CN115601331A

CN115601331A - High-rise building outer wall falling abnormity detection and risk assessment method based on infrared thermography

Info

Publication number: CN115601331A
Application number: CN202211284634.5A
Authority: CN
Inventors: 苏盛; 冯萧飞; 李彬; 王耀龙; 邹念; 李俊杰; 李想; 龙骧进; 王斌
Original assignee: Changsha University of Science and Technology
Current assignee: Changsha University of Science and Technology
Priority date: 2022-10-18
Filing date: 2022-10-18
Publication date: 2023-01-13

Abstract

A method for detecting and evaluating the falling abnormality of the outer wall of a high-rise building based on an infrared thermograph is characterized in that the characteristics that the heat dissipation characteristics of the outer wall of a beam column structure forming a bearing system and a filler wall forming a containment function in the high-rise building with a frame structure are obviously different are utilized, and the hollowing defect and the falling risk of the outer wall of the high-rise building are detected and evaluated by means of the infrared thermograph of the outer surface of the building. By analyzing the boundary invasion and the region communication condition of the boundary position between the building beam column and the filler wall region on the infrared thermograph, the characteristic difference between the abnormal region and the normal region on the infrared thermograph is extracted, the falling risk grade of the outer wall of the high-rise building is accurately evaluated, and the high-rise building with the prominent outer wall falling risk is screened out. According to the invention, through analysis and treatment of the infrared thermograph of the outer wall of the high-rise building, a large number of defects of the outer wall of the high-rise building falling off are screened and risk evaluation are realized, and decision support is provided for selective further deep detection and maintenance.

Description

High-rise building outer wall falling abnormity detection and risk assessment method based on infrared thermography

Technical Field

The invention belongs to the technical field of building safety abnormity detection of a frame structure, relates to building outer wall falling abnormity detection, and particularly relates to a high-rise building outer wall hollowing defect detection and falling risk assessment method caused by hollowing defects based on an infrared thermograph.

Background

The high-rise building is an important part of urban construction, improves the land utilization efficiency, brings higher population bearing capacity, and is the embodiment of economic and technical capacity. The high-rise building which is designed with unique originality and unique position often becomes a landmark building of a city, even a building with higher upper limit of floors is built, and the high-rise building is the target for the pursuit of each big city. However, the rise of high-rise buildings also brings great hidden dangers to urban safety, the outer wall heat insulation layer material adopted by the early high-rise buildings has great defects, and the building gradually ages and is damaged from the beginning of the building completion under the action of environmental stress such as sun, rain, heat expansion, freeze thawing and the like for a long time, and the building naturally falls off after reaching a certain degree. Over time, the problem of potential safety hazards in a large number of residences is also becoming more and more prominent. In recent years, the outer wall of a building drops off, and the brick, stone, tile and gravel falling from a high place is wrapped by huge potential energy and is smashed, so that pedestrians and vehicles in the way are often greatly damaged, news reporting the casualties of passersby or property loss caused by the outer wall bricks of the building falling from the high place exist in all parts of the country, and a precise and reliable method for quickly detecting the hidden danger of the dropping off of the outer wall of the urban high-rise building is urgently needed.

Generally, the quality detection of the outer wall of the building is to detect the sticking firmness degree of the outer wall, the finish coat, the heat insulation layer and the main body of the building. The traditional detection methods include an eye detection method, a hammering method, a drawing method, a shock wave flaw detection method and the like. Wherein, the visual inspection method and the hammering method are experience assessment methods, and technicians judge whether the risk of hollowing, cracking and falling off exists in the outer wall structure through visual observation and hollowing hammer knocking assistance on site; the drawing method is that a square or circular groove is opened on the outer side of the outer heat insulation system, epoxy resin adhesive is pasted on the surface of the outer heat insulation system, a drawing test is carried out after the metal disc is solidified, the stress of the metal disc is measured, and the health condition of the outer wall is judged; the method of the shock wave flaw detection is that by applying a small impact to the outer surface of a building, stress waves are caused at defective parts such as hollowing and cracking of an internal insulating layer, and the fluctuation characteristics such as a frequency peak value of the stress waves and a surface displacement frequency spectrum fed back are analyzed, so that whether defects exist in the structure or not can be judged, and the approximate size of the defects can be obtained. The detection method researches the correlation between the health degree of the building material and vibration, sound wave and appearance so as to judge and locate the defects, has a certain reference value, but has low detection efficiency, the accuracy and reliability depend on the subjective experience of operators, part of methods may cause further damage to the outer wall of the building, when the high-rise building is operated, a scaffold needs to be set up by a hammering method and a drawing method, and technicians often face extremely high safety risks.

In recent years, infrared thermal imaging technology is gradually applied to the field of building abnormity detection. Under the irradiation of sunlight, the health state of the building material is closely related to the temperature change, the infrared thermal imaging technology can present the temperature field distribution of the surface of the building outer wall on an infrared thermal image, the temperature field distribution is represented by different colors, and the long-distance, non-contact and large-range measurement characteristics of the infrared thermal imaging technology enable the building damage detection and analysis effect to be obvious by applying the method, and the nondestructive detection can be realized. At present, the position and the size of an abnormal area are usually judged by observing an abnormal high-temperature point in an image according to an infrared thermograph, the abnormal high-temperature point is commonly called as hot spot, and the temperature of the hollow area is higher than that of a surrounding normal area due to the fact that air exists in the hollow area, the heat transfer coefficient is changed, and heat is accumulated under the irradiation of sunlight. Or the judgment is carried out by observing abnormal low-temperature points in the image, commonly called cold spots, because the crack area absorbs water due to water seepage, the outer facing layer material absorbs water, the heat transfer coefficient is changed, the heat capacity is increased, and the temperature is slowly increased under the irradiation of sunlight, so that the temperature of the crack area is lower than that of the surrounding normal area.

The infrared imaging method detection process is usually judged by on-site shooting of a worker handheld device, the temperature distribution extraction function of the thermal infrared imager is relied on through the priori knowledge of the worker, the abnormal temperature position is searched spontaneously in a wall surface area, although certain effectiveness is achieved, large errors still exist, and the reliability of a detection result depends on the performance of the device and the technical level of operators. The method for distinguishing the abnormity through the temperature difference has a good detection effect only on small-amplitude abnormity points with obvious temperature difference. The abnormal information obtained by infrared imaging is subjected to detailed and comprehensive detection by the traditional detection method so as to evaluate the damage degree in detail, and the process is complicated, time-consuming and labor-consuming. Background noise is very serious in actual field environment, and windows, metal frames and air conditioner outdoor units can become abnormal points of temperature under the irradiation of sunlight, so that the identification precision is seriously disturbed. The urban dense population is often built in forest, the high-rise buildings are hundreds of thousands of, the damage aging speed and the damage aging degree of the buildings in different areas, different material structures and different use strengths are greatly different, and the infrared thermal image detection method relying on manual assistance has a certain detection speed, but still is a cup car pay relative to a huge building group. The digital driving abnormity automatic identification method based on the traditional infrared imaging abnormity detection is difficult to work. When the outer wall of the building falls off to cause personal injury or property damage, the problem exposed building is specifically analyzed and repaired to avoid the phenomenon of sheep casualty.

On the other hand, the temperature abnormality of the outer wall defect is usually submerged by various noises in the infrared thermograph of the whole outer wall, but for a high-rise building with a frame structure, the material difference between the beam column and the filler wall is influenced by sunlight, so that different heat dissipation characteristics are presented, and the accurate and efficient detection of the outer wall defect abnormality is expected to be carried out by focusing on the complete and clear temperature uniform area distribution rule on the surface of the outer wall of the building.

In a word, novel wisdom city should be can intelligent perception city building life health situation, and real time monitoring masters the building damage degree, in time warns management scheduling personnel, prevents to suffer from in the bud. The existing building defect detection method can not detect and analyze urban building groups in a large area, at a high speed and with high precision, and the excavation of urban elements and information is not deep enough, so that a digital detection method capable of quickly and effectively probing the damage degree of the outer wall of a high-rise building is urgently needed.

Disclosure of Invention

The invention aims to provide a high-rise building outer wall falling abnormality detection and risk assessment method based on an infrared thermograph, aiming at the defects of the prior art, and the method is used for extracting the boundary position abnormal change of a filler wall area in the building outer wall infrared thermograph according to the distribution rule characteristics of a building beam column area and the filler wall area of a frame structure by combining the infrared thermograph and a visible light image, so as to realize the building outer wall falling abnormality detection and falling risk assessment.

In order to achieve the purpose, the invention combines a thermal infrared imager and a visible light imaging device to shoot the outer wall of the urban high-rise building with a frame structure, carries out image segmentation based on a visible light image, locates the maximum effective area belonging to the outer wall surface of the building in the image, extracts the distribution characteristics of boundary pixels between the building beam column and the filling wall area in the corresponding outer wall infrared thermal image through an image processing algorithm, further judges whether the area has the abnormity of the characteristic empty drum defects such as boundary intrusion, area communication and the like, counts the total number of the abnormal areas, evaluates the falling risk grade of the outer surface of the building, generates a diagnosis report of the outer wall of the building, and further screens out urban problem building groups rapidly.

Specifically, the technical scheme adopted by the invention is as follows: a high-rise building outer wall falling abnormality detection and risk assessment method based on an infrared thermography comprises the following steps:

A. and establishing a geographical distribution map of the urban high-rise building, and collecting visible light images and corresponding infrared thermographs of the outer walls of the high-rise buildings.

The urban high-rise building is a high-rise building with a frame structure, the position distribution condition of the urban high-rise building is researched, the geographical distribution map of the high-rise building is established, the cruising ability of a line patrol vehicle and the operation and maintenance management cost are considered, an urban area is divided into a plurality of plates according to the area, the high-rise building concentrated area is taken as the center, the high-rise building scattered rare area is taken as the plate boundary, each plate is independently operated and maintained, and an operation and maintenance base point is arranged. And drawing an optimal patrol detection route by taking the large-width complete wall surface on the east and west sides of the building as a main target according to the geographical distribution map of the high-rise building. The line patrol vehicle is equipped with a FLIR M500 ultra high performance multi-sensor thermal imager system with 14 times continuous optical thermal zoom and 30 times electronic zoom color high definition camera, can detect thermal signals at the nearest 8.3nm, at the farthest 15.4km, can perform excellent short range and ultra long range target detection and identification, and its 360 ° azimuth and +/-90 ° pitch angle and automatically adjustable digital image detail enhancement can generate thermal images with clear texture and wide range. The automatic line patrol vehicle for the street periodically runs according to a fixed patrol track, shoots the outer wall of a high-rise building along the way, adjusts the direction and the elevation angle of the camera when reaching a preset fixed point position, obtains the outer wall image of the surrounding high-rise building, obtains the image which not only has rich color distribution information of the outer wall of the building, but also has the temperature distribution information of the surface layer of the outer wall, and simultaneously records basic attributes such as the longitude and latitude coordinates, the horizontal inclination angle, the vertical inclination angle, the orientation of the camera, the shooting distance, the orientation of a target wall surface and the like of the vehicle. The shooting time is arranged in sunny and cloudy days as much as possible so as to obtain the optimal temperature distribution image of the outer wall of the building as much as possible; the shooting angle is kept forward as much as possible, and the distance is moderate, so that the proportion of the building outer wall in the image is appropriate.

B. Segmenting the visible light image, and extracting the maximum effective area representing the exterior wall surface of the building in the image; and acquiring an infrared thermography corrected image (infrared corrected image for short) of the infrared thermography area corresponding to the maximum effective area.

In order to ensure timeliness of image acquisition and image processing, the image acquisition process is configured at the inspection vehicle terminal, and the image processing process and abnormal state evaluation are carried out at the cloud. After the high-rise outer wall image is collected, the obtained image is transmitted to a Faster-R-CNN neural network model trained by using the existing building outer wall data set, the visible light image is segmented, the area representing the building outer wall in the image is extracted, background noise interference of sky, street and the like in the image is eliminated, the processing area is reduced to the building outer wall, and the shot infrared thermograph is sorted and cut. The Faster-R-CNN neural network divides a plurality of candidate areas which accord with the characteristics of the outer wall surface of the building in the collected image according to the prior knowledge, and extracts the maximum effective area which represents the outer wall surface of the building. The maximum building wall candidate area obtained by image recognition and segmentation is calculated, the area of the maximum building wall candidate area can be considered as an effective target object only when the area of the maximum building wall candidate area occupies more than 40% of the area of the whole image, the condition that a plurality of small-amplitude ineffective building outer wall areas are recognized in the same image can be avoided, and then the effective infrared thermal image of the building outer wall area is obtained. By adopting a matlab data analysis platform, basic image processing operations such as Gaussian filtering, graying, expansion corrosion and the like are carried out on the image, image noise points are filtered, and the definition of the object appearance structure and the expression of the regional outline in the image can be enhanced.

The infrared correction image is obtained by extracting four straight lines belonging to the boundary of the outer wall of the building in the infrared thermal image region corresponding to the maximum effective region of the outer wall of the building by improved Hough transform, calculating coordinates of intersection points of the straight lines to obtain four corner points of the outer wall of the building, and then carrying out perspective transformation to eliminate the perspective distortion phenomenon in the image and obtain the front-view infrared thermography of the outer wall of the target building. Specifically, four corner point coordinates A, B, C and D of the external wall of the building are calculated through four side lines which are intersected pairwise on the external contour of the external wall of the building in the infrared thermography region by adopting improved Hough transformation and perspective transformation. And resetting the coordinates of the corner points of the corrected building outer wall as A ', B', C 'and D', substituting the coordinates of the four groups of mapping points into a perspective transformation equation set, solving the size of a parameter in a perspective transformation matrix, wherein the parameter is a perspective transformation parameter which accords with the image, and carrying out perspective transformation on all pixel coordinates in an area formed by four edges in the image based on the calculated perspective transformation matrix, so as to obtain the infrared corrected image of the building outer wall area.

C. And extracting boundary pixel distribution characteristics of building beam column and filled wall areas in the infrared correction image through an image processing algorithm. Specifically, the image processing algorithm comprises the following steps, and the implementation process of each step is conventional in the art:

(1) The infrared correction image is subjected to color enhancement by adopting R, G and B component piecewise linear adjustment, the image contrast is mainly enhanced, and the outlines and boundaries of a building beam column structure and a filler wall are enhanced, so that the colors of all temperature uniform areas are more vivid;

(2) Filtering the color-enhanced image by using a sobel operator, and calculating a color component change gradient image;

(3) Setting a minimum threshold value by adopting an extended minimum transformation method to partition and filter a low-gradient area and extracting boundary outlines of a beam column area and a filler wall area;

(4) And extracting a single-pixel-width boundary pixel coordinate matrix of the beam column and the filler wall area by adopting a marked watershed algorithm.

And B, in the infrared correction image of the building outer wall area obtained in the step B, the boundary between the building beam column and the filler wall area is clear and visible, but the sharpness between the areas with uniform temperature is insufficient, the transition area is wide and unclear, and the subsequent boundary coordinate extraction is not facilitated, so that the image enhancement is carried out by adopting the piecewise linear adjustment of the components R, G and B. In the infrared correction graph, three component values of R, G and B representing the temperature uniform region of the position of the building filling wall are higher, wherein R belongs to [240, 255], G belongs to [160, 250], B belongs to [120, 180], the region color is mainly white, the R component value of the reddish region of the position of the building beam column structure is higher and is consistent with the filling wall, the green and blue components are lower, R belongs to [240, 255], G belongs to [40, 120], B belongs to [50, 100], and the region color is mainly red. And in the process of transition from the white area at the position of the filler wall to the red area at the position of the beam column structure, the R component is approximately unchanged, and the G component and the B component show a remarkable increasing trend. Therefore, the areas are divided on the basis of 0-255, the R, G and B components in the filler wall area and the R component in the beam and column area are increased in a piecewise linear mode, the G and B components in the beam and column area are decreased in a piecewise linear mode, the colors of the two areas are enhanced, the boundary is sharpened, and the contrast between the areas is more vivid.

In order to extract the single-pixel width boundary of the building beam column and the filler wall area, the invention adopts a marked watershed algorithm to calculate the position of the area boundary, and adopts gradient amplitude transformation to preprocess the image. Because the infrared correction image is still an RGB three-channel image after being subjected to piecewise linear adjustment by adopting R, G and B components, and direct graying can lose a large amount of temperature distribution information, firstly, a sobel operator is adopted to filter the infrared image after color enhancement, then a color change gradient map of the infrared image is calculated, an extended minimum value transformation method is adopted to set the minimum value threshold value, a three-channel binary image can be obtained without damaging image information, morphological reconstruction is carried out by combining the gradient map and the binary image, if the pixel value in the binary image is 1, the gradient value of the corresponding position in the gradient map is set as the minimum value of an area, the outline of each temperature area boundary in the modified gradient map is clear and visible, the obtained area outline is composed of multiple pixels and is not beneficial to the next step of boundary accumulation, and a single-pixel width boundary of the outline of a beam column and a filling wall area is obtained by adopting a marked watershed algorithm.

D. According to pixel distribution of boundary positions of building beam columns and filled wall areas in the infrared correction image, a pixel stacking line graph is established, whether boundary invasion representing hollowing defects and area communication abnormity exist in the beam columns and the filled wall areas or not is detected, the total number of abnormal areas is counted, then the falling risk level of the outer surface of the building is evaluated, and a building outer wall health diagnosis report is generated, so that urban problem building groups can be screened out quickly.

According to the histogram theory in the image theory, the invention provides the boundary pixel accumulation theory for the first time. The histogram is an accurate graphic representation reflecting numerical data distribution, and a certain characteristic index of the data is used as a bottom edge; and counting the object frequency meeting the section of characteristic indexes, and taking the object frequency as a height to visually observe the distribution condition of the data. According to the method, the frequency of boundary pixels in each row and each column is counted aiming at the pixels positioned at the regional boundary position in the image, a pixel accumulation line graph is established, the physical significance of data distribution in the line graph is explored, and the positions of filled wall regions, beam column regions, and abnormal positions such as boundary intrusion and region communication can be positioned through related data characteristics.

And in the obtained infrared correction image of the building outer wall, establishing a longitudinal stacking broken line diagram of boundary pixels by taking the pixel row coordinate as a bottom edge and taking the lower boundary pixel frequency of the same row coordinate as a height. Extracting local accumulation maximum value points which are used for representing boundaries of building beam columns and filling wall areas in the original thermal image in the horizontal direction in the accumulation graph, defining that a longitudinal normal accumulation interval is formed between local accumulation maximum values at two ends in the vertical direction, and the accumulation of boundary pixels outside the interval can be judged to be abnormal, so that abnormal boundary pixels are obtained, adopting a mean-shift clustering algorithm according to the Euclidean distance between the pixels, taking the span of the normal accumulation interval as a clustering bandwidth, clustering abnormal boundary pixels adjacent in position into one class, and representing the number of abnormal areas invaded by the boundaries outside the longitudinal normal accumulation interval in a clustering result.

Counting the width of each zero accumulation interval in the longitudinal normal accumulation interval, comparing to obtain the maximum value of the width of the zero accumulation interval, taking the rest zero accumulation intervals as a reference, and considering the zero accumulation interval with the width lower than 60% of the maximum value as a boundary invasion abnormity; and calculating the shortest distance between each local accumulation maximum value except the local accumulation maximum values at the two ends of the normal accumulation section and the adjacent zero accumulation section, and if the distance exceeds the maximum width of the zero accumulation section, determining that the local accumulation maximum value is abnormal, namely that the corresponding area has internal area communication abnormality in the horizontal direction.

In the obtained boundary pixel coordinate matrix of the thermal image area of the outer wall, a boundary pixel transverse stacking line graph is established by taking pixel column coordinates as a bottom edge and taking the boundary pixel frequency number under the same column coordinates as a height. Local accumulation maximum value points which represent boundaries of building beams and columns and filled wall areas in the vertical direction in the original thermal image in the accumulation area are extracted, a transverse normal accumulation interval is defined between local accumulation maximum values on two sides in the horizontal direction, accumulation outside the interval can be judged to be abnormal, and therefore abnormal boundary pixels are obtained. According to the Euclidean distance between pixels, a mean-shift clustering algorithm is adopted, the minimum distance between all zero accumulation areas in a longitudinal accumulation graph is taken as a clustering bandwidth, abnormal boundary pixels adjacent in position are clustered into a class, and the number of the class clusters in a clustering result is the number of abnormal areas representing boundary intrusion outside a transverse normal accumulation interval.

Therefore, the total number of abnormal areas of the thermal image corrected on the outer wall of the building is the sum of the number of abnormal areas invaded by the boundary outside the vertical normal stacking section, the number of abnormal areas invaded by the boundary of the zero stacking section, the width of which is less than 60 percent of the width of the maximum zero stacking section, in the vertical normal stacking section, the number of the abnormal areas communicated with the inner area, the shortest distance between the local stacking maximum value and the adjacent zero stacking section exceeds the maximum value of the width of the zero stacking section, in the vertical normal stacking section, and the number of the abnormal areas invaded by the boundary outside the horizontal normal stacking section.

According to the position distribution characteristics of the horizontal direction boundary of the beam column and the temperature uniform area of the filler wall and the relation between the position distribution characteristics and the floors, in the obtained building outer wall longitudinal stacking graph, the local stacking maximum value points can be known to represent the positions of the upper boundary and the lower boundary of each floor, and therefore the number of the local stacking maximum value points is in a double relation with the number N of the floors in the corrected image. Therefore, the number of building layers in the current infrared correction image of the building outer wall can be extracted according to the information in the longitudinal stacking line graph, the falling risk rating of the outer wall caused by the empty drum defect can be carried out on the target wall surface by combining the total number of abnormal areas of the infrared correction image of the building outer wall obtained in the abnormality detection process, and the damage degree of the building outer wall in the empty drum abnormality aspect is judged, wherein the falling risk rating A of the building outer wall is defined as the ratio of the total number of the abnormal areas in the detected infrared correction image to the number of building layers. From the above abnormal region detection process, the maximum value of the number of effective abnormal regions existing in the infrared correction image should be the sum of the number of abnormal regions outside the longitudinal normal stacking region (the number of abnormal regions is 2 at most), the number of region intrusion and region communication abnormality in the longitudinal normal stacking region (the maximum effective number of abnormal regions is L-1, L is the number of floors), and the number of abnormal regions outside the transverse normal stacking region (the number of abnormal regions is 2L at most), i.e., 3L +1. Thus, the range of the exterior wall peel risk rating A is considered approximately between [0,3 ]. Further, defining A belonging to [0,0.5] as low falling risk, A belonging to [0.5,1] as medium falling risk and A belonging to [1,3] as high falling risk, and carrying out risk grade evaluation on the infrared thermography abnormal detection result of the exterior wall of the building.

Positioning the geographical position of the problem building according to attributes such as longitude and latitude coordinates, shooting angles and the like of the shot image, generating a building outer wall abnormity detection report through FLIR Tools software, presenting the building outer wall image and corresponding hollowing defect data according to requirements, and finally generating a high-rise building detection list of the outer wall hollowing defect abnormity. The management library of the whole life cycle of the urban high-rise building can be established, an annual inspection list, a high-risk list and a rush-repair list are set according to different building conditions, one-floor first-grade is realized, and data support is provided for further management, analysis, detection and repair. Timely discovering buildings with larger hidden dangers in the city, arranging overhaul in advance and the like without rain and silk.

The invention has the following beneficial effects:

1. and extracting abnormal changes of the boundary representation hollowing defects of the beam column region and the filler wall region in the infrared thermography of the outer wall of the building according to the structural rule characteristics of the high-rise building frame, and establishing a relation between the hollowing defects on the surface of the outer wall and the abnormal changes of the boundary of the temperature region in the thermography.

2. According to the histogram theory, a method for establishing a boundary pixel stacking line graph of a temperature uniform area is provided, the positions of a filler wall and a beam column area are positioned according to longitudinal and transverse distribution characteristics of boundary pixel stacking, and the abnormity of boundary intrusion and area communication representing hollowing defects existing in the filler wall and the beam column area is detected, so that the health condition of the outer wall surface of a building is effectively judged.

3. The method is characterized in that a method for rapidly primary screening of defects of the outer wall of the urban high-rise building is designed, an urban automatic line patrol vehicle is combined, a data driving method is adopted, abnormal conditions of the outer wall of the building are automatically identified, a building outer wall abnormal identification report is generated, a defective building list is established, the safety state of the outer facing layer of the high-rise building is dynamically grasped, the falling risk of the outer facing brick wall can be controlled, and the potential hazard of high-altitude falling objects is timely eliminated.

The invention will be further explained and explained with reference to the drawings.

Drawings

FIG. 1 is an infrared thermal image of an exterior wall of a building.

Fig. 2 is an infrared thermal image of a frame structure of an outer wall of a large-breadth building without foreign matters.

FIG. 3 is a diagram of a boundary model of a beam column area and a filler wall area of a defective building.

Fig. 4 is a partition view of the outer wall area of a building.

FIG. 5 is a vertical stacking line chart of the boundary pixels of the outer wall of the building with the marks.

FIG. 6 is a horizontal stacking line diagram of the pixels on the boundary of the outer wall of the building with marks.

Fig. 7 shows the clustering result of abnormal boundary pixels at two ends in the vertical direction of the building exterior wall, where the number of abnormal clusters is 2.

Fig. 8 shows the clustering result of the abnormal boundary pixels on both sides of the building outer wall in the horizontal direction, where the number of the abnormal clusters is 6.

Detailed Description

The invention utilizes the characteristic that the heat dissipation characteristics of the outer walls of a filler wall and a beam column in a high-rise building with a frame structure are obviously different, and carries out hollowing defect abnormity detection and falling risk assessment caused by hollowing defects on the outer wall of the high-rise building by means of the infrared thermograph of the outer wall of the building. The method comprises the steps of firstly analyzing the boundary invasion and the region communication condition of the boundary position between a building beam column and a filler wall region on an infrared thermograph, thereby extracting the characteristic difference between an abnormal region and a normal region on the infrared thermograph, accurately evaluating the grade of the falling risk of the outer wall of the high-rise building, and screening the high-rise building with the prominent outer wall falling risk. Specifically, the method for detecting the falling abnormality of the outer wall of the high-rise building and evaluating the risk based on the infrared thermography comprises the following steps:

The urban high-rise building refers to a high-rise building with a frame structure. The urban area is divided into a plurality of plates according to the area, each plate is independently operated and maintained, a patrol detection route is formulated in each plate, an automatic line patrol vehicle carrying visible light imaging equipment and infrared thermal imaging equipment periodically runs according to a fixed running track, and the exterior wall of a high-rise building along the way is shot.

B. Segmenting the visible light image, and extracting the maximum effective area representing the outer wall surface of the building in the image; and acquiring an infrared thermography corrected image (infrared corrected image) of the infrared thermography area corresponding to the maximum effective area.

In order to ensure timeliness of image acquisition and image processing, the images are acquired on the vehicle side, and the image processing process and abnormal state evaluation are carried out at the cloud. After the high-rise outer wall image is collected, the visible light image is segmented by adopting an Faster R-CNN target detection algorithm, the processing area is reduced to the outer wall surface of the building, the infrared thermograph representing the effective area of the outer wall surface of the building is identified, the infrared thermograph outer wall area extracted from the visible light image and corresponding to the effective area of the outer wall surface of the building, which accounts for more than 40% of the whole infrared thermograph in which the infrared thermograph outer wall area is located, is regarded as the effective area of the extracted outer wall surface of the building, and the maximum effective area representing the outer wall surface of the building is extracted accordingly. And carrying out basic image processing operations such as Gaussian filtering, graying, expansion corrosion and the like on the extracted effective region infrared thermograph by adopting a matlab data analysis platform, filtering image noise points, and then obtaining an infrared correction image of the building outer wall region by adopting improved Hough transform and perspective transform.

C. And extracting boundary pixel distribution characteristics of building beam column and filled wall areas in the infrared correction image through an image processing algorithm. The specific process is as follows: the infrared correction image is subjected to color enhancement by adopting R, G and B component piecewise linear adjustment, and the image contrast is mainly enhanced to strengthen the outline and the boundary of a building beam column structure and a filling wall, so that the color of each temperature uniform area is more vivid; filtering the color-enhanced image by using a sobel operator, and calculating a color component change gradient image; setting a minimum threshold value by adopting an extended minimum transformation method to partition and filter a low-gradient area and extracting boundary outlines of a beam column area and a filler wall area; and extracting a single-pixel-width boundary pixel coordinate matrix of the beam column and the filled wall area by adopting a marked watershed algorithm to obtain the boundary pixel distribution characteristics of the building beam column and the filled wall area.

D. Establishing a pixel stacking line graph according to pixels of boundary positions of building beam columns and filled wall areas in the infrared correction image, detecting whether the areas have abnormity such as boundary intrusion and area communication representing hollowing defects, counting the total number of the abnormal areas, evaluating the falling risk grade of the outer surface of the building, and generating a building outer wall health diagnosis report.

Fig. 1 is an infrared thermography of an external wall of a building with a certain frame structure, wherein abnormalities such as hot spots, boundary damage and the like exist in the image. The red hot spot indicated by the white arrow on the right side and the adjacent temperature uniform area have sharp temperature jump on the boundary, the traditional temperature anomaly detection method can detect the anomaly at the position, but the area indicated by the gray arrow on the left side has a remarkable sign that the temperature uniform area of the filling wall invades the beam column area, but the invaded part and the lower filling wall area belong to the same temperature uniform area, the traditional temperature anomaly detection method is divided by a threshold value, the anomaly at the position cannot be effectively detected by contour extraction, and the condition of false judgment missing exists. Because the traditional infrared detection method depends on local temperature abnormity and has defects, in the field of building outer wall flaw detection, the method can be combined with the distribution characteristics of a high-rise building frame structure, set constraints according with on-site natural laws, analyze the relation between surface layer damage and boundary damage in an infrared thermograph, and establish a model of boundary damage in a temperature-uniform area to effectively judge whether surface layer defects exist in the building outer wall.

In fig. 2, the black line marked area is a building filler wall area, which is built by common hollow bricks, has small specific heat capacity, is heated quickly under the irradiation of sunlight, has uniform temperature distribution, and is a white wide rectangular area on a thermography; the gray-white line marking area is a building beam column area and is formed by pouring steel bars with common concrete, the specific heat capacity is large, the temperature rise is slow under the irradiation of sunlight, the temperature distribution is uniform, and the gray-white line marking area is a red narrow frame area on an infrared thermograph (the color image is not submitted, and the gray color image is shown in the area in the figure 2).

In the infrared thermograph of the building outer wall with good health condition and no obvious abnormity, the boundary of the filler wall and the beam column region should be regular and orderly, and the temperature uniform region has no obvious boundary invasion damage or region communication phenomenon.

Between the building beam column structure and the filling wall, because of different materials and structures and different construction processes, the environmental stress resistance of the materials at the boundary is necessarily weaker than that of the materials at the boundary, natural defects such as cracking, hollowing and the like are generated firstly under the environmental stress erosion, and the observation of an infrared thermograph used in the past research can also find that temperature abnormal areas are often distributed beside the boundary of the two, few defects contained in one area exist, the boundary of the abnormal area is usually connected with the temperature areas of the two, and the surface abnormality starts from the beginning or is related to the temperature abnormal areas.

When the beam column structure side generates hollowing, a high-temperature area is formed and connected with the temperature area of the adjacent filler wall, and the high-temperature area invades to the low-temperature area. Therefore, if the phenomenon of fuzzy boundary and invasion exists between the building surface filling wall and the beam column in the infrared thermograph, the damage or hollowness of the internal structure is indicated. It can be judged that there is a high probability of the empty drum defect.

In view of the fact that the infrared thermograph can detect the hollowing defects on the surface of the building outer wall and the temperature change of the outer wall skin shedding caused by the hollowing defects, and the temperature difference exists between the abnormal area and the normal area, the regional boundary invasion and the regional communication are proposed, and therefore the abnormity of the hollowing defects of the building outer wall is represented.

Based on the structural characteristics of the building outer wall, a boundary plane model of a building outer surface filler wall and a beam column region is constructed, the inside of a communication region formed by the boundary is a filler wall region, the outside of the communication region is a beam column region, and four kinds of boundary abnormity, namely longitudinal boundary invasion at two ends in the vertical direction, transverse boundary invasion at two sides in the horizontal direction, internal boundary invasion and internal region communication exist on the communication region. In the model shown in fig. 3, two abnormal longitudinal boundary intrusions at two ends in the vertical direction, six abnormal transverse boundary intrusions at two sides in the horizontal direction, one abnormal internal boundary intruding position, and one abnormal internal region communicating position.

Since the abnormal condition of the outer wall in the building outer wall abnormality detection cannot be predicted, in order to detect the four kinds of boundary abnormalities and the number of abnormal areas thereof, the building outer surface image shown in fig. 3 is divided into three areas, as shown in fig. 4, (1) outside the longitudinal normal accumulation section in the vertical direction, (2) inside the longitudinal normal accumulation section, and (3) outside the transverse accumulation section in the horizontal direction. And according to the histogram theory and the boundary pixel distribution characteristics of the building beam column and the filler wall area, a method for establishing a temperature-uniform area boundary pixel stacking line graph is provided. According to the established pixel accumulation line graph, the positions of the filler wall and the beam column area are positioned according to the longitudinal and transverse distribution characteristics of boundary pixel accumulation, the boundary invasion and area communication abnormality in the filler wall and the beam column area are respectively detected, and then the falling risk grade of the building outer wall surface can be effectively judged. The building outer surface falling risk grade A in the infrared correction image is defined as the ratio of the total number of abnormal areas of the infrared correction image to the number of building layers contained in the image, the number of building layers of the infrared correction image is defined as 1/2 of the number of local accumulation maximum points of a longitudinal normal accumulation interval in a boundary pixel longitudinal accumulation line graph, and if the calculated number of layers is a decimal, the whole is taken upwards; and defining A is from 0,0.5 to low risk of falling, A is from 0.5,1 to medium risk of falling, and A is from 1,3 to high risk of falling.

In the infrared correction chart of the building outer wall, a longitudinal stacking broken line chart of boundary pixels is established by taking the coordinates of pixel rows as the bottom edge and the frequency of the boundary pixels below the same row of coordinates as the height; and establishing a boundary pixel transverse stacking line graph by taking the pixel column coordinate as a bottom edge and the boundary pixel frequency number under the same column coordinate as a height. The longitudinal stacking line graph and the transverse stacking line graph of the boundary pixels drawn according to the building exterior surface image shown in fig. 3 are respectively shown in fig. 5 and 6.

The boundary of the area is a boundary pixel dense area, so that the frequency of row and column boundary pixels where the boundary is located is high, and the local stacking maximum value is presented in the stacking line graph, while in the beam column area in the longitudinal stacking line graph, the boundary pixel on each row is 0, and the zero stacking interval is presented in the stacking line graph.

In fig. 5, the position of a zero stacking section within a vertical normal stacking section is detected, data points are marked with a symbol "o", and a zero stacking section list nos [ zero _1, zero \2 \8230; ], zero _ i = [ up _ i, low _ i ] is extracted. Each zero accumulation section zero _ i is stored in the list, up _ i represents the ith zero accumulation section upper limit, and low _ i represents the zero accumulation section lower limit. If the interval width is 1, the value is stored as a single value. The maximum value of the interval width of the zero-deposit interval is None _ max, and the average value of the interval width of the zero-deposit interval is None _ aver.

Extracting a vertical local piled-up maximum list Peaks _ V = [ peak _ V _1, peak _v _ _2 _8230; \8230; ], peak _ V _ i = (num _ i, loc _ i), num _ i represents the maximum size (i.e., pixel pile-up value), and loc _ i represents the longitudinal position (i.e., row number) of the maximum. And calculating the average value peak _ v _ aver of the local accumulation maximum values as a threshold value for segmentation, and filtering the local accumulation maximum values of which the accumulation values are lower than the peak _ v _ aver in the list. And taking the maximum value of the zero interval width None _ max as an interval tolerance, comparing each local accumulation maximum value peak _ V _ j reserved in Peaks _ V after the average value is divided with the local accumulation maximum value possibly existing in the interval [ loc _ j-None _ max, loc _ j + None _ max ], reserving a larger value, and filtering out the adjacent accumulation maximum value, wherein the list of the reserved local accumulation maximum values is marked as Peaks _ V _ bound. The local piled-up maxima retained after two screenings are indicated by the symbol "+". These peaks may characterize the lateral boundaries of each infill wall region.

Aiming at the external abnormality of the normal accumulation interval in the vertical direction of the filler wall, the top and bottom local accumulation maximum value positions loc _ max and loc _ min are used as the upper limit and the lower limit of the interval to obtain a longitudinal accumulation interval Veitical _ tile, whether abnormal accumulation exists outside the accumulation interval is detected and marked by a symbol'. The phenomenon that whether boundary invasion exists at the top and bottom filler wall area boundaries of the accumulation interval can be judged. And the returned abnormal accumulation interval can be used for positioning the row coordinates of the abnormal position, further performing column traversal on the areas where the rows are located, and searching whether boundary pixels exist and column coordinates corresponding to the pixels. Because the boundary pixel positions of the same abnormal area are close, the Euclidean distance between abnormal pixel points is taken as a characteristic, the stacking interval span of the transverse stacking line graph is set to be the size of the bandwidth K value, the mean-shift clustering algorithm is adopted, the abnormal pixels belonging to the same abnormal area can be clustered into one class, the clustering result is shown in figure 7, and the number of the abnormal invasion of the outer boundary of the longitudinal normal stacking area at two ends of the building outer wall in the vertical direction is num1=2.

For the internal abnormity of the stacking interval of the filler wall in the vertical direction, the beam column area is a zero stacking interval, and the two sides of the beam column area are usually the boundaries of the filler wall and are local stacking maximum values. It can be considered that if the width of the zero-pile interval is less than 0.6 × none _max, it indicates that there is a more serious boundary intrusion in the beam-column region. The larger the degree of intrusion, the narrower the zero pile-up interval. Three zero-pile intervals can be observed in fig. 5, wherein the width of the zero-pile interval at the row 32 is 1, which is much smaller than the other two zero-pile intervals, so as to obtain the number num2=1 of beam-column regions subjected to obvious invasion in the region, i.e. the abnormal number of invasion of the boundary inside the longitudinal normal pile interval.

The pixel distance dist _ m between the remaining local pile-up maximum values peak _ V _ m and the zero pile-up section zero _ n closest thereto is calculated regardless of the local pile-up maximum values located at the top and bottom in the vertical direction (located at loc _ max and loc _ min) in the peak _ V _ bound. At the normal position, the zero interval span is large, the local accumulation maximum values corresponding to the two sides are extremely close to the zero interval, and even if boundary invasion exists, the distance does not exceed the None _ max. If the filled wall area is communicated, the zero accumulation interval representing the beam-column area disappears and becomes a low accumulation area, and the distance between the accumulation peak and the nearest zero accumulation interval at the position is at least close to the width of one filled wall plus the width of one beam-column area. Therefore, with None _ max as a threshold, a local accumulation maximum value whose dist _ m is greater than the threshold can be determined as abnormal. And counting the number of the distance abnormal maximum values and dividing by 2 due to symmetry, and if the obtained peak number is an odd number, dividing by 2 and then rounding upwards to obtain the number of the internal region communication abnormal values. In fig. 5, it can be seen that the distance between the two local stacking maximum values located in the row 92 and the row 103 and the adjacent zero stacking section far exceeds the maximum zero section width, so as to determine that there is an intra-region communication anomaly, and the number of intra-region communication anomalies num3=1 in the vertical normal stacking section can be known from the above calculation method.

In fig. 6, the horizontal local piled-up maximum list Peaks _ H is extracted, [ peak _ H _1, peak _h _ _2 _8230; \8230; ], peak _ H _ i = (num _ i, loc _ i), num _ i represents the size of the maximum value, and loc _ i represents the position of the maximum value. Calculating the average value peak _ h _ aver of the transverse accumulation maximum values as a threshold value for segmentation, reserving the accumulation maximum values which are larger than the peak _ h _ aver in the list, and marking the accumulation maximum values with a symbol "+"; and taking the maximum value loc _ max of the transverse accumulation maximum value position and the minimum value loc _ min of the position in the list as the upper limit and the lower limit of the interval, positioning the Horizontal accumulation interval Horizontal _ tile of the building filler wall main body, reversely detecting the area outside the accumulation interval, and marking with a symbol' if the accumulation exists.

And the returned abnormal accumulation interval can be used for positioning the column numbers of the abnormal positions, further traversing the areas where the columns are positioned and searching whether boundary pixels and corresponding row numbers exist or not. Thereby accurately locating the abnormal position of the lateral invasion. Because the boundary pixel positions of the same abnormal area are close, the Euclidean distance of the pixels is taken as a characteristic, the minimum distance between zero accumulation intervals in the longitudinal accumulation line graph is the bandwidth K value, and the minimum distance can be gathered into one class by adopting a mean-shift clustering algorithm to obtain the external abnormal number of the transverse area of the building outer wall. As can be seen in fig. 6, outside of local maximum values accumulated on both sides of the accumulation section, significant pixel accumulation still exists, it can be known that there is significant lateral area boundary intrusion on both sides, the section where the column of the abnormal pixel is located is [4,8], [40,44] can be obtained by the lateral accumulation broken line diagram, so that row traversal is performed on the corresponding column in the image, the boundary pixel points existing therein are detected, a lateral abnormal boundary pixel point set is obtained, and then according to the clustering method, the abnormal pixels with close distances are clustered into one class, and the clustering result is shown in fig. 8, and it can be known that the number num4=6 of the external boundary intrusion abnormal number of the lateral normal accumulation area on both sides in the horizontal direction of the building exterior wall is obtained.

The number of filling wall layers L = N/2=5 in the obtained image can be determined by counting the total number of empty drum defect abnormalities num = num1+ num2+ num3+ num4=10, and counting the number of local accumulation maximum values N =10 in the peak _ V _ bound list obtained by vertically accumulating the broken line diagram. And further calculating the risk grade A = num/L =2 of the outer wall of the target building, which is a quite serious outer wall abnormal condition.

Claims

1. A high-rise building outer wall falling abnormality detection and risk assessment method based on an infrared thermograph is characterized by comprising the following steps:

A. establishing a geographical distribution map of the urban high-rise building, and collecting visible light images and corresponding infrared thermographs of the outer walls of the high-rise buildings;

B. segmenting the visible light image, and extracting the maximum effective area representing the outer wall surface of the building in the image; acquiring an infrared correction image of an infrared thermography area corresponding to the maximum effective area;

C. extracting boundary pixel distribution characteristics of building beam column and filler wall areas in the infrared correction image through an image processing algorithm;

D. according to pixel distribution of boundary positions of a filling wall area in an infrared correction image, a pixel stacking line graph is established, whether boundary invasion representing hollowing defects and area communication abnormity exist between a beam column and the filling wall area or not is detected, the total number of abnormal areas is counted, then the falling risk grade of the outer surface of a building corresponding to the infrared correction image is evaluated, and a health diagnosis report of the outer wall of the building is generated.

2. The infrared thermography-based method for detecting the falling abnormality and evaluating the risk of the outer wall of the high-rise building as claimed in claim 1, wherein the specific process of detecting the region boundary abnormality in the step D is as follows: according to pixel distribution of boundary positions of a filler wall and a beam column region in an infrared correction image, establishing a longitudinal stacking line diagram of boundary pixels by taking pixel row coordinates as a base and the frequency of the boundary pixels under the same row coordinates in the correction image as a height, and detecting the abnormality of boundary intrusion outside a longitudinal normal stacking interval; counting the width of each zero accumulation interval in the longitudinal normal accumulation intervals, calculating the maximum value of the width of the zero accumulation interval, and if the width of one zero accumulation interval is lower than 60% of the maximum value of the width of the zero accumulation interval, judging that the infrared image area corresponding to the position of the zero accumulation interval has boundary invasion abnormality; if the shortest distance between a certain local accumulation maximum value and an adjacent zero accumulation interval in the longitudinal normal accumulation interval exceeds the maximum width of the zero accumulation interval, judging that the internal area communication abnormality exists in the boundary of the upper area of the image corresponding to the local accumulation maximum value; and meanwhile, taking the pixel column coordinate as the base, correcting the lower boundary pixel frequency of the same column coordinate in the image as the height, establishing a boundary pixel transverse stacking line graph, and detecting the abnormity of boundary intrusion outside a transverse normal stacking interval.

3. The method for detecting the falling abnormality of the outer wall of the high-rise building and evaluating the risk based on the infrared thermograph as claimed in claim 2, wherein when the total number of the abnormal regions is counted, the number of the abnormal regions invaded by the boundary outside the longitudinal normal stacking section is obtained by clustering by using the span of the normal stacking section in the longitudinal stacking section as a clustering bandwidth according to the Euclidean distance between the abnormal pixels outside the normal stacking section detected in the longitudinal stacking broken line graph, the abnormal boundary pixels adjacent to the normal stacking section in the longitudinal stacking section are clustered into one class, and the number of the clusters in the clustering result is the number of the abnormal regions; the number of abnormal areas invaded by the boundary outside the transverse normal stacking interval is obtained by clustering by using a mean-shift clustering algorithm and using the minimum distance between the zero stacking areas in the longitudinal stacking diagram as a clustering bandwidth according to the Euclidean distance between abnormal pixels outside the normal stacking interval detected in the transverse stacking line diagram, the abnormal boundary pixels adjacent to the position are clustered into one type, and the number of clusters in a clustering result is the number of the abnormal areas; the total number of abnormal areas is the sum of the number of abnormal areas invaded by the boundary outside the longitudinal normal accumulation section, the number of abnormal areas invaded by the boundary of the zero accumulation section, the width of which is less than 60 percent of the width of the maximum zero accumulation section, in the longitudinal normal accumulation section, the number of internal area communication abnormal areas, the shortest distance between the local accumulation maximum value and the adjacent zero accumulation section exceeds the maximum value of the width of the zero accumulation section, in the longitudinal normal accumulation section, and the number of abnormal areas invaded by the boundary outside the transverse normal accumulation section.

4. The method for detecting and assessing the falling abnormality of the outer wall of the high-rise building based on the infrared thermography as claimed in claim 3, wherein the falling risk grade A of the outer surface of the building in the infrared corrected image is defined as the ratio of the total number of the abnormal areas of the infrared corrected image to the number of the building layers contained in the image, the number of the building layers of the infrared corrected image is defined as 1/2 of the number of the local accumulation maximum points in the longitudinal normal accumulation section in the longitudinal accumulation line graph of the boundary pixel, and if the calculated number of the layers is a decimal number, the number is rounded up; and defining A is from 0,0.5 to low risk of falling, A is from 0.5,1 to medium risk of falling, and A is from 1,3 to high risk of falling.

5. The infrared thermography-based high-rise building exterior wall falling abnormality detection and risk assessment method as claimed in claim 1, wherein the specific process of the step C is as follows:

(1) Performing color enhancement on the infrared correction image by adopting R, G and B component piecewise linear adjustment;

(2) Filtering the image after color enhancement by using a sobel operator, and calculating a color component change gradient image;

(3) Setting a minimum threshold value by adopting an extended minimum transformation method to partition and filter a low-gradient region and extracting boundary profiles of a beam column region and a filler wall region;

6. The infrared thermography-based method for detecting and assessing the exfoliation abnormality of the exterior wall of high-rise buildings as claimed in claim 1, wherein the effective area image area extracted in step B, which belongs to the exterior wall surface of the building, should occupy more than 40% of the whole visible light image area.

7. The method for detecting the falling abnormality of the outer wall of the high-rise building and evaluating the risk based on the infrared thermography as claimed in claim 1, wherein the infrared correction image in the step B is obtained by processing the extracted infrared thermography area corresponding to the maximum effective area of the outer wall surface of the building by improved Hough transform and perspective transform.

8. The method for detecting the falling abnormality and evaluating the risk of the outer wall of the high-rise building based on the infrared thermography as claimed in claim 1, wherein before the infrared correction image is obtained in the step B, an infrared thermography corresponding to an effective area of the outer wall surface of the building is filtered by using a matlab data analysis platform to remove image noise, so that the definition of the outline structure of the object and the expression of the area contour in the image are enhanced.

9. The infrared thermography-based high-rise building exterior wall drop abnormality detection and risk assessment method as claimed in claim 1, wherein in the step a, the exterior wall image is collected by dividing an urban area into a plurality of blocks according to the area size, and an automatic line patrol vehicle carrying a dual-photothermal imaging device in each block periodically runs according to a fixed running track so as to shoot the exterior wall of the high-rise building along the way.

10. The infrared thermography-based method for detecting the falling abnormality of the outer wall of the high-rise building and evaluating the risk as claimed in claim 1, wherein in the step B, a Faster R-CNN target detection algorithm is adopted to segment the visible light image.