CN107610174A - A kind of plane monitoring-network method and system based on depth information of robust - Google Patents
A kind of plane monitoring-network method and system based on depth information of robust Download PDFInfo
- Publication number
- CN107610174A CN107610174A CN201710866367.5A CN201710866367A CN107610174A CN 107610174 A CN107610174 A CN 107610174A CN 201710866367 A CN201710866367 A CN 201710866367A CN 107610174 A CN107610174 A CN 107610174A
- Authority
- CN
- China
- Prior art keywords
- plane
- detection
- growth
- seed block
- effective
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
Abstract
The present invention is applied to image procossing, there is provided a kind of plane monitoring-network method based on depth information of robust, including：Depth map is received, extracts some germinative seed blocks in the depth map；Region growing is carried out according to the germinative seed block, obtains the generation plane of the germinative seed block；The generation plane of the germinative seed block is carried out outgrowth correction or owes growth to correct, obtains effective growth plane of the germinative seed block；Output includes the depth map detection plane of effective growth plane of the germinative seed block.The embodiment of the present invention carries out outgrowth correction or owes growth to correct by the growth plane for obtaining germinative seed block region growing, improves the accuracy and robustness of plane monitoring-network method.
Description
Technical Field
The invention belongs to the field of image processing and computer vision, and particularly relates to a robust depth information-based plane detection method and system.
Background
Since the plane carries information of the direction and size of objects in the 3D scene, 3D reconstruction can be performed using planar detection techniques. The 3D reconstruction can be simply summarized as a process of plane detection of indoor and outdoor scenes and building a segmented plane model. In addition, the plane detection technology is also widely used for object detection in a robot navigation system and computer vision.
In early plane detection work, texture information of planes was mainly utilized, but when the colors or textures of the planes are not consistent, a great challenge is brought to the method. The distance information of the depth map is utilized to solve the problems in the prior art, and the result also proves that the method can effectively deal with the complex situation. The depth map can be generated directly by a depth camera (e.g. swissfinger SR40001, microsoft Kinect) or synthesized by software, and different values in the map reflect distance information of objects in the scene relative to the shooting camera. Since the depth map represents spatial information for each point in the scene, points from the same plane will have similar spatial features, such as gradients and normal vectors. Based on this, there is a work that a plane of a significant object in a scene is segmented by using a local gradient, and then a ground is segmented by using a Random Sample Consensus (RANSAC). In addition, the real-time plane detection is realized by extracting three components of normal vectors of each point and clustering points with similar directions, but the accuracy and the robustness of the three components are poor.
According to the working principle, the plane detection method can be divided into three categories: the iterative plane fitting method is based on a Hough transform method and a region growing method. The iterative plane fitting method is a common method for plane detection, and is typically represented by RANSAC, in which a fitting model is initialized from several randomly selected points. The method has good effect on detecting the large plane and strong robustness on noise, but the calculation amount is too large, and the complex plane can be excessively simplified in the calculation process. Hough transform-based methods are commonly used for parametric object detection, especially for lines and circles in 2D planes. In order to make this type of method usable in 3D space and reduce computational consumption, a variety of hough transform-based derivation algorithms have emerged. For example, the 3D hough transform method represents a plane by using the slopes of the plane in the x-axis and y-axis directions and the distance from the coordinate origin, but has a high calculation cost when searching for parameters of a fitting model, and the problem is more prominent particularly when the input data is large or the accumulator is sensitive. The Random Hough Transform (RHT) calculates parameters using a probability model, thereby avoiding high calculation cost in finding optimal parameters. The main idea of the region growing method is to use the correlation between adjacent points to construct a plane, and there is work to propose an algorithm based on the growth of two seed points, and gradually update plane parameters through the centroid and covariance matrix of a grown region, but the calculation amount is too large. Work also proposes a Cached-Octree Region Growing algorithm (CORG) to segment the point cloud into a plurality of planar regions, but the planar overgrowth problem may occur in the intersecting planar regions. There are two other work in which two growth strategies are proposed: a sub-window growing algorithm in the point cloud of the structured environment and a hybrid growing algorithm in the unstructured environment. When the window size is set properly, the algorithm is faster than the rate of point-based growth. In addition, work has been directed to Robust Principle Component Analysis (RPCA) to separate all 3D points into boundary line points, corner points and bins. The plane growth process will start with one bin, and if the angle between the growing bin and its neighboring bin is less than a certain threshold, the neighboring bin will be included in the currently growing plane. The method works normally when the included angle of adjacent surface elements is acute, but the method may fail when the included angle of adjacent surface elements is obtuse.
Disclosure of Invention
The invention aims to solve the technical problem of providing a robust depth information-based plane detection method and system, and aims to solve the problem that the accuracy and robustness of the conventional plane detection algorithm are poor in a complex scene.
The invention is realized in such a way that a robust depth information-based plane detection method comprises the following steps:
receiving a depth map, and extracting a plurality of effective seed blocks in the depth map;
performing region growth according to the effective seed block to obtain a generation plane of the effective seed block;
performing overgrowth correction or under-growth correction on the generation plane of the effective seed block to obtain an effective growth plane of the effective seed block;
and outputting a depth map detection plane containing the effective growth plane of the effective seed block.
The invention also provides a robust depth information-based plane detection system, which comprises:
the extraction unit is used for receiving the depth map and extracting a plurality of effective seed blocks in the depth map;
the growth unit is used for carrying out region growth according to the effective seed blocks to obtain the generation planes of the effective seed blocks;
a correcting unit, configured to perform overgrowth correction or under-growth correction on the generation plane of the effective seed block to obtain an effective growth plane of the effective seed block;
and the output unit is used for outputting a depth map detection plane containing the effective growth plane of the effective seed block.
Compared with the prior art, the invention has the beneficial effects that: according to the embodiment of the invention, the effective seed blocks of the depth map are extracted, the region growth is carried out according to the extracted effective seed blocks to obtain the growth plane of each effective seed block, the overgrowth correction or the under-growth correction is carried out on the growth plane of each effective seed block to obtain the effective growth plane, and the depth map detection plane is output according to the effective growth plane. According to the embodiment of the invention, the overgrowth correction or the under-growth correction is carried out on the growth plane obtained by the growth of the effective seed block area, so that the accuracy and the robustness of the plane detection method are improved.
Drawings
FIG. 1 is a flowchart of a robust depth information-based plane detection method according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a set of neighboring points provided by an embodiment of the present invention;
FIG. 3 is a diagram of neighboring points of an effective seed block provided by an embodiment of the present invention;
FIG. 4 is a flow chart of an overgrowth correction method provided by an embodiment of the invention;
FIG. 5 is a flow chart of an under-growth correction method provided by an embodiment of the invention;
fig. 6 is a schematic structural diagram of a robust depth information-based planar detection system according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Fig. 1 shows a robust depth information-based plane detection method provided by an embodiment of the present invention, which includes:
s101, receiving a depth map, and extracting a plurality of effective seed blocks in the depth map;
s102, performing region growing according to the effective seed blocks to obtain a generating plane of the effective seed blocks;
s103, performing overgrowth correction or under-growth correction on the generation plane of the effective seed block to obtain an effective growth plane of the effective seed block;
s104, outputting a depth map detection plane containing the effective growth plane of the effective seed block.
In order to solve the problem that the accuracy and robustness of the existing Plane Detection algorithm are poor in a complex scene, the embodiment of the invention provides a robust Depth information-driven Plane Detection method (DPD), namely, the Depth information is used for Plane Detection, and the Plane Detection method comprises two parts: and (3) performing plane detection based on seed block growth and a post-processing process for further enhancing the robustness of the algorithm. The plane detection method starts from the seed block with the highest smoothness, and guides the growth process by using a plane equation and a dynamic threshold function of a fitting plane of the growth plane. Under the action of the mechanism, when the growth of the seed block reaches the maximum range, the next seed block starts to grow, and the growth process is continuously iterated until all planes are detected. The accuracy and robustness of the method can be improved by utilizing a dynamic threshold function and a post-processing process of an enhancement mechanism, wherein the latter is provided for the problems of plane overgrowth and under-growth which are easy to occur in a growth-based plane detection method.
Specifically, the plane equation of the fitting plane refers to a plane equation of the growing plane obtained by a linear least square plane fitting method, and the plane equation of the fitting plane is assumed to be z = p 00 +p 10 x+p 01 y, has p 00 +p 10 x+p 01 y-z =0. For n pixels on the plane, (x) i ,y i ,z i ) I =1,2 \ 8230n, and fitting the above-mentioned plane equation with the n pixel points according to the linear least square plane fitting method, the value of the following formula is minimized:
namely should satisfyCalculating the coefficient p 00 ,p 10 ,p 01 So as to obtain the plane equation z, the normal vector n of the plane i ＝(p 10 ,p 01 ,-1)，d i ＝p 00 。
The following examples of the invention are further illustrated:
generating an effective seed block:
under the influence of the shooting scene and the shooting camera, holes are sometimes generated on the depth map collected by the depth camera, that is, there are no pixels with depth values. If the seed blocks are randomly selected to grow, the problems of model fitting failure and high calculation cost are easy to occur. Thus, a key step in region growing is to select a growing seed block, i.e., to select a region without holes.
Therefore, in this step, on the input depth map, an L × L rectangular window is slid one pixel point at a time in a raster scanning manner to traverse the depth map, and it is checked whether all points in the rectangular window are hole points at the position of each pixel point, when there is no hole point in the rectangular window, a point set in the rectangular window is regarded as an effective seed block, and a plane equation and a corresponding smoothness of the effective seed block are calculated by a Linear Least Squares (LLS) method, which is a best matching function for finding data by minimizing a sum of Squares of errors. Through such continuous scanning, all effective seed blocks in the input depth map are extracted.
In 3D space, a plane can be represented by a normal vector n and a distance D from the origin of the coordinate system. The fitting plane of the effective seed block can be obtained by using a linear least square plane fitting method, wherein the fitting plane of the effective seed block is assumed to be S, the plane equation of the effective seed block is f (t), and S = f (t), namely the plane equation expresses the fitting plane in a mathematical mode. Assuming that p is any point on the ith valid seed block, the fitting error between the fitting plane of the ith valid seed block and the point p is defined as:
e i (p)＝|n i ·p+d i |；
wherein n is i Normal vector representing the fitted plane of the ith valid seed block, d i The distance of the fitted plane representing the ith valid seed block from the origin of the coordinate system.
Root mean square fitting error delta i Is defined as follows:
where P represents the set of points for all pixels in the valid seed block and | P | represents the number of pixels in the valid seed block.
In the present invention, the root mean square fitting error δ i Used to represent the smoothness of the effective seed block, and thus a smaller root mean square fitting error δ i Meaning a higher smoothness. Since the plane growing process starts with the most smooth effective seed block and can ensure accurate growth of the plane, the root mean square error δ is used in the present embodiment i All the effective seed blocks are sorted from small to large.
And (3) a region growing process: region growing refers to an iterative process of generating a plane from one valid seed block. In a specific growing process, not all effective seed blocks are finally given the opportunity to grow, because the region growing starts from the effective seed block with the highest smoothness, and the growing plane of the effective seed block which is firstly grown contains some later unused effective seed blocks, and the contained effective seed blocks do not have the opportunity to grow.
The key of the region-based growth method is to distinguish the inner point and the local outer point of the current growth plane, and the judgment basis is a threshold value T, wherein the threshold value T is the output of the dynamic threshold function provided by the embodiment of the invention. And when detecting that the adjacent point set of the growth plane is an empty set or no point suitable for the current growth plane exists in the adjacent point set, determining that the growth of the current plane reaches the maximum range, and ending the growth of the current plane immediately. As shown in fig. 2, the neighboring point set refers to a set of all pixel points neighboring to the current growth plane, that is, the neighboring point set of the current growth plane is composed of points 1 to 16.
Specifically, the main contents of the growth process include:
in the first stage, all neighboring points of the most smooth valid seed block are detected, and in this step, as shown in fig. 3, the neighboring points of the 3 × 3 rectangular window are taken as the center, and the total of 8 neighboring points are taken, i.e., point 1 to point 8. When it is detected that the neighboring point is not a hole point and does not belong to the previous pointWhen the detected plane exists, the adjacent point is substituted into the plane equation of the current growth plane, and the error formula e is fitted i (p)＝|n i ·p+d i And | obtaining corresponding fitting errors, substituting the adjacent points into a fitting error formula, namely replacing the point p with the point coordinates of the adjacent points, and calculating to obtain the fitting errors. And when the fitting error is smaller than the threshold value T, merging the adjacent points into the current growth plane, and otherwise, regarding the adjacent points as local points of the current growth plane. In this stage, the fitting error needs to be calculated once for each neighboring point.
And in the second stage, updating parameters of a plane equation by using a linear least square plane fitting method LLS, and updating the root mean square fitting error. Specifically, at the first stage, the primary growth plane has been fitted, and the fitting error is calculated. When new neighboring points merge into the current growth plane, the LLS is reused to fit a new growth plane, and a new fitting error is calculated, which is an iterative process. The plane can be optimized and adjusted once and again in the growth process. In this stage, the plane equation is updated once after each growth, and a root mean square fit error is calculated for all points on the plane.
In the prior art, most plane detection algorithms adopt a fixed value as a threshold value, so that the problem of plane overgrowth or overgrowth is easily caused. Algorithms have since proposed threshold functions based on depth map camera noise models, but the thresholds proposed in these methods increase monotonically as the depth values increase. However, this does not fully correspond to reality, and the following three situations arise: 1) When the threshold value is small, the growth plane is easy to lack growth; 2) When the threshold value is larger, the facet at the far end is easy to cause overgrowth; 3) When the growth plane is parallel to the camera plane, the threshold value will become a fixed value since the depth distance of the plane does not change.
To overcome the above disadvantages, the embodiment of the present invention designs a dynamic threshold function based on the noise model and the plane size.
The dynamic threshold function is defined as follows:
wherein I represents the depth map, I d The depth value of a point which represents the fitting error of the depth map and is calculated by substituting a plane equation into the depth map and utilizing a fitting error formula, in the embodiment, the point is an adjacent point of the current growth plane, tau represents the allowable maximum roughness of the growth plane, lambda determines the increasing speed of a threshold value, H and W respectively represent the height and the width of the depth map, alpha and k are constants, j represents the iteration number of the plane in the growth process, j =1 is initialized, and the threshold value T with the maximum output is represented by I d And (6) determining.
The dynamic threshold function solves well the noise model based problem, such as: when the growth plane is a small plane at the far end, due to the consideration of the size of the growth plane, overgrowth caused by a large threshold value can be avoided; when the growth plane is parallel to the camera plane, the threshold will not be a fixed value due to noise accumulation taking into account the growth process.
And (3) post-treatment process of an enhancement mechanism:
specifically, the post-processing procedure of the forcing mechanism includes an overgrowth correction procedure and an under-growth correction procedure, which are explained below:
1. and (3) overgrowth correction:
during planar growth, one of the growth planes will preferentially grow to its intersection with the other growth plane. The overgrowth refers to a phenomenon that if the fitting error of the adjacent pixel point of the intersection line of the planes and the current plane is smaller than the current threshold value, the current growing plane grows onto the intersection plane by mistake. Wherein, the growth direction can be divided into longitudinal growth and transverse growth.
Using the two-plane equation, the intersection can be expressed as a parametric equation:
where x represents the cross product of two vectors,indicating the detection plane S i The normal vector of (a) is calculated,represents the maximum plane S u T denotes the unknown parameter,indicating the direction vector of the intersecting line, p 0 Is a point on the intersecting line defined as:
wherein, the first and the second end of the pipe are connected with each other,denotes S i The distance from the origin of the coordinate system,denotes S i Distance from the origin of the coordinate system.
The main contents of the correction process comprise:
first, the overgrowth region S is accurately detected 0 。
The degree of overgrowth mainly depends on the included angle of the intersecting planes, the accuracy of depth data and a threshold value T, wherein the included angle theta is an inverse cosine value multiplied by normal vector points of the two planes, and the threshold value T is the output maximum value of a dynamic threshold function. The theoretical width w of the overgrowth area is equal to the ratio of the threshold T to the sine value of the included angle theta, and in order to ensure that all pixel points in the overgrowth area are in the scanning range, the actual width is taken as [ w ] (1 + epsilon) in the embodiment of the invention, wherein [ w ] represents a large integer close to w, and epsilon is greater than 0, so that the actual width can be ensured to be larger than the theoretical width.
And secondly, reallocate.
Reassignment refers to the over-growth of the region S 0 Substituting the points on the farthest boundary into a plane equation of the two planes, respectively calculating the fitting errors from the points to the planes, comparing the sizes of the two fitting errors, and subjecting the overgrowth region S 0 And merging the two planes with smaller fitting errors.
Finally, the erroneously grown plane is corrected and its plane equation is updated. Likewise, the correction process is applied to other detected planes.
The flow of the overgrowth correction step is shown in fig. 4, and comprises:
(1) Taking the growth plane of the effective seed block currently corrected as the detection plane S i Finding out the detection plane S i Adjacent planes of (a);
(2) Finding out the detection plane S through the parameter equation of the intersecting line i Lines of intersection with adjacent planes;
(3) Judging whether the intersection line exists in the two-dimensional 2D range of the current depth map, and if so, setting the width of a detection scanning bar; otherwise, moving to the next adjacent plane, finding out the intersection line and judging whether the intersection line is in the range of the current depth map 2D;
(4) Sliding the detection scanning bar along the intersecting line to scan the pixel point, and obtaining an overgrowth region S o ；
(5) According to the overgrowth region S 0 Judging the current detection plane S i Whether overgrowth is made, if overgrowth is made, overgrowth region S 0 Merging with adjacent planes, or vice versa, overgrowing the region S 0 And a detection plane S i And (6) merging. In this step, i.e. if the region S is overgrown o When the points are substituted into the adjacent plane, the fitting error is smaller than that of the current detection plane, and the overgrowth area is supposed to belong to the adjacent plane, so that the overgrowth problem of the current detection plane is deduced.
(6) In the merging process, the overgrowth region S o It may be fragmented into small isolated chunks that are detected and reassigned. Specifically, in the detection of the previous step (5), it is generatedPrimary, secondary overgrowth regions, and after the primary regions have been reallocated, the secondary regions may become isolated regions. The points established in the isolated region are reassigned into the adjacent plane equations, and the plane with the least root mean square fit error is the plane to which the isolated region belongs.
(7) Respectively updating the current detection plane S i Plane equations of the planes adjacent thereto. Since each plane has its own plane equation, the plane equations of both need to be updated separately in this step.
(8) And iterating the correction process until the growth planes of all the effective seed blocks are detected, and finishing overgrowth correction.
2. An under-growth correction process:
in order to solve the under-growth condition of the growth planes, the embodiment of the invention provides a plane merging method, namely when two growth planes simultaneously meet three conditions of parallel, coplanar and adjacent, the two growth planes can be merged into a larger plane. These relationships are detected sequentially in an under-growth correction method.
Parallel: because the error and the depth noise of the fitting equation of the growth planes need to be considered in the actual situation, when the included angle between the two growth planes is smaller than the included angle between the normal vector of the current plane and the normal vector of the maximum error estimation plane, the two growth planes are judged to be parallel.
Coplanar: and respectively calculating the fitting error of each point on the two growth planes and the detection plane, namely the fitting error from the point of the smaller growth plane to the larger growth plane in the two growth planes and the fitting error from the point of the larger growth plane to the self, and judging that the two growth planes are coplanar when the Hellinger distance of the two calculated fitting errors is less than a threshold value. The Hellinger distance is typically used to measure the similarity of two probability distributions. However, if the difference between the areas of the two growth planes is too large, even if the two growth planes are coplanar under the actual condition, a large Hellinger distance can be obtained, so that the judgment threshold value T is utilized for the two growth planes with large difference in size m To determine if they are coplanar.
Adjacent: and expanding the current growth plane by one pixel, and detecting that the edge of the current growth plane has an intersection point with the edge of the other growth plane, wherein the two growth planes are adjacent planes.
The flow of the under-growth correction step is shown in fig. 5, and includes:
(1) All the growth planes are used as detection planes, all the detection planes are arranged in a descending order according to the area size and are stored, and the detection plane S with the largest area is found out u Wherein, the area size of the detection planes after descending order is stored in a list form;
(2) Calculating the included angle theta of all two detection planes ij And will be theta ij Putting the data into an upper triangular matrix;
(3) Using the upper triangular matrix to find out all the detection planes S u Parallel detection planes. In this step, whether the included angle between the two detection planes is smaller than the detection plane S or not can be determined u Judging whether the planes are parallel by an included angle between the normal vector and the maximum error estimation plane normal vector, if so, judging that the two detection planes are parallel, otherwise, judging that the two detection planes are not parallel;
(4) Will and detect plane S u The parallel detection planes are arranged in descending order according to the area size. Specifically, in the overgrowth correction and the under-growth correction, the correction principle is to start processing from a plane with a large area, and the plane equation is more accurate due to the plane with the larger area.
(5) With S k Is represented by the formula u Any of the detection planes in parallel, the assumed plane pair S k And S u Parallel to each other, respectively calculate the detection plane S k Each pixel point is located with the detection plane S u Fitting error e of ku And a detection plane S u Fitting error e of each pixel point and the point uu ；
(6) Judging the detection plane S u Whether it is a larger plane, which refers to a plane whose size exceeds one sixth of the depth map;
(7) If the plane S is detected u Representing the two fitting errors in step (5) by histogram instead of the larger plane, and calculating the histogramThe Hellinger distance of the two curves in the figure. Judging whether the hellinger distance is smaller than a hellinger distance judgment threshold T h When it is smaller than the hellinger distance judgment threshold T h When it is, then S k And S u Coplanar, if not less than, then S k And S u Non-coplanar, the hellinger distance judging threshold T h Is a constant;
(8) If the plane S is detected u If it is a larger plane, the detection plane S is calculated k Each pixel point is located with the detection plane S u Fitting error e of ku . When the fitting error e ku Less than the judgment threshold T m Then S k And S u Coplanar, if not less than, then S k And S u Not coplanar, the determination threshold T m Is a constant number
(9) If S k And S u Coplanar, further judge S k And S u Whether adjacent, adjacent coplanar pairs may merge;
(10) If in common with S k And S u Adjacent to each other, the plane S will be detected k And the detection plane S u Merging and updating S u The plane equation of (a);
(11) If S k And S u Parallel, then S k Must also be parallel to the detection plane S u Thus will S k Parallel plane of (3) into the detection plane S u A parallel group of (a);
(12) All S u Whether all the parallel planes are detected, if so, aiming at the detection plane S u The under-growth correction step is ended, otherwise, the next parallel plane is moved, and the loop is started from the step (5) until all the parallel planes are detected.
(13) If it is directed to the detection plane S u And after the under-growth correction is finished, detecting the next detection plane according to the area size arrangement sequence of all the detection planes until all the detection planes are corrected.
When the embodiment provided by the invention detects the plane with complex texture, the depth information is used for driving the plane detection model to adopt the seed growing method for the depth map generated by the depth camera, and the post-processing is carried out on the plane after the growth of the seed block is completed, so that the plane detection of various indoor scenes and depth noise is more robust, and the accuracy of the plane detection is improved. In addition, when the adjacent planes need precise boundaries, the post-processing procedure of the enhancement mechanism can reassign the correct boundary points to the planes.
The above embodiments provided by the present invention can be applied to, for example: robust plane detection (wall surface, desktop, etc.), indoor scene reconstruction and position recognition, robot navigation system, object recognition in the computer vision field, etc.
Fig. 6 shows a robust depth information-based planar detection system provided by an embodiment of the present invention, which includes:
an extracting unit 601, configured to receive a depth map, and extract a number of valid seed blocks in the depth map;
a growing unit 602, configured to perform region growing according to the effective seed block to obtain a generating plane of the effective seed block;
a correcting unit 603, configured to perform overgrowth correction or under-growth correction on the generation plane of the effective seed block, so as to obtain an effective growth plane of the effective seed block;
an output unit 604, configured to output a depth map detection plane including the effective growth plane of the effective seed block.
Further, the extracting unit 601 is specifically configured to:
traversing a rectangular window with a preset size through the depth map in a raster scanning mode by taking each pixel point of the depth map as a center;
when each pixel point of the depth map is traversed, whether all points in the rectangular window are hole points is checked;
if no hole point exists in the rectangular window, taking the point set in the rectangular window as an effective seed block;
and calculating the plane equation and the smoothness of each effective seed block by a linear least square plane fitting method.
Wherein the fitting plane is expressed by a mathematically plane equation, the extraction unit is further configured to:
obtaining a fitting plane of the effective seed block by utilizing a linear least square plane fitting method, wherein the fitting plane is represented by a normal vector n and a distance d from the origin of a coordinate system;
calculating a fitting error of a fitting plane of the effective seed block and a point on the effective seed block;
denote any point on the ith valid seed block by p, n i Normal vector representing the fitting plane of the i-th valid seed block, d i Representing the distance of the fitted plane of the ith valid seed block from the origin of the coordinate system, e i (p) represents the fitting error of the fitting plane of the ith valid seed block to the point p on the ith valid seed block, then e i (p)＝|n i ·p+d i |；
Obtaining a root mean square fitting error according to the fitting error, and representing smoothness of the effective seed block by the root mean square fitting error;
at delta i According to the root mean square fitting error, thenWherein P represents the point set of all pixel points in the effective seed block, and P represents the number of pixel points in the effective seed block;
and sequencing the effective seed blocks according to the root mean square fitting error of the effective seed blocks from small to large.
Further, the growing unit 602 is specifically configured to:
detecting all adjacent points of the effective seed block with the highest smoothness;
if the adjacent point of the effective seed block detected currently is not a hole and does not belong to the growth plane of other effective seed blocks, substituting the adjacent point of the effective seed block detected currently into the plane equation of the effective seed block, and fitting an error formula e i (p)＝|n i ·p+d i I, calculating to obtain a corresponding fitting error;
judging whether the fitting error is smaller than a threshold value T or not, if so, merging adjacent points of the currently detected effective seed block into a currently growing growth plane of the effective seed block, and if so, regarding the adjacent points as local points of the currently growing growth plane;
updating the parameters of the equation of the effective seed block by using a linear least square plane fitting method, and updating the root mean square fitting error according to the parameters;
when the adjacent point set of the growth plane of the effective seed block is judged to be an empty set or no adjacent point of the growth plane suitable for the current growth exists in the adjacent point set, stopping growth to obtain the growth plane of the effective seed block;
wherein the threshold T is an output of a dynamic threshold function, and the dynamic threshold function is:
i denotes the depth map, I d Representing the depth map, substituting a plane equation into the depth map, calculating the depth value of a point of a fitting error by using a fitting error formula, representing tau, representing the allowable maximum roughness of a growth plane, determining the growth speed of a threshold value by lambda, representing the height and the width of the depth map by H and W respectively, representing alpha and k as constants, representing j, representing the iteration times of the growth plane in the growth process, initializing to enable j =1, outputting the maximum threshold value T from I d And (6) determining.
Further, the step of performing overgrowth correction on the generated plane of the effective seed block by the correction unit 603 includes:
taking the growth plane of the currently corrected effective seed block as a detection plane, and searching an adjacent plane of the detection plane;
finding out the intersection line of the detection plane and the currently detected adjacent plane through a parameter equation of the intersection line;
judging whether the intersection line exists in the two-dimensional range of the depth map, if not, detecting the next adjacent plane, and finding out the intersection line of the detection plane and the currently detected adjacent plane;
if the detection scanning line exists, setting the width of the detection scanning line, and scanning pixel points of the detection scanning line along the intersecting line to obtain an overgrowth area;
judging whether the detection plane grows excessively according to the overgrowth area, if so, merging the overgrowth area with the adjacent plane, and if not, merging the overgrowth area with the detection plane to obtain an effective growth plane; wherein, in the merging process, isolated blocks generated by fragmentation are detected and redistributed;
respectively updating the plane equations of the detection plane and the adjacent plane;
iterating the overgrowth correction steps until all the growth planes complete overgrowth detection and correction;
the step of performing under-growth correction on the generated plane of the effective seed block by the correction unit 603 includes:
all the growth planes are used as detection planes, all the detection planes are arranged in a descending order according to the area size and are stored, and the detection plane S with the largest area is found out u ；
Calculating the included angle theta of all two detection planes ij And all the calculated included angles theta are calculated ij Putting the obtained product into an upper triangular matrix;
finding and detecting plane S by using the upper triangular matrix u A parallel detection plane;
will and detect the plane S u The parallel detection planes are arranged in a descending order according to the area;
with S k Is represented by the formula u Calculating S for any parallel detection plane k Each pixel point and S u Fitting error e of ku And S u Fitting error e of each pixel point to itself uu ；
Judging the detection plane S u Whether it is a larger plane with an area size exceeding one sixth of the area of the depth map;
if the detection plane S u Not a larger plane, the fitting error e is determined ku And fitting error e uu Representing by a histogram, and calculating the hellinger distance of two curves in the histogram;
judging whether the hellinger distance is smaller than a hellinger distance judgment threshold T h If less than, then S k And S u Coplanar, if not less than, then S k And S u Non-coplanar, said hellinger distance judging threshold T h Is a constant;
if the detection plane S u If the plane is a large plane, the fitting error e is judged ku Whether it is less than the judgment threshold T m If less than, then S k And S u Coplanar, if not less than, then S k And S u Not coplanar, the judgment threshold T m Is a constant;
if S k And S u Coplanar, further determination of S is required k And S u Whether or not adjacent, if so, then S k And S u Merging to obtain effective growth plane and updating S u Of the plane equation of (c) while simultaneously dividing S k Parallel planes of (2) add S u A parallel group of (a);
judging the detection plane S u Whether all the parallel planes are detected is finished, if so, finishing aiming at the detection plane S u If not, detecting the next parallel plane until all parallel planes are detected;
if the plane S is detected u And after the under-growth correction is finished, detecting the next detection plane according to the area size arrangement sequence of all the detection planes until all the detection planes are corrected.
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 and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (10)
1. A robust depth information-based plane detection method is characterized by comprising the following steps:
receiving a depth map, and extracting a plurality of effective seed blocks in the depth map;
performing region growth according to the effective seed block to obtain a generation plane of the effective seed block;
performing overgrowth correction or under-growth correction on the generation plane of the effective seed block to obtain an effective growth plane of the effective seed block;
and outputting a depth map detection plane containing the effective growth plane of the effective seed block.
2. The method of claim 1, wherein the extracting a number of valid seed blocks in the depth map comprises:
traversing a rectangular window with a preset size through the depth map in a raster scanning mode by taking each pixel point of the depth map as a center;
when each pixel point of the depth map is traversed, checking whether all points in the rectangular window are hole points;
if no hole point exists in the rectangular window, taking the point set in the rectangular window as an effective seed block;
and calculating the plane equation and the smoothness of each effective seed block by a linear least square plane fitting method.
3. The plane detection method as claimed in claim 2, wherein the fitting plane is expressed by a mathematical plane equation, and the calculating the plane equation and the smoothness of each of the effective seed blocks by the linear least squares plane fitting method comprises:
obtaining a fitting plane of the effective seed block by utilizing a linear least square plane fitting method, wherein the fitting plane is represented by a normal vector n and a distance d from the origin of a coordinate system;
calculating a fitting error of a fitting plane of the effective seed block and a point on the effective seed block;
denote any point on the ith valid seed block by p, n i Normal vector representing the fitted plane of the ith valid seed block, d i Represents the distance of the fitted plane of the ith valid seed block from the origin of the coordinate system, e i (p) represents the fitting error of the fitting plane of the ith valid seed block to the point p on the ith valid seed block, then e i (p)＝|n i ·p+d i |；
Obtaining a root mean square fitting error according to the fitting error, and representing smoothness of the effective seed block by the root mean square fitting error;
at delta i According to the root mean square fitting error, thenWherein P represents the point set of all pixel points in the effective seed block, and P represents the number of pixel points in the effective seed block;
and sequencing the effective seed blocks according to the order of the root mean square fitting errors of the effective seed blocks from small to large.
4. The method of claim 3, wherein the performing region growing according to the effective seed block to obtain the generated plane of the effective seed block comprises:
detecting all adjacent points of the effective seed block with the highest smoothness;
if the adjacent point of the effective seed block detected currently is not a hole and does not belong to the growth plane of other effective seed blocks, substituting the adjacent point of the effective seed block detected currently into the plane equation of the effective seed block, and fitting an error formula e i (p)＝|n i ·p+d i Calculating to obtain a corresponding fitting error;
judging whether the fitting error is smaller than a threshold value T, if so, merging adjacent points of the currently detected effective seed block into a currently growing growth plane of the effective seed block, and if so, regarding the adjacent points as local points of the currently growing growth plane;
updating the parameters of the equation of the effective seed block by using a least square plane fitting method, and updating a root mean square fitting error according to the parameters;
when the adjacent point set of the growth plane of the effective seed block is judged to be an empty set or no adjacent point of the growth plane suitable for the current growth exists in the adjacent point set, stopping growth to obtain the growth plane of the effective seed block;
wherein the threshold T is an output of a dynamic threshold function, and the dynamic threshold function is:
i denotes the depth map, I d Representing the depth map, substituting a plane equation into the depth map, calculating the depth value of a point with fitting error by using a fitting error formula, representing the allowable maximum roughness of a growth plane by tau, determining the growth speed of the threshold value by lambda, representing the height and the width of the depth map by H and W respectively, representing the height and the width of the depth map by alpha and k as constants by j, representing the iteration times of the growth plane in the growth process, initializing to enable j =1, outputting the maximum threshold value T, and calculating the depth value of the point with fitting error by I d And (6) determining.
5. The planar inspection method as set forth in claim 1, wherein the step of performing overgrowth correction on the generated plane of the effective seed block comprises:
taking the growth plane of the currently corrected effective seed block as a detection plane, and searching an adjacent plane of the detection plane;
finding out the intersection line of the detection plane and the currently detected adjacent plane through a parameter equation of the intersection line;
judging whether the intersection line exists in the two-dimensional range of the depth map, if not, detecting the next adjacent plane, and finding out the intersection line of the detection plane and the currently detected adjacent plane;
if the detection scanning line exists, setting the width of the detection scanning line, and scanning pixel points of the detection scanning line along the intersecting line to obtain an overgrowth area;
judging whether the detection plane grows excessively according to the overgrowth area, if so, merging the overgrowth area with the adjacent plane, and if not, merging the overgrowth area with the detection plane to obtain an effective growth plane; wherein, in the merging process, isolated blocks generated by fragmentation are detected and redistributed;
respectively updating the plane equations of the detection plane and the adjacent plane;
and iterating the overgrowth correction steps until all the growth planes complete overgrowth detection and correction.
6. The planar inspection method as set forth in claim 1, wherein the step of correcting the under-growth of the generated plane of the effective seed block includes:
all the growth planes are used as detection planes, all the detection planes are arranged in a descending order according to the area size and are stored, and the detection plane S with the largest area is found out u ；
Calculating the included angle theta of all two detection planes ij And all the calculated included angles theta are calculated ij Putting the obtained product into an upper triangular matrix;
finding and detecting plane S by using the upper triangular matrix u Parallel detection planes;
will and detect plane S u The parallel detection planes are arranged in a descending order according to the area;
with S k Is represented by the formula u Calculating S for any parallel detection plane k Each pixel point and S u Fitting error e of ku And S u Fitting error e of each pixel point to itself uu ；
Judging the detection plane S u Whether it is a larger plane having an area size exceeding one sixth of the area of the depth map;
if the detection plane S u If not, the fitting error e is adjusted ku And fitting error e uu Using histogramsRepresenting and calculating the hellinger distance of two curves in the histogram;
judging whether the hellinger distance is smaller than a hellinger distance judgment threshold T h If less than, then S k And S u Coplanar, if not less than, then S k And S u Non-coplanar, said hellinger distance judging threshold T h Is a constant;
if the detection plane S u If it is a large plane, only the fitting error e is determined ku Whether or not less than the judgment threshold T m If less than, then S k And S u Coplanar, if not less than, then S k And S u Not coplanar, the judgment threshold T m Is a constant;
if S k And S u Coplanar, further determination of S is required k And S u Whether or not adjacent, if so, then S k And S u Merging to obtain effective growth plane and updating S u Will simultaneously with S k Parallel plane joining S u A parallel group of (a);
judging the detection plane S u Whether all the parallel planes are detected is finished, if so, finishing the detection of the plane S u And (4) an under-growth correcting step, if not, detecting the next parallel plane until the detection of all the parallel planes is completed.
If the plane S is detected u Detecting the next detection plane according to the area size arrangement sequence of all the detection planes until all the detection planes are corrected.
7. A robust depth information based planar detection system, comprising:
the extraction unit is used for receiving the depth map and extracting a plurality of effective seed blocks in the depth map;
a growing unit, configured to perform region growing according to the effective seed block to obtain a generating plane of the effective seed block;
a correcting unit, configured to perform overgrowth correction or under-growth correction on the generation plane of the effective seed block to obtain an effective growth plane of the effective seed block;
and the output unit is used for outputting a depth map detection plane containing the effective growth plane of the effective seed block.
8. The planar inspection system of claim 7, wherein the extraction unit is specifically configured to:
traversing a rectangular window with a preset size through the depth map in a raster scanning mode by taking each pixel point of the depth map as a center;
when each pixel point of the depth map is traversed, whether all points in the rectangular window are hole points is checked;
if no hole point exists in the rectangular window, taking the point set in the rectangular window as an effective seed block;
and calculating the plane equation and the smoothness of each effective seed block by a linear least square plane fitting method.
Wherein the fitting plane is expressed by a mathematically plane equation, the extraction unit is further configured to:
obtaining a fitting plane of the effective seed block by utilizing a linear least square plane fitting method, wherein the fitting plane is represented by a normal vector n and a distance d from the origin of a coordinate system;
calculating a fitting error of a fitting plane of the effective seed block and a point on the effective seed block;
denote any point on the ith valid seed block by p, n i Normal vector representing the fitted plane of the ith valid seed block, d i Representing the distance of the fitted plane of the ith valid seed block from the origin of the coordinate system, e i (p) represents the fitting error of the fitting plane of the ith valid seed block to the point p on the ith valid seed block, then e i (p)＝|n i ·p+d i |；
Obtaining a root mean square fitting error according to the fitting error, and representing smoothness of the effective seed block by the root mean square fitting error;
at delta i According to the root mean square simulationResultant error ofWherein P represents the point set of all pixel points in the effective seed block, and P represents the number of pixel points in the effective seed block;
and sequencing the effective seed blocks according to the order of the root mean square fitting errors of the effective seed blocks from small to large.
9. The planar inspection system of claim 8, wherein the growth unit is specifically configured to:
detecting all adjacent points of the effective seed block with the highest smoothness;
if the adjacent point of the effective seed block detected currently is not a hole and does not belong to the growth plane of other effective seed blocks, substituting the adjacent point of the effective seed block detected currently into the plane equation of the effective seed block, and fitting an error formula e i (p)＝|n i ·p+d i I, calculating to obtain a corresponding fitting error;
judging whether the fitting error is smaller than a threshold value T or not, if so, merging adjacent points of the currently detected effective seed block into a currently growing growth plane of the effective seed block, and if so, regarding the adjacent points as local points of the currently growing growth plane;
updating the parameters of the equation of the effective seed block by using a linear least square plane fitting method, and updating a root mean square fitting error according to the parameters;
when the adjacent point set of the growth plane of the effective seed block is judged to be an empty set or no adjacent point suitable for the growth plane of the current growth exists in the adjacent point set, stopping growth to obtain the growth plane of the effective seed block;
wherein the threshold T is an output of a dynamic threshold function, the dynamic threshold function being:
i denotes the depth map, I d Representing the depth map, substituting a plane equation into the depth map, calculating the depth value of a point with fitting error by using a fitting error formula, representing the allowable maximum roughness of a growth plane by tau, determining the growth speed of a threshold value by lambda, representing the height and the width of the depth map by H and W respectively, representing the height and the width of the depth map by alpha and k as constants by j, representing the iteration times of the growth plane in the growth process, initializing to enable j =1, outputting the maximum threshold value T from I d And (6) determining.
10. The planar inspection system of claim 7, wherein said step of said correction unit performing overgrowth correction on the generated plane of said valid seed block comprises:
taking the growth plane of the currently corrected effective seed block as a detection plane, and searching an adjacent plane of the detection plane;
finding out the intersection line of the detection plane and the currently detected adjacent plane through a parameter equation of the intersection line;
judging whether the intersection line exists in the two-dimensional range of the depth map, if not, detecting the next adjacent plane, and finding out the intersection line of the detection plane and the currently detected adjacent plane;
if the detection scanning line exists, setting the width of the detection scanning line, and scanning pixel points of the detection scanning line along the intersecting line to obtain an overgrowth area;
judging whether the detection plane overgrows according to the overgrowth area, if so, merging the overgrowth area with the adjacent plane, and if not, merging the overgrowth area with the detection plane to obtain an effective growth plane; wherein, in the merging process, isolated blocks generated by fragmentation are detected and redistributed;
respectively updating the plane equations of the detection plane and the adjacent plane;
iterating the overgrowth correction steps until all the growth planes complete overgrowth detection and correction;
the step of the correction unit performing under-growth correction on the generation plane of the effective seed block comprises the following steps:
all the growth planes are used as detection planes, all the detection planes are arranged in a descending order according to the area size and are stored, and the detection plane S with the largest area is found out u ；
Calculating the included angle theta of all pairwise detection planes ij And calculating all the included angles theta ij Putting the obtained product into an upper triangular matrix;
finding and detecting a plane S by using the upper triangular matrix u A parallel detection plane;
will and detect the plane S u The parallel detection planes are arranged in a descending order according to the area;
with S k Is represented by u Calculating S for any parallel detection plane k Each pixel point and S u Fitting error e of ku And S u Fitting error e of each pixel point to itself uu ；
Judging the detection plane S u Whether it is a larger plane having an area size exceeding one sixth of the area of the depth map;
if the detection plane S u Not a larger plane, the fitting error e is determined ku And fitting error e uu Representing by a histogram, and calculating the hellinger distance of two curves in the histogram;
judging whether the hellinger distance is smaller than a hellinger distance judgment threshold T h If less than, then S k And S u Coplanar, if not less than, then S k And S u Not coplanar, the hellinger distance judging threshold T h Is a constant;
if the detection plane S u If it is a large plane, only the fitting error e is determined ku Whether or not less than the judgment threshold T m If less than, then S k And S u Coplanar, if not less than, then S k And S u Not coplanar, the judgment threshold T m Is a constant;
if S k And S u Coplanar, then need to be furtherJudgment S k And S u Whether or not adjacent, if so, then S k And S u Merging to obtain effective growth plane and updating S u Of S, while dividing S k Parallel planes of (2) add S u A parallel group of (a);
judging the detection plane S u Whether all the parallel planes are detected is finished, if so, finishing aiming at the detection plane S u If not, detecting the next parallel plane until all parallel planes are detected;
if the plane S is detected u And after the under-growth correction is finished, detecting the next detection plane according to the area size arrangement sequence of all the detection planes until all the detection planes are corrected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710866367.5A CN107610174B (en) | 2017-09-22 | 2017-09-22 | Robust depth information-based plane detection method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710866367.5A CN107610174B (en) | 2017-09-22 | 2017-09-22 | Robust depth information-based plane detection method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107610174A true CN107610174A (en) | 2018-01-19 |
| CN107610174B CN107610174B (en) | 2021-02-05 |
Family
ID=61061944
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710866367.5A Active CN107610174B (en) | 2017-09-22 | 2017-09-22 | Robust depth information-based plane detection method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107610174B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109214348A (en) * | 2018-09-19 | 2019-01-15 | 北京极智嘉科技有限公司 | A kind of obstacle detection method, device, equipment and storage medium |
| CN109359614A (en) * | 2018-10-30 | 2019-02-19 | 百度在线网络技术（北京）有限公司 | A kind of plane recognition methods, device, equipment and the medium of laser point cloud |
| CN111595328A (en) * | 2020-06-01 | 2020-08-28 | 四川阿泰因机器人智能装备有限公司 | Real obstacle map construction and navigation method and system based on depth camera |
| JP2020528134A (en) * | 2018-06-25 | 2020-09-17 | ベイジン ディディ インフィニティ テクノロジー アンド ディベロップメント カンパニー リミティッド | Calibration of integrated sensor in natural scene |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120196679A1 (en) * | 2011-01-31 | 2012-08-02 | Microsoft Corporation | Real-Time Camera Tracking Using Depth Maps |
| US8831336B2 (en) * | 2011-11-11 | 2014-09-09 | Texas Instruments Incorporated | Method, system and computer program product for detecting an object in response to depth information |
| CN104424660A (en) * | 2013-09-04 | 2015-03-18 | 国际商业机器公司 | Efficient visual surface finding method and system |
| CN105096259A (en) * | 2014-05-09 | 2015-11-25 | 株式会社理光 | Depth value restoration method and system for depth image |
| CN105359187A (en) * | 2013-06-11 | 2016-02-24 | 微软技术许可有限责任公司 | High-performance plane detection with depth camera data |
| WO2017023456A1 (en) * | 2015-08-05 | 2017-02-09 | Intel Corporation | Method and system of planar surface detection for image processing |
-
2017
- 2017-09-22 CN CN201710866367.5A patent/CN107610174B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20120196679A1 (en) * | 2011-01-31 | 2012-08-02 | Microsoft Corporation | Real-Time Camera Tracking Using Depth Maps |
| US8831336B2 (en) * | 2011-11-11 | 2014-09-09 | Texas Instruments Incorporated | Method, system and computer program product for detecting an object in response to depth information |
| CN105359187A (en) * | 2013-06-11 | 2016-02-24 | 微软技术许可有限责任公司 | High-performance plane detection with depth camera data |
| CN104424660A (en) * | 2013-09-04 | 2015-03-18 | 国际商业机器公司 | Efficient visual surface finding method and system |
| CN105096259A (en) * | 2014-05-09 | 2015-11-25 | 株式会社理光 | Depth value restoration method and system for depth image |
| WO2017023456A1 (en) * | 2015-08-05 | 2017-02-09 | Intel Corporation | Method and system of planar surface detection for image processing |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2020528134A (en) * | 2018-06-25 | 2020-09-17 | ベイジン ディディ インフィニティ テクノロジー アンド ディベロップメント カンパニー リミティッド | Calibration of integrated sensor in natural scene |
| CN109214348A (en) * | 2018-09-19 | 2019-01-15 | 北京极智嘉科技有限公司 | A kind of obstacle detection method, device, equipment and storage medium |
| CN109359614A (en) * | 2018-10-30 | 2019-02-19 | 百度在线网络技术（北京）有限公司 | A kind of plane recognition methods, device, equipment and the medium of laser point cloud |
| CN111595328A (en) * | 2020-06-01 | 2020-08-28 | 四川阿泰因机器人智能装备有限公司 | Real obstacle map construction and navigation method and system based on depth camera |
| CN111595328B (en) * | 2020-06-01 | 2023-04-25 | 四川阿泰因机器人智能装备有限公司 | Real obstacle map construction and navigation method and system based on depth camera |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107610174B (en) | 2021-02-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110443836B (en) | Point cloud data automatic registration method and device based on plane features | |
| CN107610174B (en) | Robust depth information-based plane detection method and system | |
| Oehler et al. | Efficient multi-resolution plane segmentation of 3D point clouds | |
| CN110807781B (en) | Point cloud simplifying method for retaining details and boundary characteristics | |
| CN108830888B (en) | Coarse matching method based on improved multi-scale covariance matrix characteristic descriptor | |
| CN104040590A (en) | Method for estimating pose of object | |
| CN103727930A (en) | Edge-matching-based relative pose calibration method of laser range finder and camera | |
| CN108647580B (en) | Improved SIFT-based ISAR image feature point extraction and matching method | |
| CN107240130B (en) | Remote sensing image registration method, device and system | |
| CN116740072B (en) | Road surface defect detection method and system based on machine vision | |
| US8270730B2 (en) | Target orientation | |
| CN112819883A (en) | Rule object detection and positioning method | |
| CN113487722A (en) | Automatic concrete member detection method based on three-dimensional laser scanning method | |
| CN111783722B (en) | Lane line extraction method of laser point cloud and electronic equipment | |
| CN110110687A (en) | Fruit automatic identifying method on tree based on colouring information and three-D profile information | |
| CN114387329B (en) | Building contour progressive regularization method based on high-resolution remote sensing image | |
| CN115239661A (en) | Mechanical part burr detection method and system based on image processing | |
| Wang et al. | Road detection based on illuminant invariance and quadratic estimation | |
| CN103714550A (en) | Image registration automatic optimization algorithm based on matching of curve characteristic evaluation | |
| WO2018131163A1 (en) | Information processing device, database generation device, method, and program, and storage medium | |
| CN107993193B (en) | Tunnel lining image splicing method based on illumination equalization and surf algorithm improvement | |
| CN107765257A (en) | A kind of laser acquisition and measuring method based on the calibration of reflected intensity accessory external | |
| CN115205317B (en) | Bridge monitoring photoelectric target image light spot center point extraction method | |
| CN116721156A (en) | Workpiece position positioning method, device, computer equipment and storage medium | |
| Novacheva | Building roof reconstruction from LiDAR data and aerial images through plane extraction and colour edge detection |
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 |