CN111091049B - Road surface obstacle detection method based on reverse feature matching - Google Patents
Road surface obstacle detection method based on reverse feature matching Download PDFInfo
- Publication number
- CN111091049B CN111091049B CN201911061150.2A CN201911061150A CN111091049B CN 111091049 B CN111091049 B CN 111091049B CN 201911061150 A CN201911061150 A CN 201911061150A CN 111091049 B CN111091049 B CN 111091049B
- Authority
- CN
- China
- Prior art keywords
- road
- point
- obstacle
- points
- feature
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 26
- 238000000034 method Methods 0.000 claims abstract description 33
- 230000008859 change Effects 0.000 claims abstract description 29
- 238000012216 screening Methods 0.000 claims abstract description 15
- 230000004888 barrier function Effects 0.000 claims description 8
- 230000002776 aggregation Effects 0.000 claims description 7
- 238000004220 aggregation Methods 0.000 claims description 7
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 claims description 3
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 6
- 238000012545 processing Methods 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000007547 defect Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 230000006872 improvement Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 239000004575 stone Substances 0.000 description 2
- 238000007792 addition Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000000746 purification Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/211—Selection of the most significant subset of features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
The invention discloses a road surface obstacle detection method based on reverse feature matching, which belongs to the field of remote sensing change detection, and is characterized in that the method is used for extracting change information in an image based on an aerospace high-resolution remote sensing image by using a SIFT feature point reverse matching method, and the area and position information of the road surface obstacle in the image are obtained by processing the change information. The method has the characteristics or key links that: respectively calculating SIFT feature point sets for remote sensing images before and after the road loss, carrying out feature matching, and inverting the matching result to obtain a change information feature point set; screening by using a road buffer area to obtain a characteristic point set of change information in a road area, and screening out pseudo change points and tiny change points by a sliding window method; and finally, grouping and combining the characteristic points to obtain the position information of the obstacle, and obtaining the approximate range of the obstacle by combining a region growing method to estimate the area. The key technology of the invention belongs to the category of remote sensing change detection, but the position and area information of the road obstacle can be obtained without image registration, and the degree of automation is high.
Description
Technical Field
The invention relates to a road obstacle detection problem, in particular to a road traffic obstacle problem generated by disasters such as debris flow, landslide, collapse and the like detected by using a remote sensing technology, and belongs to the technical field of remote sensing.
Background
By the end of 2018, the total mileage of China reaches 484.65 ten thousand kilometers, and the highway density reaches 50.48 kilometers per hundred square kilometers. The mountain area of China accounts for more than two thirds of the total area of China, and mountain roads are easily damaged by disasters such as debris flow, landslide, collapse, rolling stones and the like. Only 2018 has 501 geological disasters leading to road damage, resulting in huge economic losses. The obstacle information in the road passing direction is timely ascertained, the rush-through time is judged according to the characteristics, the size and other factors of the obstacle, and the advancing scheme is adjusted in real time, so that the method has great practical significance for military and civil traffic organizations.
The traditional road obstacle information acquisition mode mainly relies on manual field investigation, and has the defects of great consumption of manpower and material resources, difficult data organization, low informatization degree, poor real-time performance and the like. With the development of remote sensing technology, the remote sensing technology is more used for road damage detection at present. The main methods are three types: human-machine interaction interpretation, image-classification-based machine interpretation, and change-detection-based machine interpretation. The damage evaluation reliability is high through a man-machine interaction interpretation mode, but the automation degree is low, and the workload is large. The method based on image classification can perform qualitative analysis on the road obstacle extraction result, but the accuracy of the method depends on the accuracy of classification. The accuracy of the method based on the change detection depends on the registration accuracy, and the detection result needs further interpretation to obtain more accurate obstacle information.
Therefore, the existing road damage remote sensing detection technology has the defects of low automation degree, difficult registration, personnel operation with professional knowledge and the like.
Disclosure of Invention
The invention aims at the defects of the prior art, and provides a road obstacle detection method based on reverse feature matching, which is used for acquiring the position and area information of a road obstacle by processing feature points which are not matched with a pre-disaster image in a post-disaster image.
In order to solve the technical problems, the technical scheme of the invention is as follows, and the road surface obstacle detection method based on reverse feature matching comprises the following steps:
firstly, respectively calculating SIFT feature point sets of remote sensing images before and after the road loss, carrying out feature matching, and inverting the matching result to obtain a change information feature point set;
screening by using a road buffer area to obtain a change information feature point set in a road area;
step three, screening out pseudo-change characteristic points and tiny change characteristic points according to the distribution aggregation degree of the characteristic points;
step four, the characteristic points are initially grouped according to the point spacing, and the characteristic points belonging to different barriers are distinguished;
step five, detecting the preliminary grouping result obtained in the previous step, and obtaining a final grouping result through similarity judgment;
step six, obtaining the position information of each obstacle according to the grouping result, and obtaining the distribution range of the obstacles by combining a multi-seed point region growing method and estimating the area.
In a further improvement of the present invention, in the first step, the reverse feature matching formula is:
Q u ＝C Q Q m (1)
point set Q u Is the point set Q m Complement in Q, i.eWherein Q is a characteristic point set of the post-disaster image, Q m Is a feature point set which can be matched with the feature point of the pre-disaster image in the post-disaster image, Q u And the characteristic point set is the image change information.
In the second step, the road buffer is generated by using the existing road vector data, and the change characteristic point set Q in the buffer is screened ur The formula is:
where Poly is a road buffer polygon, i=0, 1, …, k, k is a point set Q u Is a feature point total number of (a).
In the third step, the specific implementation steps of screening feature points according to the distribution aggregation degree are as follows:
and 3.1, selecting a circular detection window with the radius r=15 pixels, and continuously moving the detection window by taking each unmatched characteristic point as a circle center.
And 3.2, counting the number N1 of unmatched feature points and the number N2 of all feature points in the window, calculating the proportion of the unmatched points to the total number of the feature points, namely the value of N1/N2, and if N1 is more than or equal to N and N1/N2 is more than m, reserving the feature points.
The reference formula of the step is:
through the step, an obstacle characteristic point set Q can be obtained obs 。
In the fourth step, the preliminary grouping of the feature points according to the point spacing mainly comprises the following steps:
and 4.1, determining the road direction. The inclination angle theta (theta epsilon 0 pi) of the road center line can be calculated from the road center line data. If it isThe link direction is defined as the vertical direction. If->The link direction is defined as a horizontal direction.
And 4.2, sorting the characteristic points. The feature points of the road segments in the vertical direction are ordered according to the coordinates y, and the feature points of the road segments in the horizontal direction are ordered according to the coordinates x.
And 4.3, searching backwards from the first point, wherein the distance between two adjacent points is not more than R, and the two adjacent points belong to the same group until a certain point q is found, and after the point q, no other characteristic points exist in the range of the distance R, so that the first group of searching is completed. Taking the median value of all the characteristic point coordinates in the group as a candidate coordinate point O of the obstacle i . And the next feature point is taken as the first point.
Step 4.4, repeating step 4.3 until the Q is checked obs And obtaining the candidate coordinate point set O of the obstacle according to all the characteristic points in the model.
In the fifth step, the detailed grouping result is determined by similarity, and the detailed steps are as follows:
and 5.1, obtaining a standard road sample, intercepting a road center line by using the obstacle candidate coordinate point set O obtained in the last step, and intercepting the distance of a road section from a locating point R/2 so as to avoid taking out a roadblock pixel point. Taking each point on the extracted road center line as a center, taking eight adjacent areas as a road sample, as shown in fig. 2. And counting the mean value and variance of all road samples, sorting the samples by the two statistics, and taking the sample corresponding to the statistics mode as a standard road surface sample.
Marking coordinate point properties, defining the forward direction of the road as the forward direction of the coordinate axis, taking R/2 as the step length along the forward direction of the road, and marking each coordinate point O i The front side point and the rear side point are respectively compared with the standard pavement sample and the similarity of the obstacle point, and different labels are given to the positioning points according to four different conditions as shown in table 1.
TABLE 1 Properties of obstacle candidate points
And 5.3, merging, judging that two points are points positioned on two sides of the same obstacle if two continuous coordinate points are respectively +1 and-1 along the road advancing direction, and taking the average value of the two coordinates as the positioning point of the obstacle. Obtaining a final obstacle locating point set R obs ：
In c i ，c i+1 Respectively represent obstacle coordinate points O i ，O i+1 Is a property of (a).
In the sixth step, a multi-seed point region growing method is adopted: taking the locating point and a neighborhood point with the difference between the locating point and the standard road sample mean value smaller than the threshold value alpha as seed points; and the difference between the point to be grown and any seed point is smaller than the threshold value alpha, so that the growth condition is met. After the growth is finished, the distribution range of the obstacles can be obtained, the number of pixels of each obstacle in the road area is counted, and the actual occupied area of each obstacle can be estimated approximately by combining the image resolution.
In the step 3.2, the value of n is generally 3-6; m is generally 0.5.
In the step 4.3, the threshold value R is set according to the project requirement and the image resolution, and the higher the precision requirement is, the smaller R is, and the larger R is otherwise.
In a further improvement of the present invention, in the seventh step, the threshold α is generally between 30 and 70.
According to the scheme, the empty and sky remote sensing images are utilized, the road surface change area feature point set is obtained through the reverse feature point matching technology, the change feature point set is clustered according to the feature point distribution density and the adjacent point distance, barriers at different positions can be segmented, and then positioning and area calculation are carried out on each barrier. The scheme is improved in the aspects of barrier segmentation, positioning, unique judgment of positioning points of various barriers and the like. The obstacle recognition and positioning method provided by the application has the advantages of wide search range, no need of image registration, high automation degree, high detection accuracy and the like. As shown in the case: the road obstacle omission ratio is 0, the false detection ratio is lower than 10%, the area estimation error is lower than 10%, and the road obstacle omission ratio can provide basis for rescue and relief route planning and road traffic engineering quantity estimation.
Drawings
FIG. 1 is a schematic flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of the extraction of the standard road sample in step 4.1;
FIG. 3 is a schematic representation of the nature of the identified obstacle candidate points in step 4.2;
FIG. 4 is a schematic diagram of the distribution of obstacles in post-disaster images according to an embodiment of the present invention;
FIG. 5 is a diagram of a road buffer screening result according to an embodiment of the present invention;
FIG. 6 is a graph of a sliding window screening result according to an embodiment of the present invention;
FIG. 7 is a graph of the grouping result of the adjacent point spacing method according to an embodiment of the present invention;
fig. 8 is a diagram of the obstacle detection result of the embodiment of the present invention.
Detailed Description
The invention will be described in further detail below with reference to the drawings and the detailed description, for the purpose of enhancing the understanding and appreciation of the invention, but the invention is not limited to the examples given.
Example 1: referring to fig. 1, a road surface obstacle detection method based on reverse feature matching, the method comprising the steps of:
firstly, respectively calculating SIFT feature point sets of remote sensing images before and after the road loss, carrying out feature matching, and inverting the matching result to obtain a change information feature point set;
screening by using a road buffer area to obtain a change information feature point set in a road area;
step three, screening out pseudo-change characteristic points and tiny change characteristic points according to the distribution aggregation degree of the characteristic points;
step four, the characteristic points are initially grouped according to the point spacing, and the characteristic points belonging to different barriers are distinguished;
step five, detecting the preliminary grouping result obtained in the previous step, and obtaining a final grouping result through similarity judgment;
step six, obtaining the position information of each obstacle according to the grouping result, and obtaining the distribution range of the obstacles by combining a multi-seed point region growing method and estimating the area.
In a further improvement of the present invention, in the first step, the reverse feature matching formula is:
Q u ＝C Q Q m (1)
point set Q u Is the point set Q m Complement in Q, i.eWherein Q is a characteristic point set of the post-disaster image, Q m Is a feature point set which can be matched with the feature point of the pre-disaster image in the post-disaster image, Q u And the characteristic point set is the image change information.
In the second step, the road buffer is generated by using the existing road vector data, and the change feature point set Q in the buffer is screened ur The formula is:
where Poly is a road buffer polygon, i=0, 1, …, k, k is a point set Q u Is a feature point total number of (a).
In the third step, the specific implementation steps of screening feature points according to the distribution aggregation degree are as follows:
and 3.1, selecting a circular detection window with the radius r=15 pixels, and continuously moving the detection window by taking each unmatched characteristic point as a circle center.
And 3.2, counting the number N1 of unmatched feature points and the number N2 of all feature points in the window, calculating the proportion of the unmatched points to the total number of the feature points, namely the value of N1/N2, and if N1 is more than or equal to N and N1/N2 is more than m, reserving the feature points.
The reference formula of the step is:
through the step, an obstacle characteristic point set Q can be obtained obs 。
In the fourth step, the preliminary grouping of the feature points according to the point distances mainly includes the following steps:
and 4.1, determining the road direction. The inclination angle theta (theta epsilon 0 pi) of the road center line can be calculated from the road center line data. If it isThe link direction is defined as the vertical direction. If->The link direction is defined as a horizontal direction.
And 4.2, sorting the characteristic points. The feature points of the road segments in the vertical direction are ordered according to the coordinates y, and the feature points of the road segments in the horizontal direction are ordered according to the coordinates x.
Step 4.3, searching backward from the first point, wherein the distance between two adjacent points is not more than R, and the two adjacent points belong to the same group until a certain point q is found, at this pointNo other feature points exist within the range of the post-R distance, the first set of searches is completed. Taking the median value of all the characteristic point coordinates in the group as a candidate coordinate point O of the obstacle i . And the next feature point is taken as the first point.
Step 4.4, repeating step 4.3 until the Q is checked obs And obtaining the candidate coordinate point set O of the obstacle according to all the characteristic points in the model.
In the fifth step, the specific steps of the refined grouping result through similarity judgment are as follows:
and 5.1, obtaining a standard road sample, intercepting a road center line by using the obstacle candidate coordinate point set O obtained in the last step, and intercepting the distance of a road section from a locating point R/2 so as to avoid taking out a roadblock pixel point. Taking each point on the extracted road center line as a center, taking eight adjacent areas as a road sample, as shown in fig. 2. And counting the mean value and variance of all road samples, sorting the samples by the two statistics, and taking the sample corresponding to the statistics mode as a standard road surface sample.
Marking coordinate point properties, defining the forward direction of the road as the forward direction of the coordinate axis, taking R/2 as the step length along the forward direction of the road, and marking each coordinate point O i And the front side point and the rear side point are respectively compared with the standard pavement sample and the obstacle point, as shown in fig. 3. The anchor points are given different labels for four different situations as shown in table 1.
TABLE 1 Properties of obstacle candidate points
And 5.3, merging, judging that two points are points positioned on two sides of the same obstacle if two continuous coordinate points are respectively +1 and-1 along the road advancing direction, and taking the average value of the two coordinates as the positioning point of the obstacle. Obtaining a final obstacle locating point set R obs ：
In c i ，c i+1 Respectively represent obstacle coordinate points O i ，O i+1 Is a property of (a).
In the sixth step, a multi-seed point region growing method is adopted: taking the locating point and a neighborhood point with the difference between the locating point and the standard road sample mean value smaller than the threshold value alpha as seed points; and the difference between the point to be grown and any seed point is smaller than the threshold value alpha, so that the growth condition is met. After the growth is finished, the distribution range of the obstacles can be obtained, the number of pixels of each obstacle in the road area is counted, and the actual occupied area of each obstacle can be estimated approximately by combining the image resolution.
In the step 3.2, the value of n is generally 3-6; m is generally 0.5.
In the step 4.3, the threshold R is set according to the project requirement and the image resolution, and the higher the accuracy requirement is, the smaller R is, and the larger R is otherwise.
In the seventh step, the value of the threshold value alpha is generally between 30 and 70.
Application example 1: a pavement obstacle detection method based on reverse feature matching, the method comprising the steps of: the technical flow chart of the invention is shown in fig. 1. The key steps in fig. 1 will be described separately for the flow chart.
Step one, a SIFT algorithm is used for detecting feature point sets (respectively marked as P and Q) of images in two stages before and after a disaster and carrying out feature point matching, in order to ensure the accuracy of extracting the feature points of subsequent barriers, a ratio purification method is used for reducing mismatching, and an RANSAC algorithm is used for removing outlier matching and further purifying a matching result. After reverse feature matching, a change information feature point set Q of the post-disaster image is obtained u ；
Step two, the change information characteristic point set Q in the road area is extracted after the road buffer area is screened ur . FIG. 4 is a graph showing the result of extracting a road domain change feature point set using a road ROI;
step three, according to the image resolution and the texture characteristics of the road area, each parameter of the sliding window screening method is r=15 pixels, n=5 (the number of unmatched points in the window is more than or equal to 5), m=0.5 (the number of unmatched points is more than half of the total number of characteristic points), and fig. 5 (a) and (b) are examples of screening results of the sliding window, so that a plurality of characteristic point clusters with high aggregation degree are obtained;
step four, the road obstacle points within 10m of the case setting field are not distinguished, and the threshold r=20 because the resolution of the experimental image is 0.5 m. The road is calculated to be in a horizontal trend by the road vector data, so the obstacle characteristic point set Q obs Ordering according to the coordinate x, and Q according to R obs The method is divided into 15 groups, and 15 obstacle candidate points are obtained after median value is found in each group. 6 (a) and (b) are examples of obstacle candidate points obtained by using a neighboring point pitch grouping method;
as can be seen from fig. 5 (b), the detected feature points belong to the same obstacle, but two candidate points are grouped and located as shown in fig. 6 (b), and the two candidate points need to be combined, and the results of calibrating the properties of the 15 candidate points according to the implementation steps of the similarity comparison method are shown in table 2:
TABLE 2 calibration results of candidate Point Properties
And (5) obtaining 4 groups of coordinate points to be combined through comparison and statistics. And taking the combined new coordinate values as positioning points of the 4 obstacles, keeping the rest 7 coordinate points unchanged, and finally obtaining positioning point coordinates of the 11 obstacles, wherein the positioning point coordinates are shown in Table 3.
Step six, after obtaining accurate obstacle locating points, the shape information of each obstacle can be obtained through a plurality of sub-region growth methods, and the white part in fig. 8 is the distribution range of each obstacle. As can be seen from fig. 8, the firstNo growth of the obstacle is resulted from the +.>The neighborhood of the obstacle locating point is very similar to the standard pavement sample, and no obstacle locating point existsA suitable seed point is found, and the area of the obstacle there is thus set to 0. The remaining 10 obstacles are subjected to area estimation by counting the number of pixels of which the area growth results are positioned in the road area, and the calculation results are shown in Table 3; the reference area in table 3 is the area calculated by interpreting the range of the outlined obstacle boundary using human vision. The area is the area constructed by the road buffer zone boundary and the extracted obstacle boundary, and can be directly used as the reference of the rescue earth and stone quantity.
TABLE 3 obstacle extraction results
The above embodiments are merely preferred embodiments of the present invention, and are not intended to limit the present invention in any other way, but modifications, additions or the like are made according to the technical spirit of the present invention, and still fall within the scope of the present invention as claimed.
Claims (8)
1. The road surface obstacle detection method based on reverse feature matching is characterized by comprising the following steps of:
firstly, respectively calculating SIFT feature point sets of remote sensing images before and after the road loss, carrying out feature matching, and inverting the matching result to obtain a change information feature point set;
screening by using a road buffer area to obtain a change information feature point set in a road area;
step three, screening out pseudo-change characteristic points and tiny change characteristic points according to the distribution aggregation degree of the characteristic points;
step four, the characteristic points are initially grouped according to the point spacing, and the characteristic points belonging to different barriers are distinguished;
step five, detecting the preliminary grouping result obtained in the previous step, and obtaining a final grouping result through similarity judgment;
step six, obtaining the position information of each obstacle according to the grouping result, and obtaining the distribution range of the obstacles by combining a multi-seed point region growing method and estimating the area;
in the third step, the specific implementation steps of screening feature points according to the distribution aggregation degree are as follows:
step 3.1, selecting a circular detection window with the radius r=15 pixels, and continuously moving the detection window by taking each unmatched characteristic point as a circle center;
step 3.2, counting the number N1 of unmatched feature points and the number N2 of all feature points in the window, calculating the proportion of the unmatched points to the total number of the feature points, namely the value of N1/N2, and if N1 is more than or equal to N and N1/N2 is more than m, reserving the feature points;
the reference formula of the step is:
through the step, an obstacle characteristic point set Q can be obtained obs ；
In the sixth step, a multi-seed point region growing method is adopted: taking the locating point and a neighborhood point with the difference between the locating point and the standard road sample mean value smaller than the threshold value alpha as seed points; and when the difference between the growing point and any seed point is smaller than the threshold value alpha, namely the growing condition is met, the distribution range of the obstacles can be obtained after the growth is finished, the number of pixels of each obstacle in the road area is counted, and the actual occupied area of each obstacle can be estimated approximately by combining the image resolution.
2. The road surface obstacle detection method based on reverse feature matching as claimed in claim 1, wherein: in step one, the inverse feature matching formula is Q u ＝C Q Q m Point set Q u Is the point set Q m Complement in Q, i.e. Q u ：Wherein Q is a characteristic point set of the post-disaster image, Q m Is a feature point set which can be matched with the feature point of the pre-disaster image in the post-disaster image, Q u And the characteristic point set is the image change information.
3. The method for detecting a road obstacle based on reverse feature matching according to claim 1, wherein in the second step, the road buffer used is generated from existing road vector data to obtain the change feature point set Q in the road buffer ur The reference formula is:
poly is a road buffer polygon, i=0, 1, ··, k, k is the point set Q u Is a feature point total number of (a).
4. The method for detecting a road surface obstacle based on reverse feature matching according to claim 1, wherein in the fourth step, the preliminary grouping of the feature points according to the point pitch mainly comprises the steps of:
step 4.1, determining the road direction, calculating the inclination angle theta (theta E [0, pi ]) of the road section center line according to the road center line data, ifThe road direction is defined as vertical if +.>The road section direction is specified as the horizontal direction;
step 4.2, sorting the characteristic points of the road sections in the vertical direction according to the coordinates y, and sorting the characteristic points of the road sections in the horizontal direction according to the coordinates x;
step 4.3, searching backwards from the first point, wherein the distance between two adjacent points is not more than R, and the two adjacent points belong to the same group until a certain point q is found, no other characteristic points exist within the range of the distance R after the point, the first group of searching is completed, and the median value of the coordinates of all the characteristic points in the group is taken as a candidate coordinate point O of an obstacle at the position i And takes the next feature point asA first point;
step 4.4, repeating step 4.3 until the Q is checked obs And obtaining the candidate coordinate point set O of the obstacle according to all the characteristic points in the model.
5. The method for detecting a road surface obstacle based on reverse feature matching according to claim 1, wherein in the fifth step, the specific step of determining the refined grouping result by similarity is as follows:
step 5.1, obtaining a standard road sample, intercepting a road center line by using the obstacle candidate coordinate point set O obtained in the previous step, and intercepting the distance from a road section to a locating point R/2 so as to avoid taking out a roadblock pixel point; taking each point on the extracted road center line as a center, taking eight adjacent areas of the point as a road sample, counting the mean value and the variance of all the road samples, sequencing the samples according to the two statistics, and taking the sample corresponding to the statistics mode as a standard road sample;
marking coordinate point properties, defining the forward direction of the road as the forward direction of the coordinate axis, taking R/2 as the step length along the forward direction of the road, and marking each coordinate point O i Respectively taking one point from the front side and the rear side of the road surface, respectively comparing the similarity of the front side point and the rear side point with the standard road surface sample and the obstacle point, and giving different labels to positioning points according to four different conditions, wherein the labels are shown in the table 1:
TABLE 1 Properties of obstacle candidate points
Step 5.3, merging, judging two points to be points positioned on two sides of the same obstacle if two continuous coordinate points are respectively +1 and-1 along the road advancing direction, and taking the average value of the two coordinates as the positioning point of the obstacle at the position; obtaining a final obstacle locating pointSet R obs ：
In c i ，c i+1 Respectively represent obstacle coordinate points O i ，O i+1 Is a property of (a).
6. The method for detecting a road surface obstacle based on reverse feature matching according to claim 5, wherein in step 3.2, n has a value of 3 to 6; m is 0.5.
7. The method for detecting a road surface obstacle based on reverse feature matching according to claim 6, wherein in step 4.3, the threshold R is set according to the project requirement and the image resolution, and the higher the accuracy requirement, the smaller R, and vice versa.
8. The method for detecting a road surface obstacle based on reverse feature matching according to claim 7, wherein the threshold α has a value between 30 and 70.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911061150.2A CN111091049B (en) | 2019-11-01 | 2019-11-01 | Road surface obstacle detection method based on reverse feature matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911061150.2A CN111091049B (en) | 2019-11-01 | 2019-11-01 | Road surface obstacle detection method based on reverse feature matching |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111091049A CN111091049A (en) | 2020-05-01 |
| CN111091049B true CN111091049B (en) | 2024-02-09 |
Family
ID=70393483
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911061150.2A Active CN111091049B (en) | 2019-11-01 | 2019-11-01 | Road surface obstacle detection method based on reverse feature matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111091049B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111783700B (en) * | 2020-07-06 | 2023-11-24 | 中国交通通信信息中心 | Automatic recognition and early warning method and system for pavement foreign matters |
| CN111858342A (en) * | 2020-07-23 | 2020-10-30 | 深圳慕智科技有限公司 | Fuzzy test data generation method based on intelligent traffic image input feature recognition |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103914688A (en) * | 2014-03-27 | 2014-07-09 | 北京科技大学 | Urban road obstacle recognition system |
| WO2016076449A1 (en) * | 2014-11-11 | 2016-05-19 | Movon Corporation | Method and system for detecting an approaching obstacle based on image recognition |
| CN109785307A (en) * | 2019-01-09 | 2019-05-21 | 武汉大学 | A kind of unmanned plane image road Damage assessment method based on vector guidance |
| CN109815859A (en) * | 2019-01-10 | 2019-05-28 | 中交信息技术国家工程实验室有限公司 | A kind of inferior grade road automatically extracts and mutation analysis method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4406381B2 (en) * | 2004-07-13 | 2010-01-27 | 株式会社東芝 | Obstacle detection apparatus and method |
-
2019
- 2019-11-01 CN CN201911061150.2A patent/CN111091049B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103914688A (en) * | 2014-03-27 | 2014-07-09 | 北京科技大学 | Urban road obstacle recognition system |
| WO2016076449A1 (en) * | 2014-11-11 | 2016-05-19 | Movon Corporation | Method and system for detecting an approaching obstacle based on image recognition |
| CN109785307A (en) * | 2019-01-09 | 2019-05-21 | 武汉大学 | A kind of unmanned plane image road Damage assessment method based on vector guidance |
| CN109815859A (en) * | 2019-01-10 | 2019-05-28 | 中交信息技术国家工程实验室有限公司 | A kind of inferior grade road automatically extracts and mutation analysis method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111091049A (en) | 2020-05-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108519605B (en) | Road edge detection method based on laser radar and camera | |
| CN109949326B (en) | Building contour line extraction method based on knapsack type three-dimensional laser point cloud data | |
| CN112287807B (en) | Remote sensing image road extraction method based on multi-branch pyramid neural network | |
| CN108197583B (en) | Building change detection method based on graph cut optimization and image structure characteristics | |
| WO2017120897A1 (en) | Object surface deformation feature extraction method based on line scanning three-dimensional point cloud | |
| CN111784657A (en) | Digital image-based system and method for automatically identifying cement pavement diseases | |
| CN108921076B (en) | Pavement crack disease self-adaptive constant false alarm detection method based on image | |
| CN111091049B (en) | Road surface obstacle detection method based on reverse feature matching | |
| CN111145157B (en) | Road network data automatic quality inspection method based on high-resolution remote sensing image | |
| EP4120123A1 (en) | Scan line-based road point cloud extraction method | |
| CN105069395B (en) | Roadmarking automatic identifying method based on Three Dimensional Ground laser scanner technique | |
| CN112819809B (en) | Method for quantifying morphology of mineral particles in rock | |
| CN112070756B (en) | Three-dimensional road surface disease measuring method based on unmanned aerial vehicle oblique photography | |
| CN115115627B (en) | Soil saline-alkali soil monitoring method based on data processing | |
| CN112964712A (en) | Method for rapidly detecting state of asphalt pavement | |
| CN112801075A (en) | Automatic rural road boundary line extraction method based on aerial image | |
| CN111524154B (en) | Image-based tunnel segment automatic segmentation method | |
| CN114387328A (en) | RQD calculation method based on deep learning model and core image | |
| CN110705595B (en) | Parking space automatic extraction method based on knapsack type three-dimensional laser point cloud data | |
| JP6600615B2 (en) | Clustering device, artifact identification device, trunk detection device, method, and program | |
| Huang et al. | Crack fundamental element (CFE) for multi-scale crack classification | |
| CN112906627B (en) | Green pricklyash peel identification method based on semantic segmentation | |
| CN115294066A (en) | Sandstone particle size detection method | |
| CN110070109B (en) | Method for automatically clustering ground feature scattered points in building arrangement dense area | |
| CN109308468B (en) | Lane line detection method |
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 |