CN115620169B - Building main angle correction method based on regional consistency - Google Patents
Building main angle correction method based on regional consistency Download PDFInfo
- Publication number
- CN115620169B CN115620169B CN202211609894.5A CN202211609894A CN115620169B CN 115620169 B CN115620169 B CN 115620169B CN 202211609894 A CN202211609894 A CN 202211609894A CN 115620169 B CN115620169 B CN 115620169B
- Authority
- CN
- China
- Prior art keywords
- angle
- main
- building
- building structure
- vector
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 71
- 238000012937 correction Methods 0.000 title claims abstract description 11
- 238000000605 extraction Methods 0.000 claims abstract description 39
- 238000012805 post-processing Methods 0.000 claims abstract description 19
- 238000012545 processing Methods 0.000 claims abstract description 10
- 230000011218 segmentation Effects 0.000 claims description 10
- 238000005192 partition Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 3
- 238000000638 solvent extraction Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 10
- 230000001788 irregular Effects 0.000 abstract description 10
- 238000004364 calculation method Methods 0.000 abstract description 5
- 238000005457 optimization Methods 0.000 abstract description 5
- 238000009435 building construction Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000003595 spectral effect Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000008961 swelling Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/13—Satellite images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/761—Proximity, similarity or dissimilarity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Remote Sensing (AREA)
- Astronomy & Astrophysics (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a building main angle correction method based on regional consistency, and relates to the field of remote sensing image processing. The method is used as a post-processing method for building extraction, the main angle of the irregular building structure is corrected according to the trend consistency of the building structure in the area range of the irregular building structure, the accuracy of main angle calculation in post-processing is improved, and therefore the post-processing optimization effect of building structure extraction can be greatly improved.
Description
Technical Field
The invention relates to the field of remote sensing image processing, in particular to a building main angle correction method based on regional consistency.
Background
At present, with the arrival of the big data era of smart cities and the continuous development of satellite remote sensing technology, earth observation, satellite remote sensing, ecological assessment and homeland supervision are gradually developed towards the macroscopic direction, the dynamic direction and the fine direction, so that the large-range and high-efficiency satellite remote sensing image interpretation has practical requirements. The building extraction in the high-resolution remote sensing image has important significance for the applications of illegal building monitoring, urban area automatic extraction, map updating, urban change monitoring, urban planning, three-dimensional modeling, digital urban establishment and the like. The high-resolution remote sensing image improves the spectral characteristics of the ground features, highlights the information of the structure, texture, details and the like of the ground features, and simultaneously, due to the fact that the ground features are shielded due to the problem of the observation angle of the satellite, the scale is increased to bring about the serious problem of foreign matter co-spectrum, and meanwhile, the noise of the image is increased, so that the precision of building extraction is limited, the visual interpretation method is still the most commonly used interpretation method, although the precision of the visual interpretation method is relatively guaranteed, the defects of low efficiency, time and labor consumption severely restrict the large-scale application of the high-resolution remote sensing image, and the image data are greatly wasted.
The deep learning is a new stage of machine learning development in artificial intelligence, and effectively solves the problems of characterization of complex object features, correlation analysis of complex scenes and the like. The deep learning method for high-resolution remote sensing image building extraction can automatically extract the feature information of a building, and high-precision and high-efficiency building extraction is realized. However, due to the complexity of the remote sensing image, the buildings are affected by noise, occlusion, shadow and low contrast, the current automatic building extraction method has low reliability of results, and has obvious defects that the buildings and roads cannot be effectively distinguished because of similar spectral characteristics, and the extraction results with higher precision cannot be directly obtained by a full-automatic method. In the prior art, the other mode is to combine a computer automatic extraction technology with manual interaction, namely interactive surface feature extraction, and the extraction precision is guaranteed to a certain extent. However, the method needs to manually give the initial position of the building, depends on the edge information to extract the building, is complex in interaction, has high requirement on the precision of the manually given position, and is difficult to implement.
In order to further improve the precision of building automatic extraction and apply the extraction result to engineering practice, researchers have conducted various researches in the post-processing direction of building extraction. And performing graphical correction processing on the building extraction result based on the geometric shape, spatial distribution and other regular features of the building, wherein the optimized result can be directly used for a specific engineering project.
In the post-processing method, the accuracy of extracting the building by the deep learning method can be improved by calculating the obtained main angle of the building structure. However, the existing main angle calculation method still has two problems which are difficult to overcome, one is that the main angle of the building is difficult to obtain due to various forms of the building and errors caused by automatic extraction; and secondly, different main angle solving methods have certain application range and limitation, and particularly for buildings with poor extraction effect, the solving results of different solving methods are greatly different, so that better post-processing effect is difficult to obtain.
Disclosure of Invention
The invention provides a building main angle correction method based on regional consistency. According to the method, the main angle of the irregular building structure is corrected according to the trend consistency of the building structure in the area range of the irregular building structure, the accuracy of calculation of the main angle in post-processing is improved, and therefore the post-processing optimization effect of extraction of the building structure can be greatly improved.
In order to achieve the technical purpose, the technical scheme of the invention is as follows:
a building main angle correction method based on regional consistency comprises the following steps:
s1, obtaining a target remote sensing image, and obtaining a candidate main angle of a house building structure of the target remote sensing image through a preset method, wherein the candidate main angle comprises a first candidate value and a second candidate value;
s2, calculating a difference value between the first candidate value and the second candidate value, dividing the building structure into a standard building structure and a building structure to be corrected according to the relation between the difference value and a first preset threshold value, and acquiring a real main angle of the standard building structure;
s3, obtaining a region main angle of the building structure of the house to be repaired through a region statistical strategy, wherein the region statistical strategy comprises external expansion main angle statistics and line segment histogram statistics;
and S4, calculating a difference value between the candidate main angle and the area main angle, and screening out the real main angle of the building structure to be corrected from the candidate main angles according to the relation between the difference value and a second preset threshold value.
Further, step S1 includes:
s11, acquiring angles of two long axis center lines of a circumscribed rectangle with the minimum area of the building structure as a first candidate value of the building structure;
s12, obtaining a vector contour line segment of the building structure through a vector contour extraction method, and obtaining an angle of the vector contour line segment;
s13, grouping the vector contour line segments with similar angles into a group, screening out a group of vector contour line segments with the longest total length as a reference vector contour line segment group, and obtaining a second candidate value of the building structure through a vertical equivalence method based on the angles of the reference vector contour line segments;
the vector contour line segments of similar angles include vector contour line segments having an angle difference of [0 °,2 ° ] U [88 °,90 ° ].
Further, the vector contour extraction method includes:
performing semantic segmentation, binarization processing and mathematical morphology processing on the remote sensing image to obtain a post-processing image;
and carrying out boundary tracking and vector extraction on the post-processed image to obtain a vector contour line segment of the building structure.
Further, the vertical equivalence method comprises:
clustering the reference vector contour line segment group based on the angle of the reference vector contour line segment to obtain a main vector contour line segment group and an auxiliary vector contour line segment group, wherein the total line segment length of the main vector contour line segment group is greater than the total line segment length of the auxiliary vector contour line segment group, and respectively averaging the angle of the main vector contour line segment group and the angle of the auxiliary vector contour line segment group to obtain a main vector angle and an auxiliary vector angle;
and performing vertical transformation on the auxiliary vector angle to the main vector angle to obtain a vertical auxiliary vector angle, and taking the mean value of the main vector angle and the vertical auxiliary vector angle as a second candidate value.
Further, step S2 includes:
if the difference value between the first candidate value and the second candidate value is smaller than a first preset threshold value, taking the building structure as a standard building structure, and taking the second candidate value as a real main angle of the standard building structure;
and if the difference value of the first candidate value and the second candidate value is larger than a first preset threshold value, taking the building structure as the building structure to be corrected.
Further, the region statistical strategy is an outward expansion main angle statistics, and step S3 includes:
the minimum external rectangle of the building structure to be corrected is extended by a preset distance to obtain an extended rectangle;
taking the standard building construction inside the outward-expanding rectangle and intersected with the outward-expanding rectangle as a reference building construction;
and taking the mean value of the real main angles of the reference house building structure as the main angles of the area of the house building structure to be corrected.
Further, the region statistical strategy is a line segment histogram statistics, and step S3 includes:
partitioning the housing construction structure in the target remote sensing image by a hierarchical clustering method based on a distance matrix;
obtaining angles of vector contour line segments of the building structures in each partition, wherein the building structures are composed of a plurality of vector contour line segments;
respectively constructing angle histograms of all the subareas by utilizing angles of vector contour line segments of the house building structures in all the subareas;
and taking the angle corresponding to the highest frequency in the angle histogram of the partition where the house building structure to be corrected is located as the main angle of the area of the house building structure to be corrected.
Further, step S4 includes:
respectively calculating the difference between the first candidate value and the main angle of the area and the difference between the second candidate value and the main angle of the area, and selecting a smaller difference as a target difference;
if the target difference value is smaller than a second preset threshold value, selecting a candidate main angle close to the main angle of the area as a real main angle of the building structure of the house to be corrected;
and if the target difference is larger than a second preset threshold value, selecting a second candidate value as a real main angle of the building structure to be corrected.
The invention has the beneficial effects that: the invention provides a brand-new building main angle correction method based on regional consistency. And calculating to obtain the regional main angle of the building structure to be corrected through a regional statistical strategy, and selecting the optimal value of the candidate main angle by taking the regional main angle as a reference to obtain the real main angle. The method is used as a post-processing method for building extraction, solves the problem of difficulty in calculating the main angle of an irregular building, and can greatly improve the post-processing optimization effect of building extraction in the remote sensing image.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of a method for correcting a main angle of a building based on regional consistency according to the present invention;
FIG. 2 is a schematic diagram of an original distribution of a building according to a first embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a result of correcting a main angle of a building according to a first embodiment of the present invention;
fig. 4 is a schematic diagram of the original distribution of the building according to the third embodiment of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived from the embodiments of the present invention by a person skilled in the art, are within the scope of the present invention.
Referring to fig. 1, fig. 1 is a schematic flow chart of an embodiment of a method for correcting a main angle of a building based on regional consistency according to the present invention, the method includes the following steps:
s1, obtaining candidate main angles of a building structure by a preset method, wherein the candidate main angles comprise a first candidate value and a second candidate value;
s2, dividing the building construction into a standard building construction and a building construction to be corrected according to the difference value of the first candidate value and the second candidate value;
s3, acquiring the real main angle of the standard building structure in the preset range, and calculating the regional main angle of the building structure to be corrected according to the real main angle of the standard building structure in the preset range;
s4, selecting a numerical value closest to the main angle of the area of the building to be corrected from the candidate main angles of the building to be corrected, and obtaining the real main angle of the building to be corrected.
The technical idea of the invention is as follows: 1. the existing building extraction method based on remote sensing images can only extract regular buildings with obvious characteristics, the universality is poor, the extraction effect is not ideal under the condition that the buildings are dense or irregular buildings are encountered, and the accuracy of the extracted buildings is optimized by appropriate post-processing steps. Wherein, the main angle of the building is acquired, which greatly helps the post-processing optimization. The method for acquiring the main angle of the building generally comprises the following steps: 1) Acquiring a minimum external rectangle of a building, wherein the directions of central lines of two long axes of the minimum external rectangle are used as first candidate values of main angles of the building; 2) And acquiring contour line segments of the building, calculating the total length of the contour line segments with all similar angles, and taking the direction with the longest total length as a second candidate value of the main angle of the building. For irregular buildings, the results obtained by different methods for calculating the main angle may also be different, and how to further determine the most true main angle is emphasized in the subsequent steps of the invention; 2. the present invention introduces the concept of a principal angle of area to help determine the true principal angle of a building. The regional principal angle is an average of principal angles of a set of building structures having similar principal angles over a region in which the building is located. The main angle of the area can reflect the area consistency of the building area, and provides a guiding function for the main angle of the building to be corrected, which is consistent with the area consistency; 3. and selecting the candidate main angle of the building to be corrected, which is closest to the main angle of the area, as the real main angle. The method can determine the real main angle of the irregular building to be corrected by taking the regional consistency of the building area as guidance, and further improve the effect of post-processing of building extraction, thereby realizing high-precision building extraction.
Example one
The embodiment provides a method for obtaining an area principal angle of a building structure to be corrected based on outward expansion principal angle statistics, and meanwhile, the embodiment provides a method for obtaining a real principal angle of the building structure to be corrected under the condition that a target difference value is smaller than a second preset threshold value:
s1, obtaining a target remote sensing image, and obtaining candidate main angles of a building structure of the target remote sensing image by a preset method, wherein the candidate main angles comprise a first candidate value and a second candidate value;
s11, acquiring angles of two long axis center lines of a circumscribed rectangle with the minimum area of the building structure as a first candidate value of the building structure;
s12, obtaining a vector contour line segment of the house building through a vector contour extraction method, and obtaining an angle of the vector contour line segment;
s13, grouping the vector contour line segments with similar angles into a group, screening out a group of vector contour line segments with the longest total length as a reference vector contour line segment group, and obtaining a second candidate value of the building structure through a vertical equivalence method based on the angles of the reference vector contour line segments;
the vector contour line segment with the similar angle is a vector contour line segment with the angle difference of [0 degrees, 2 degrees ] < U [88 degrees, 90 degrees ].
To supplement the explanation of step S12, the vector contour extraction method includes:
performing semantic segmentation on the target remote sensing image by a deep learning method to obtain segmentation data;
in the embodiment, semantic segmentation is carried out on the target remote sensing image by using mask-RCNN. The Mask-RCNN follows the idea of fast RCNN, the structure of ResNet-FPN is adopted for feature extraction, a Mask prediction branch is additionally added, a semantic segmentation task with high quality can be completed, the extraction effect on common buildings is good, and the effect is obviously reduced only when the complex irregular buildings are faced. In this example, sample preparation was performed by manual sample drawing using ArcGIS. In the training process, pixel blocks cut to 512 × 512 are used as samples, about 7000 instances are used as a training data set, and about 3000 instances are used as a verification data set. Computer configuration: intel Xeon (R) CPU E5-2620 v4 @ 2.10GHz x 32; a display card: quadro M4000; memory: 128G; operating the system: ubuntu 16.04. The training process takes more than 40 hours and iterates for 9 ten thousand times;
carrying out binarization processing on the segmentation data, and taking 0.5 as a binarization threshold value of a building target probability map to obtain a segmentation binarization image;
the segmentation binary image of the building structure is used for highlighting the outline of a target building and compressing the data volume of the whole image;
post-processing the segmented binary image by a mathematical morphology method to obtain a post-processed image:
the mathematical morphology method comprises a group of morphological algebraic operators, and the basic operation comprises the following steps: swelling, erosion, opening and closing;
the segmentation binarization image is denoised by a mathematical morphology method, classification noise, small non-building structures and other useless information are removed, and the analysis and processing of the image shape and structure can be carried out based on the mathematical morphology method, wherein the analysis and processing comprises image segmentation, feature extraction, edge detection, image filtering, image enhancement and recovery.
Generating a contour line segment of the building structure by tracking the boundary of the post-processing image through a Moore field boundary tracking method;
and performing vector extraction on the post-processed image based on the contour line segment of the building structure to obtain a vector contour line segment of the building structure.
Supplementary explanation is made to step S12, and the vertical equivalence method includes:
clustering the reference vector contour line segment group based on the angle of the reference vector contour line segment to obtain a main vector contour line segment group and an auxiliary vector contour line segment group;
the total length of the line segments of the primary vector contour line segment group is greater than that of the line segments of the secondary vector contour line segment group;
respectively solving an angle mean value of the main vector contour line segment group and the auxiliary vector contour line segment group to obtain a main vector angle and an auxiliary vector angle;
the inter-group line segment angle difference between the main vector outline line segment group and the auxiliary vector outline line segment group is [0 degrees and 2 degrees ], and the inter-group line segment angle difference is [88 degrees and 90 degrees ], so that the main vector angle and the auxiliary vector angle are close to a vertical relation;
vertically transforming the auxiliary vector angle to the main vector angle to obtain a vertical auxiliary vector angle;
the vertical transformation is represented as follows:
wherein x is i Is the angle of the dominant vector, y i As a subsidiary vector angle, f (x) i ) Is a vertical auxiliary vector angle;
after the vertical change operation, the angle difference between the angle of the obtained vertical auxiliary vector and the angle of the main vector is [0 degrees, 2 degrees ];
taking the mean value of the main vector angle and the vertical auxiliary vector angle as a second candidate value b 2 。
And S2, calculating a difference value between the first candidate value and the second candidate value, and dividing the building into a standard building and a building to be corrected according to the relation between the difference value and a first preset threshold value.
If the difference value between the first candidate value and the second candidate value is smaller than a first preset threshold value, taking the building structure as a standard building structure, and taking the second candidate value as a real main angle of the standard building structure;
and if the difference value of the first candidate value and the second candidate value is larger than a first preset threshold value, taking the building structure as the building structure to be corrected.
Supplementary explanation is performed on step S2, in this embodiment, the preset threshold is set to 3 °, and if the difference between the first candidate value and the second candidate value is smaller than 3 °, it is determined that the main direction of the building structure does not need to be modified, the building structure is used as a standard building structure, and the second candidate value is used as a true main angle of the standard building structure.
And S3, acquiring the real main angle of the standard house building structure within the preset range, and acquiring the main angle of the area of the house building structure to be corrected according to the shape information of the standard house building structure within the preset range.
In a first embodiment, the region statistical strategy is an outward expansion main angle statistics, and step S3 includes:
obtaining a true principal angle a of a reference housing construction 2 、a 3 、a 4 、a 5 、a 6 、a 7 ；
Referring to fig. 2, a is a building structure to be modified, other buildings within a certain distance d from the periphery are reference building structures, and the buildings participating in the calculation of the main angle of the area include all the buildings within the distance d from a and intersecting the range line.
Clustering the real main angles of the reference building structure by a neighbor clustering method based on the real main angles:
(1) Taking any one of the real main angles of the 6 reference building structures as a first clustering center, for example, let z 1 =a 2 Wherein z is 1 Is reference group A 1 The cluster center of (a);
(2) Current cluster center is z 1 Calculating a 3 To the clustering center z 1 Euclidean distance dist (a) 3 ,z 1 )；
If dist (a) 3 ,z 1 )∈[0,3]Or U [87,93 ]]Then a is 2 ∈A 1 ；
(3) Current cluster center is z 1 Calculating a 5 To the clustering center z 1 Euclidean distance dist (a) 4 ,z 1 )；
If dist (a) 4 ,z 1 ) E (3, 87), then a 4 Defined as the second cluster center z 2 ，z 2 =a 4 Wherein z is 2 Is reference group A 2 The cluster center of (a);
(1) And analogizing until the clustering of the real main angles of the 8 reference house building structures is completed, wherein the clustering result is A 1 ={a 2 、a 3 、a 5 、a 6 、a 7 }}，A 2 ={a 4 }。
The nearest neighbor clustering method classifies according to Euclidean distance of a real main angle, and the Euclidean distance is expressed as follows:
wherein, dist (a) i ,a j ) As a true principal angle a i To the true principal angle a j The Euclidean distance of (c);
selecting a group of reference groups with the largest number of reference house building structures in the group, and calculating the mean value of the real main angles of the reference house building structures in the group to obtain the main angles of the area of the house building structures to be corrected;
wherein the reference group with the largest number of reference building structures in the group is A 1 Calculating the mean value of the main angles of the group of building structures to obtain the main angles of the area of the building structures to be corrected
。
S4, selecting a numerical value closest to the main angle of the area of the building to be corrected from the candidate main angles of the building to be corrected, and obtaining the real main angle of the building to be corrected.
In the first embodiment, the target difference is smaller than the second preset threshold, and step S4 includes:
respectively calculating the difference between the first candidate value and the main angle of the area and the difference between the second candidate value and the main angle of the area, and selecting a smaller difference as a target difference;
and selecting the candidate main angle corresponding to the target difference value as the real main angle of the house building structure to be corrected when the target difference value is smaller than a second preset threshold value.
The acquisition of the region principal angle in this embodiment is shown in fig. 2. In this embodiment, the target difference is smaller than the second preset threshold, which is visually represented that the candidate principal angle of the building structure to be modified is consistent with the regional principal angle thereof, that is, a better principal angle modification result can be obtained by the method of the present invention through the regional principal angle.
And then selecting the numerical value of the main angle of the area closest to the main angle of the building to be corrected from the 2 candidate main angles of the main angle of the building to be corrected to obtain the real main angle of the building to be corrected.
The result of the main angle correction in the embodiment is shown in fig. 3.
Example two
The embodiment provides a method for acquiring the main angle of an area of a building structure to be repaired based on line segment histogram statistics
The method of (1).
In the second embodiment, the region statistical strategy is a line histogram statistics, and step S3 includes:
partitioning the standard building structure in a preset range by a hierarchical clustering method based on a distance matrix:
the values of the elements in the distance matrix are obtained by weighting a space distance matrix and an angular distance matrix between buildings, and the formula is as follows:
wherein,d i,j for the element values in the distance matrix corresponding to the ith building and the jth building,
is the spatial distance between the ith building and the jth building, is based on>
For the angular distance between the ith building and the jth building>
Is a weight value of the spatial distance>
Is a weight value of the angular distance>
。
The shortest distance between the corner points between the boundary outlines of the two buildings is taken as the space distance between the two buildings, and the formula is as follows:
wherein,x p,i 、y p,i representing X and Y coordinates of a pth angular point in an ith building contour curve;x q,j 、y q,j the X and Y coordinates representing the q-th corner point in the jth building contour curve,N i the number of segments representing the ith building contour curve,N j the number of segments representing the jth building contour curve;
the difference of the angle mean values between the boundary outlines of the two buildings is used as the angular distance of the two buildings, and the formula is as follows:
in the formula,θ k,i representing the angle of the kth line segment in the outer contour curve of the ith building;θ l,j the outer contour curve representing the jth buildinglThe angle of the bar segment;S i denotes the firstiIndividual constructionThe number of segments of the building is,S j denotes the firstjThe number of segments of the individual building outline curves.
Obtaining an angle of a vector outline line segment of a reference building structure, wherein the standard building structure consists of a plurality of vector outline line segments;
and respectively constructing angle histograms of all the partitions by utilizing angles of vector contour line segments of the reference house building structures in all the partitions, wherein the angle corresponding to the highest frequency in each partition histogram is used as the main regional angle of the corresponding building.
Example three:
the embodiment provides a method for obtaining a real main angle of a building structure to be corrected under the condition that a target difference value is greater than a second preset threshold value, comprising the following steps:
referring to fig. 4, in the third embodiment, the target difference is greater than the second predetermined threshold:
and S3, obtaining the main angle of the area of the building structure of the house to be repaired through the statistics of the outward expansion main angle.
A is the house to be corrected, and the real main angle of other houses within a certain distance d of the periphery is a 1 、a 2 、…、a 8 . The houses participating in the calculation of the regional principal angle include all houses within the range of distance d and intersecting the range line. Counting the main angles of the building structures within the range of the building structures to be corrected, grouping the building structures with similar main angles into a group, and obtaining 4 groups of results: a. The 1 ={a 1 、a 2 、a 3 、a 4 、a 7 }，A 2 ={a 5 }，A 3 ={a 6 }，A 4 ={a 8 Selecting a group of building structures with the largest quantity as A 1 。
Calculating the average value of the main angles of the set of building structures to obtain the main angles of the area of the building structures to be corrected
。
And S4, calculating a difference value between the candidate main angle and the area main angle, and screening out the real main angle of the building structure to be corrected from the candidate main angles according to the relation between the difference value and a second preset threshold value.
Respectively calculating the difference between the first candidate value and the main angle of the area and the difference between the second candidate value and the main angle of the area, and selecting a smaller difference as a target difference;
and selecting the second candidate value as the real main angle of the building structure to be corrected when the target difference is larger than a second preset threshold value.
As can be seen from FIG. 4, the principal angle candidate b for House A 1 And b 2 The difference between the main angle a and the main angle a is large, so the main angle of the area of the building structure to be corrected cannot be used as a guide parameter for correcting the main angle of the building structure to be corrected. So that the second candidate value b of the main angle of the building structure of the house to be corrected is finally used 2 As the final true principal angle.
The invention has the beneficial effects that: the invention provides a brand-new building main angle correction method based on regional consistency. Calculating to obtain the main angle of the area of the to-be-corrected building structure through the main angles of the standard building structures in a certain range around the to-be-corrected building structure, and selecting the optimal value in the candidate main angles by taking the main angle of the area as the reference to obtain the real main angle. The method is used as a post-processing method for building extraction, solves the problem of difficulty in calculating the main angle of an irregular building, and can greatly improve the post-processing optimization effect of building extraction in the remote sensing image.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a component of' 8230; \8230;" does not exclude the presence of another like element in a process, method, article, or apparatus that comprises the element.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (5)
1. The method for correcting the main angle of the building based on the area consistency is characterized by comprising the following steps of:
s1, obtaining a target remote sensing image, and obtaining a candidate main angle of a house building structure of the target remote sensing image through a preset method, wherein the candidate main angle comprises a first candidate value and a second candidate value;
the step S1 comprises the following steps:
s11, acquiring angles of two long axis center lines of a circumscribed rectangle with the minimum area of the building structure as a first candidate value of the building structure;
s12, obtaining a vector contour line segment of the house building through a vector contour extraction method, and obtaining an angle of the vector contour line segment;
s13, grouping the vector contour line segments with similar angles into a group, screening out a group of vector contour line segments with the longest total length as a reference vector contour line segment group, and obtaining a second candidate value of the building structure through a vertical equivalence method based on the angles of the reference vector contour line segments;
the vector contour line segments of the similar angle include vector contour line segments having an angle difference of [0 °,2 ° ] ℃ [88 °,90 ° ];
s2, calculating a difference value between the first candidate value and the second candidate value, dividing the building structure into a standard building structure and a building structure to be corrected according to the relation between the difference value and a first preset threshold value, and acquiring a real main angle of the standard building structure;
s3, obtaining the main angle of the area of the building structure to be repaired through an area statistical strategy, wherein the area statistical strategy comprises outward expansion main angle statistics or line segment histogram statistics;
the region statistical strategy is outward expansion main angle statistics, and the step S3 comprises the following steps:
the minimum external rectangle of the building structure to be corrected is extended by a preset distance to obtain an extended rectangle;
taking the standard house building structures in the external expansion rectangle and intersected with the external expansion rectangle as reference house building structures;
taking the mean value of the real main angles of the reference house building as the main angles of the area of the house building to be corrected;
the region statistical strategy is line segment histogram statistics, and the step S3 comprises the following steps:
partitioning the housing construction structure in the target remote sensing image by a hierarchical clustering method based on a distance matrix;
obtaining angles of vector contour line segments of the building structures in each partition, wherein the building structures are composed of a plurality of vector contour line segments;
respectively constructing an angle histogram of each partition by utilizing the angles of vector contour line segments of the building structures in each partition;
taking the angle corresponding to the highest frequency in the angle histogram of the partition where the house building structure to be corrected is located as the main angle of the area of the house building structure to be corrected;
and S4, calculating a difference value between the candidate main angle and the area main angle, and screening out the real main angle of the building structure to be corrected from the candidate main angles according to the relation between the difference value and a second preset threshold value.
2. The method of claim 1, wherein the vector contour extraction method comprises:
performing semantic segmentation, binarization processing and mathematical morphology processing on the remote sensing image to obtain a post-processing image;
and carrying out boundary tracking and vector extraction on the post-processed image to obtain a vector contour line segment of the building structure.
3. The method of building principal angle correction based on regional uniformity of claim 1, wherein the vertical equivalence method comprises:
clustering the reference vector contour line group based on the angles of the reference vector contour line segments to obtain a main vector contour line segment group and an auxiliary vector contour line segment group, wherein the total line segment length of the main vector contour line segment group is greater than that of the auxiliary vector contour line segment group, and averaging the angles of the main vector contour line segment group and the angles of the auxiliary vector contour line segment group respectively to obtain a main vector angle and an auxiliary vector angle;
and performing vertical transformation on the auxiliary vector angle to the main vector angle to obtain a vertical auxiliary vector angle, and taking the mean value of the main vector angle and the vertical auxiliary vector angle as a second candidate value.
4. The method for correcting the main angle of the building based on the regional uniformity as claimed in claim 1, wherein the step S2 comprises:
if the difference value between the first candidate value and the second candidate value is smaller than a first preset threshold value, taking the building structure as a standard building structure, and taking the second candidate value as a real main angle of the standard building structure;
and if the difference value of the first candidate value and the second candidate value is larger than a first preset threshold value, taking the building structure as the building structure to be corrected.
5. The method for correcting the main angle of the building based on the regional uniformity as claimed in claim 1, wherein step S4 comprises:
respectively calculating the difference between the first candidate value and the main angle of the area and the difference between the second candidate value and the main angle of the area, and selecting a smaller difference as a target difference;
if the target difference value is smaller than a second preset threshold value, selecting a candidate main angle close to the main angle of the area as a real main angle of the building structure of the house to be corrected;
and if the target difference is larger than a second preset threshold value, selecting a second candidate value as a real main angle of the building structure to be corrected.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609894.5A CN115620169B (en) | 2022-12-15 | 2022-12-15 | Building main angle correction method based on regional consistency |
| PCT/CN2023/137754 WO2024125434A1 (en) | 2022-12-15 | 2023-12-11 | Regional-consistency-based building principal angle correction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211609894.5A CN115620169B (en) | 2022-12-15 | 2022-12-15 | Building main angle correction method based on regional consistency |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115620169A CN115620169A (en) | 2023-01-17 |
| CN115620169B true CN115620169B (en) | 2023-04-07 |
Family
ID=84879859
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211609894.5A Active CN115620169B (en) | 2022-12-15 | 2022-12-15 | Building main angle correction method based on regional consistency |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN115620169B (en) |
| WO (1) | WO2024125434A1 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620169B (en) * | 2022-12-15 | 2023-04-07 | 北京数慧时空信息技术有限公司 | Building main angle correction method based on regional consistency |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102938066A (en) * | 2012-12-07 | 2013-02-20 | 南京大学 | Method for reconstructing outer outline polygon of building based on multivariate data |
| CN109583284A (en) * | 2017-09-29 | 2019-04-05 | 中国科学院遥感与数字地球研究所 | Urban skyscraper object height extracting method and device based on High Resolution SAR Images |
| CN111652892A (en) * | 2020-05-02 | 2020-09-11 | 王磊 | Remote sensing image building vector extraction and optimization method based on deep learning |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2016204648A1 (en) * | 2009-05-15 | 2016-07-21 | Eagle View Technologies, Inc. | Pitch determination systems and methods for aerial roof estimation |
| WO2012037576A2 (en) * | 2010-09-17 | 2012-03-22 | Seer Technology, Inc. | Building perpendicularity testing and adjustment |
| JP6233869B2 (en) * | 2012-06-07 | 2017-11-22 | 日本電気株式会社 | Image processing apparatus, image processing apparatus control method, and program |
| CN103065295B (en) * | 2012-12-04 | 2016-01-20 | 南京大学 | A kind of aviation based on buildings angle point self-correction and ground lidar data high-precision automatic method for registering |
| CN108106594B (en) * | 2017-12-06 | 2020-05-15 | 上海建工五建集团有限公司 | Real-time measuring method for deformation of building |
| CN113139453B (en) * | 2021-04-19 | 2023-04-07 | 中国地质大学（武汉） | Orthoimage high-rise building base vector extraction method based on deep learning |
| CN114494905A (en) * | 2022-01-26 | 2022-05-13 | 中科星图股份有限公司 | Building identification and modeling method and device based on satellite remote sensing image |
| CN114241338A (en) * | 2022-02-15 | 2022-03-25 | 中航建筑工程有限公司 | Building measuring method, device, equipment and storage medium based on image recognition |
| CN114898119B (en) * | 2022-07-08 | 2022-11-01 | 浙江大华技术股份有限公司 | Building outline drawing method, device, equipment and medium |
| CN115272882A (en) * | 2022-08-03 | 2022-11-01 | 山东省国土测绘院 | Discrete building detection method and system based on remote sensing image |
| CN115620169B (en) * | 2022-12-15 | 2023-04-07 | 北京数慧时空信息技术有限公司 | Building main angle correction method based on regional consistency |
-
2022
- 2022-12-15 CN CN202211609894.5A patent/CN115620169B/en active Active
-
2023
- 2023-12-11 WO PCT/CN2023/137754 patent/WO2024125434A1/en unknown
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102938066A (en) * | 2012-12-07 | 2013-02-20 | 南京大学 | Method for reconstructing outer outline polygon of building based on multivariate data |
| CN109583284A (en) * | 2017-09-29 | 2019-04-05 | 中国科学院遥感与数字地球研究所 | Urban skyscraper object height extracting method and device based on High Resolution SAR Images |
| CN111652892A (en) * | 2020-05-02 | 2020-09-11 | 王磊 | Remote sensing image building vector extraction and optimization method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115620169A (en) | 2023-01-17 |
| WO2024125434A1 (en) | 2024-06-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | Tender tea shoots recognition and positioning for picking robot using improved YOLO-V3 model | |
| WO2024077812A1 (en) | Single building three-dimensional reconstruction method based on point cloud semantic segmentation and structure fitting | |
| CN112070769B (en) | Layered point cloud segmentation method based on DBSCAN | |
| US9036915B2 (en) | Architectural pattern detection and modeling in images | |
| CN103034863B (en) | The remote sensing image road acquisition methods of a kind of syncaryon Fisher and multiple dimensioned extraction | |
| CN110472676A (en) | Stomach morning cancerous tissue image classification system based on deep neural network | |
| CN105740798B (en) | A kind of point cloud object scene recognition methods based on structural analysis | |
| CN111652892A (en) | Remote sensing image building vector extraction and optimization method based on deep learning | |
| JP2017521779A (en) | Detection of nuclear edges using image analysis | |
| CN113223042B (en) | Intelligent acquisition method and equipment for remote sensing image deep learning sample | |
| CN112084871B (en) | High-resolution remote sensing target boundary extraction method based on weak supervised learning | |
| CN112163622A (en) | Overall situation and local fusion constrained line segment feature matching method for aviation wide-baseline stereopair | |
| CN113409332B (en) | Building plane segmentation method based on three-dimensional point cloud | |
| WO2024125434A1 (en) | Regional-consistency-based building principal angle correction method | |
| CN110175574A (en) | A kind of Road network extraction method and device | |
| Zheng et al. | Building recognition of UAV remote sensing images by deep learning | |
| CN114742957B (en) | Building facade extraction method based on point cloud data | |
| CN116152494A (en) | Building foot point identification segmentation method based on two-stage 3D point cloud semantic segmentation | |
| CN109598261B (en) | Three-dimensional face recognition method based on region segmentation | |
| CN112241676A (en) | Method for automatically identifying terrain sundries | |
| CN117670895A (en) | Immunohistochemical pathological image cell segmentation method based on section re-staining technology | |
| CN116052110B (en) | Intelligent positioning method and system for pavement marking defects | |
| CN116452604A (en) | Complex substation scene segmentation method, device and storage medium | |
| CN109117841B (en) | Scene text detection method based on stroke width transformation and convolutional neural network | |
| CN116778164A (en) | Semantic segmentation method for improving deep V < 3+ > network based on multi-scale structure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |