by receiving a separate control signal, and when driven by the imaging area control for the imaging unit, the irradiation area may be controlled in conjunction with this.

According to an embodiment of the present invention, lighting (eg, 200W lighting) with a predetermined output is installed on the left/right side of the photographing unit to irradiate light when photographing in the absence of natural light.

Since the lighting units 310a and 310b are disposed in a detachable form, it is possible to reduce the volume when stored.

According to an embodiment of the present invention, installation is simplified by using a tripod, and weight is reduced through a structure made of aluminum. As another example, as shown in FIG. 20 , as described above, by arranging the photographing equipment on the moving cart 2010, there is an effect of facilitating movement.

13 shows a system configuration diagram according to an embodiment of the present invention.

The daytime imaging equipment 401 performs motor control, measures distance, and acquires data. Daytime shooting equipment means shooting equipment used when there is natural light.

The night photographing equipment 402 performs motor control, measures distance, acquires data, and performs lighting control. Night photography equipment means photography equipment used when there is no natural light.

The smartphone 403 performs photographing, image transmission, and photographing information transmission.

The wired/wireless router 402 for daytime shooting equipment and the wired/wireless router 405 for nighttime shooting equipment create a closed Wi-Fi network and are securely connected by SSID and PW.

The notebook 406 is installed with integrated control software and analysis software.

The LTE or 5G modem 407 supports connection to the LTE or 5G network and stores images in the cloud or NAS 408 .

According to an embodiment of the present invention, by configuring Wi-Fi for the system to achieve wireless, through which equipment assembly is simplified and problems of non-coupling and misconnection are solved. A nighttime shooting device 404 is provided separately from the daytime shooting device 401, and as the shooting is performed with a smartphone 403, there is an effect of miniaturization and light weight. The notebook 406 is connected to an LTE or 5G network and automatically backs up captured images to the cloud or NAS 408.

All communication is carried out by TCP/IP communication using a wireless closed Wi-Fi network, and the integrated control program (laptop), shooting equipment control program (embedded PC), and smartphone app attempt to establish a communication connection at the same time as the program runs, and lose access. Attempts to reconnect infinitely until the end of the program.

14 shows a software configuration diagram according to an embodiment of the present invention.

The software configuration is classified into integrated control software and sub-control software that controls and measures sub-devices for the purpose of distributed control.

All subordinate control software transmits and receives control, measurement, and equipment status information through wireless communication with the integrated control software through Wi-Fi, and the integrated control software performs interface with the user and overall control of shooting.

Daytime shooting control software is installed in the embedded computer and the daytime shooting equipment 510 . A night photographing control software is installed in the embedded computer and the night photographing equipment 520 . The embedded computer program is built into and driven in the equipment, and transmits signals and data files between devices. It provides device inspection and notification check function, real-time measurement and motor control function of data acquisition equipment, and real-time signal and file relay function by wirelessly connecting to smartphones and laptops. The photographing equipment program and the lighting equipment program control the motor by receiving movement information of the motor necessary for photographing from the integrated control software, and measure distance information from the structure through a distance sensor. In addition, limit sensor information for checking the state of the motor is obtained. The daytime shooting control software and the nighttime shooting control software receive information on motor operation and transmit measurement information through communication with the integrated control software. In addition, it controls a plurality of motors and receives motor status and encoder (angle) values. In addition, distance information and motor location information are measured in real time through a distance sensor and a limit sensor.

A shooting app is installed in the smartphone 530 . The shooting app captures an image according to the control signal and transmits the measured image. It controls the camera by receiving the real-time control signal, measures the image, and transmits the measured image in real time. Camera and image status (focus, image information) is transmitted in real time. The shooting app runs automatically when the smartphone is turned on and runs in the background. After that, when communication with the integrated control software is connected, it is automatically changed to the shooting mode, and shooting proceeds according to the command of the integrated control software. In the shooting mode, information such as focus information and image files is transmitted. When the shooting is completed or the user forcibly stops it, the communication is released and converted to standby mode again.

In the notebook 540, integrated control software and analysis software are installed. The integrated control software is a program that measures and controls the entire equipment. It controls and measures the equipment and the smartphone, user interface function, real-time monitoring and control function, smartphone linkage and camera control function, and video data automatic monitoring and storage. function.

A cloud or NAS 550 stores the images.

15 shows a control and measurement device according to an embodiment of the present invention.

Components connected to the control and measurement device according to an embodiment of the present invention include a laptop computer, a battery and charger, a chassis, a module, a motor driver, an embedded computer, an access point, and a power device.

The chassis connects the control and measurement modules and measures the data.

The module measures the sensor signal and generates a control signal.

The motor driver delivers control commands to the motor and corrects errors when they occur.

The embedded computer receives the user's command, measures and measures the data of the equipment, and collects video control and video files from the smartphone and transmits them to the laptop computer.

The access point connects the components of the system to the network.

The power device is a device that supplies power to the control device and can drive AC220 or DC24.

The laptop computer has a built-in operating program, receives user information, and wirelessly measures and controls the system in real time.

Batteries and chargers are provided to use the equipment in places without wired power, and run for a maximum preset time (eg 5 hours).

16 shows data flow according to an embodiment of the present invention.

A dotted line indicates a wireless connection, and a solid line indicates a wired connection.

16, a black line shows a control signal, a blue line shows data flow through a wired LAN, a red line shows a power signal, and a purple line shows data flow through Wi-Fi.

The data acquisition device and the embedded computer are connected to the access point and wired LAN.

The power device supplies power to the motor driver, data acquisition device, access point, and embedded computer using AC 220V or battery power.

The motor driver sends control signals to the data acquisition device.

The access point is connected via Wi-Fi communication with smartphones and laptop computers.

Control commands are transmitted between the motor driver and the data acquisition device, and the data acquisition device and the embedded computer transmit and receive data through a wired LAN, and the embedded computer and the laptop computer through Wi-Fi communication.

Image data is transmitted and received through Wi-Fi communication between the smart phone and the embedded computer, and between the embedded computer and the laptop computer.

17 illustrates an image merging process according to an embodiment of the present invention.

According to an embodiment of the present invention, when merging image path information is received after merging starts, image merging (receiving image coordinate information by an analysis program), error checking (determination of result of merging completed image, providing editing function through user interface in case of error) ), preview (sequential merging images are compressed and saved to display images for preview), and user editing processes (image merging and post-processing are performed under user control if the user wants to edit when checking an error), and merging is performed. End (images are compressed and saved, and split images for the damage detection module are saved). That is, the image merging program provides a progress rate and a compressed image of a currently merged image to the user, and provides a handling function through a user interface when there is an error in merging.

Meanwhile, the damage detection method based on automatic image acquisition control according to an embodiment of the present invention may be implemented in a computer system or recorded on a recording medium. A computer system may include at least one processor, a memory, a user input device, a data communication bus, a user output device, and a storage. Each of the aforementioned components communicates data through a data communication bus.

The computer system may further include a network interface coupled to the network. The processor may be a central processing unit (CPU) or a semiconductor device that processes instructions stored in memory and/or storage.

The memory and storage may include various types of volatile or non-volatile storage media. For example, memory may include ROM and RAM.

Accordingly, the damage detection method based on the automatic control of image acquisition according to an embodiment of the present invention may be implemented as a computer-executable method. When the damage detection method based on automatic image acquisition control according to an embodiment of the present invention is executed in a computer device, computer readable instructions may perform the damage detection method based on automatic image acquisition control according to the present invention.

Meanwhile, the damage detection method based on automatic image acquisition control according to the present invention described above can be implemented as computer readable code on a computer readable recording medium. Computer-readable recording media includes all types of recording media in which data that can be decoded by a computer system is stored. For example, there may be read only memory (ROM), random access memory (RAM), a magnetic tape, a magnetic disk, a flash memory, an optical data storage device, and the like. In addition, the computer-readable recording medium may be distributed in computer systems connected through a computer communication network, and stored and executed as readable codes in a distributed manner.

Claims (10)

(a) acquiring an image based on automatic control;
(b) constructing damaged image data using similarly damaged objects; and
(c) detecting damage of at least one of concrete cracks, network cracks, exfoliation, damage, exposure of reinforcing bars, material separation, efflorescence, corrosion of steel materials, and peeling of paint using a deep learning model;
The step (b) is to collect the image of the similarly damaged object for each damage, and to construct synthetic learning data by synthesizing the damaged object with the image in which the similarly damaged object exists.
In-image acquisition automatic control-based damage detection method.

According to claim 1,
Step (a) is to obtain a high-resolution image capable of identifying a damage width of 0.1 mm or less from a preset distance in consideration of concrete durability and watertightness
In-image acquisition automatic control-based damage detection method.
According to claim 1,
Step (a) is to obtain photographing information including focusing information, photographing distance information, photographing angle information, and current location information.
In-image acquisition automatic control-based damage detection method.
According to claim 3,
In the step (a), image stitching is performed using the photographing distance information and photographing angle information, matching feature points in the image to be connected, and calculating distortion information by determining distortion between images using matching information, Inversely calculating the distortion information, rearranging the image pixels, filling the empty space between the images with the average value of neighboring images, correcting the matched image using the average value of brightness, and then cutting and storing the image to a predetermined pixel size.
In-image acquisition automatic control-based damage detection method.
delete According to claim 1,
In step (b), labeling is performed using the damage candidate detection result, and training data is constructed using feedback on the displayed damage candidate.
In-image acquisition automatic control-based damage detection method.
According to claim 1,
The step (c) is to perform damage detection by using a plurality of damage detection models in parallel using a segmentation-based artificial intelligence model.
In-image acquisition automatic control-based damage detection method.
According to claim 7,
Step (c) is to perform detection through grouping based on similarity for each damage.
In-image acquisition automatic control-based damage detection method.
According to claim 1,
In the step (c), linear damage and area-type damage are separately detected, a damaged pixel is determined for the linear damage, damage information is calculated in units of pixels, and a damage area for the area-type damage is calculated in units of pixels. to do
In-image acquisition automatic control-based damage detection method.
According to claim 9,
The step (c) quantifies the damage information in real units by converting the damage information calculated in units of pixels into actual units using at least one of shooting distance information and length information of an object identified in the image.
In-image acquisition automatic control-based damage detection method.

Images