CN108428250A - A kind of X angular-point detection methods applied to vision positioning and calibration - Google Patents
A kind of X angular-point detection methods applied to vision positioning and calibration Download PDFInfo
- Publication number
- CN108428250A CN108428250A CN201810077053.1A CN201810077053A CN108428250A CN 108428250 A CN108428250 A CN 108428250A CN 201810077053 A CN201810077053 A CN 201810077053A CN 108428250 A CN108428250 A CN 108428250A
- Authority
- CN
- China
- Prior art keywords
- corner point
- pixel
- value
- point
- sample data
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 23
- 238000005070 sampling Methods 0.000 claims abstract description 8
- 230000000007 visual effect Effects 0.000 claims description 23
- 238000004364 calculation method Methods 0.000 description 5
- 238000013178 mathematical model Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10016—Video; Image sequence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30204—Marker
- G06T2207/30208—Marker matrix
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Geometry (AREA)
- Image Analysis (AREA)
Abstract
Description
技术领域technical field
本发明涉及一种应用于视觉定位和标定的X角点检测方法，属于计算机视觉应用的技术领域。The invention relates to an X corner point detection method applied to visual positioning and calibration, and belongs to the technical field of computer vision applications.
背景技术Background technique
视觉定位和标定是三维计算机视觉的重要组成部分。计算机视觉的基本任务之一是从摄像机获取的图像信息出发计算三维空间中物体的几何信息，并由此重建和识别物体，而空间物体表面某点的三维几何位置与其在图像中对应点之间的相互关系是由摄像机成像的几何模型决定的，这些几何模型参数就是摄像机参数。在大多数条件下，这些参数必须通过实验与计算才能得到，这个过程被称为视觉标定。标定过程就是确定摄像机的几何和光学参数，摄像机相对于世界坐标系的方位；视觉定位过程就是根据视觉标定的参数通过二维图像信息计算物体的三维信息。标定精度的大小，直接影响着计算机视觉的定位精度。Visual localization and calibration are important components of 3D computer vision. One of the basic tasks of computer vision is to calculate the geometric information of the object in the three-dimensional space from the image information obtained by the camera, and then reconstruct and recognize the object, and the three-dimensional geometric position of a point on the surface of the space object and its corresponding point in the image The interrelationship of is determined by the geometric model of camera imaging, and these geometric model parameters are camera parameters. Under most conditions, these parameters must be obtained through experiments and calculations. This process is called visual calibration. The calibration process is to determine the geometric and optical parameters of the camera, and the orientation of the camera relative to the world coordinate system; the visual positioning process is to calculate the three-dimensional information of the object through the two-dimensional image information according to the visual calibration parameters. The size of the calibration accuracy directly affects the positioning accuracy of computer vision.
平面标定法是一种常用的摄像机视觉标定方法，是借助已知的棋盘格标定板，即标定板的尺寸和形状已知，利用标定板上的X角点与拍摄其所获得图像上的对应点之间的对应关系建立数学模型，用此数学模型来标定摄像机内外参数。棋盘格标定板由于制作经济、简单而被广泛应用于摄像机标定中；另外，带有X角点视觉标志的光学定位系统应用也比较广泛。The plane calibration method is a commonly used camera visual calibration method. It uses a known checkerboard calibration board, that is, the size and shape of the calibration board is known, and uses the X corner point on the calibration board to correspond to the image obtained by shooting it. The corresponding relationship between the points establishes a mathematical model, and uses this mathematical model to calibrate the internal and external parameters of the camera. The checkerboard calibration board is widely used in camera calibration because of its economical and simple manufacture; in addition, the optical positioning system with the X corner visual mark is also widely used.
对于X角点的检测，目前已经有一些方法提出，常用的有基于Harris角点检测的方法、基于Hessian矩阵的检测方法以及基于改进Susan角点检测的方法。已经提出的方法主要是通过各种不同方式的特征计算来判断X角点的强弱，算法运算量大，不适合并行批量处理。For the detection of X corners, some methods have been proposed at present, and the commonly used methods are based on Harris corner detection, Hessian matrix-based detection methods and improved Susan corner detection methods. The proposed method is mainly to judge the strength of the X corner point through various feature calculations. The algorithm has a large amount of calculation and is not suitable for parallel batch processing.
发明内容Contents of the invention
针对现有技术的不足，本发明提供了一种应用于视觉定位和标定的X角点检测方法；Aiming at the deficiencies of the prior art, the present invention provides an X corner point detection method applied to visual positioning and calibration;
本发明提高了角点检测算法的运算速度、抗干扰力及准确度。The invention improves the operation speed, anti-interference ability and accuracy of the corner point detection algorithm.
术语解释：Explanation of terms:
X角点：视觉标定用的国际象棋盘是由黑白颜色突变区域组合而成，其中相邻黑白棋盘格交界的临界点，即为X角点。X corner point: The chessboard used for visual calibration is composed of black and white color mutation areas, and the critical point at the junction of adjacent black and white checkerboard grids is the X corner point.
本发明的技术方案为：Technical scheme of the present invention is:
一种应用于视觉定位和标定的X角点检测方法，包括：An X-corner detection method applied to visual positioning and calibration, comprising:
S1：采集图像，采用回形窗口对图像进行采样；设定该回形窗口取样的边长为2r个像素点，该回形窗口为正方形，则该回形窗口所取样本共含有8r-4个像素点，r小于图像中最小的X角点边长的一半；将回形窗口的所有像素点计入一个环形数据队列，回形窗口的所有像素点即样本数据，记第i个像素点为Pi，Pi的灰度值为fi，i＝1,2...(8r-4)；S1: Collect images, use the paper window to sample the image; set the side length of the paper window sampling to 2r pixels, and the paper window is a square, then the samples taken by the paper window contain a total of 8r-4 pixels, r is less than half of the side length of the smallest X corner point in the image; all the pixels of the paper window are included in a circular data queue, and all the pixels of the paper window are sample data, and the ith pixel is recorded is P i , the gray value of P i is f i , i=1,2...(8r-4);
S2：根据X角点的图像特征，初步判断样本数据中是否包含X角点，如果满足判断条件，则计算出X角点的亚像素级位置，否则，进入步骤S5；S2: According to the image characteristics of the X corner point, preliminarily judge whether the sample data contains the X corner point, if the judgment condition is met, calculate the sub-pixel position of the X corner point, otherwise, go to step S5;
S3：根据步骤S2得到的X角点的亚像素级位置，进一步判断样本数据是否包含X角点，并排除重复判断的X角点；S3: According to the sub-pixel position of the X corner point obtained in step S2, further judge whether the sample data contains the X corner point, and exclude the repeatedly judged X corner point;
S4：以X角点作为回形窗口中心，重新获取样本数据，并判断数据是否满足X角点对称性条件，满足则用曲线拟合的方法计算出X角点的亚像素级位置，设置X角点重复检测标志；S4: Take the X corner point as the center of the circular window, reacquire the sample data, and judge whether the data meets the symmetry condition of the X corner point. If it is satisfied, use the curve fitting method to calculate the sub-pixel position of the X corner point, and set X Corner duplicate detection flag;
S5：使回形窗口在图像上移动获取新的样本数据，每次间隔n个像素，n∈(1,2r)，重复步骤S2到S4，检测出所有的X角点。S5: Make the paper window move on the image to obtain new sample data, each interval is n pixels, n∈(1,2r), repeat steps S2 to S4, and detect all X corner points.
根据本发明优选的，n＝r。Preferably according to the invention, n=r.
根据本发明优选的，所述步骤S2，包括：Preferably according to the present invention, said step S2 includes:
S21：依次对样本数据进行灰度化；阈值可以自适应选取。S21: Grayscale the sample data in turn; the threshold can be selected adaptively.
S22：将样本数据的灰度值进行两次二值化，计算步骤S21处理后的样本数据的阶跃次数Ns，如果Ns＝4，则执行步骤S23，否则，执行步骤S5；S22: Binarize the gray value of the sample data twice, calculate the number of steps N s of the sample data processed in step S21, if N s =4, execute step S23, otherwise, execute step S5;
S23：以样本数据灰度值的均值作为阈值，对样本数据的灰度值二值化；设定步骤S22计算得到的样本数据灰度值产生阶跃时的像素为台阶A、台阶B、台阶C、台阶D，计算这四个像素的索引值之间的距离LAB、LBC、LCD、LDA，如果LAB、LBC、LCD、LDA均小于max_T且LAB、LBC、LCD、LDA均大于min_T，max_T∈(10,15),min_T∈(5,10)，则初步判断样本数据中包含X角点，继续执行步骤S24，否则，执行步骤S5；S23: Binarize the gray value of the sample data by using the average value of the gray value of the sample data as the threshold; set the pixels when the gray value of the sample data calculated in step S22 produces a step as step A, step B, step C. Step D, calculate the distance L AB , L BC , L CD , L DA between the index values of these four pixels, if L AB , L BC , LCD , L DA are all less than max_T and L AB , L BC , L CD , L DA are all greater than min_T, max_T∈(10,15), min_T∈(5,10), then it is preliminarily judged that the sample data contains the X corner point, and continue to step S24, otherwise, go to step S5;
S24：由摄影几何及对称性原理，计算X角点的亚像素级位置L，即直线AC和BD的交点。计算公式为L＝AC×BD。S24: Calculate the sub-pixel position L of the X corner point, that is, the intersection of the straight lines AC and BD, based on the principle of photographic geometry and symmetry. The calculation formula is L=AC×BD.
步骤S2中计算得到的X角点位置是亚像素级的(精确到小数点后一位)，位置精度比较高。The position of the X corner point calculated in step S2 is at the sub-pixel level (accurate to one decimal place), and the position accuracy is relatively high.
根据本发明优选的，所述步骤S22中，二值化阈值为mean±Δ，mean为样本数据灰度值的均值，Δ为阈值调节值，Δ的取值范围为20—160像素。Δ的值与整个图像的亮度有关，加上Δ作为阈值调节值，可以避免由于图像噪声的影响造成的错误判断，增强了算法的鲁棒性。Preferably according to the present invention, in the step S22, the binarization threshold is mean±Δ, mean is the mean value of the gray value of the sample data, Δ is the threshold adjustment value, and the value range of Δ is 20-160 pixels. The value of Δ is related to the brightness of the entire image, adding Δ as the threshold adjustment value can avoid wrong judgments caused by the influence of image noise and enhance the robustness of the algorithm.
根据本发明优选的，所述步骤S24，包括：Preferably according to the present invention, said step S24 includes:
以台阶A、台阶B、台阶C、台阶D的像素值作为点A、B、C、D的坐标值，求得它们的三维齐次坐标，点A和点C的齐次坐标叉乘得到直线AC的齐次方程的矢量表示形式，点B和点D的齐次坐标叉乘得到直线BD的齐次方程的矢量表示形式，将表示直线AC的齐次方程的矢量和直线BD的齐次方程的矢量叉乘，得到直线AC和BD交点的齐次坐标L1，设L1的坐标为(x1,x2,x3)，则点(x1/x3,x2/x3)即为交点的二维坐标，取整后即得X角点的像素值L(亚像素级位置L)。Take the pixel values of step A, step B, step C, and step D as the coordinate values of points A, B, C, and D to obtain their three-dimensional homogeneous coordinates. The homogeneous coordinates of point A and point C are cross-multiplied to obtain a straight line The vector representation of the homogeneous equation of AC, the homogeneous coordinates of point B and point D are cross-multiplied to obtain the vector representation of the homogeneous equation of straight line BD, the vector representation of the homogeneous equation of straight line AC and the homogeneous equation of straight line BD The vector cross product of the straight line AC and BD obtains the homogeneous coordinate L1 of the intersection point of the straight line AC and BD. Let the coordinates of L1 be (x1, x2, x3), then the point (x1/x3, x2/x3) is the two-dimensional coordinate of the intersection point, take After the adjustment, the pixel value L (sub-pixel level position L) of the X corner point is obtained.
根据本发明优选的，所述步骤S3，包括：Preferably according to the present invention, said step S3 includes:
S31：判断X角点重复检测标志，如果步骤S23得到的X角点的像素值L位于不活跃区，则判定该X角点已经被检测出，则跳出本次循环，执行步骤S5；否则，执行步骤S32；S31: Determine the repeated detection flag of the X corner point, if the pixel value L of the X corner point obtained in step S23 is located in the inactive area, then it is determined that the X corner point has been detected, then jump out of this loop, and execute step S5; otherwise, Execute step S32;
S32：获取X角点的像素值L邻域像素的灰度值，所述邻域是指以X角点的像素值L为中心、以r像素为半径的范围；以该邻域灰度值的均值作为阈值将该邻域二值化，计算灰度值的阶跃次数ΔVC，如果ΔVC>min_V，继续执行步骤S4，否则，执行步骤S5；min_V＝4。S32: Obtain the grayscale value of the pixel value L of the X corner point neighborhood pixel, the neighborhood refers to the range with the pixel value L of the X corner point as the center and r pixels as the radius; the neighborhood grayscale value The mean value of is used as the threshold to binarize the neighborhood, and calculate the number of grayscale steps ΔV C , if ΔV C >min_V, proceed to step S4, otherwise, proceed to step S5; min_V=4.
根据本发明优选的，所述步骤S4，具体包括：Preferably according to the present invention, said step S4 specifically includes:
S41：以X角点的像素值L作为回形窗口的中心，重新获取样本序列P′；S41: Take the pixel value L of the X corner point as the center of the circular window, and reacquire the sample sequence P';
S42：以灰度值的均值作为阈值，将样本序列P＇的灰度值二值化，记灰度值二值化产生阶跃时的像素为台阶A1、台阶B1、台阶C1、台阶D1，计算这四个像素索引值之间的距离L′A1B1、L＇B1C1、L＇C1D1、L＇D1A1，如果L＇A1B1＝L＇C1D1且L＇B1C1＝L′D1A1，继续执行步骤S42，否则，执行步骤S5；S42: Using the mean value of the gray value as the threshold value, binarize the gray value of the sample sequence P', record the pixels when the gray value binarization produces a step as step A1, step B1, step C1, and step D1, Calculate the distance L' A1B1 , L' B1C1 , L' C1D1 , L' D1A1 between these four pixel index values, if L' A1B1 =L' C1D1 and L' B1C1 =L' D1A1 , continue to step S42, otherwise , execute step S5;
S43：用曲线拟合的方法求出台阶A1、台阶B1、台阶C1、台阶D1的一维亚像素位置A′、B′、C′、D′；S43: Calculate the one-dimensional pixel positions A', B', C', and D' of the steps A1, B1, C1, and D1 by curve fitting;
S44：根据步骤S43求出的台阶A1、台阶B1、台阶C1、台阶D1的一维亚像素位置A′、B＇、C＇、D＇，以及步骤S32求出的X角点的像素值L，求出台阶A1、台阶B1、台阶C1、台阶D1的二维亚像素位置A′、B′、C′、D′；S44: According to the one-dimensional pixel positions A', B', C', D' of the steps A1, B1, C1, and D1 obtained in step S43, and the pixel value L of the X corner point obtained in step S32 , find the two-dimensional sub-pixel positions A', B', C', D' of the steps A1, B1, C1, and D1;
假设某台阶的一维亚像素位置为m，其对应X角点中心的像素为(x,y)，求出台阶A1、台阶B1、台阶C1、台阶D1的二维亚像素位置；台阶A1的二维亚像素位置为(x+A′-r+1,y-r+0.5)，台阶B1的二维亚像素位置为(x+r+0.5,y+B′-3r+1)，台阶C1的二维亚像素位置为(x-C′+5r-1,y+r+0.5)，台阶D1的二维亚像素位置为(x-r+0.5,y-D′+7r-1)；Assuming that the one-dimensional pixel position of a step is m, and the pixel corresponding to the center of the X corner point is (x, y), calculate the two-dimensional sub-pixel positions of steps A1, B1, C1, and D1; The two-dimensional sub-pixel position of B1 is (x+A'-r+1, y-r+0.5), and the two-dimensional sub-pixel position of step B1 is (x+r+0.5, y+B'-3r+1). The two-dimensional sub-pixel position of C1 is (x-C'+5r-1, y+r+0.5), and the two-dimensional sub-pixel position of step D1 is (x-r+0.5, y-D'+7r-1);
S45：按照步骤S32的方法，计算直线A＇C＇和B＇D＇的交点坐标，即X角点的像素值L的亚像素位置；S45: According to the method of step S32, calculate the intersection point coordinates of the straight lines A'C' and B'D', that is, the sub-pixel position of the pixel value L of the X corner point;
S46：计算X角点的方向信息：按逆时针方向，根据黑白变化序列得到两条边界线，包括BW(Black-to-White)线、WB(White-to-Black)线，BW线是指从黑到白跳变的边界线；WB线是指从白到黑跳变的边界线；求取BW线、WB线与水平方向的夹角θ1、θ2，即X角点的方向信息；S46: Calculate the direction information of the X corner point: in the counterclockwise direction, two boundary lines are obtained according to the sequence of black and white changes, including the BW (Black-to-White) line and the WB (White-to-Black) line. The BW line refers to The boundary line that jumps from black to white; the WB line refers to the boundary line that jumps from white to black; calculate the angles θ 1 and θ 2 between the BW line, WB line and the horizontal direction, that is, the direction information of the X corner point ;
S47：将该X角点的像素值L的邻域设为不活跃区，表示该X角点已被检测出。避免X角点重复检测。S47: Set the neighborhood of the pixel value L of the X corner point as an inactive area, indicating that the X corner point has been detected. Avoid repeated detection of X corners.
根据本发明优选的，所述步骤S43，用曲线拟合的方法求出台阶A1、台阶B1、台阶C1、台阶D1的一维亚像素位置A＇、B＇、C＇、D＇，包括：取样本序列P＇台阶A1附近的五个像素(台阶A1前面三个，台阶A1后面两个)，以这五个像素在样本序列P′中的索引值为x坐标，其灰度值的梯度为y坐标，进行二次曲线拟合，拟合出的曲线近似为二次抛物线，二次抛物线的极值点即为沿该梯度方向灰度变化最大的地方，即为台阶A1的一维亚像素位置A′；以同样的方法分别求出台阶B1、台阶C1、台阶D1的一维亚像素位置B′、C′、D′。Preferably according to the present invention, the step S43 is to obtain the one-dimensional pixel positions A', B', C', and D' of the step A1, the step B1, the step C1, and the step D1 by using a curve fitting method, including: Take five pixels near the step A1 of the sample sequence P' (three in front of the step A1 and two behind the step A1), and take the index value of these five pixels in the sample sequence P' as the x coordinate, and the gradient of the gray value is the y coordinate, and the quadratic curve fitting is performed, and the fitted curve is approximately a quadratic parabola, and the extreme point of the quadratic parabola is the place where the gray value changes the most along the gradient direction, that is, one Via of the step A1 Pixel position A'; the one-dimensional pixel positions B', C', and D' of steps B1, C1, and D1 are calculated in the same way.
本发明的有益效果为：The beneficial effects of the present invention are:
1、本发明对图像采样时，每次间隔采样窗口边长的一半，提高了检测速度，且不会出现遗漏X角点的情况。假设回形窗口半径为r＝10，那么采用本发明方法的检测速度将是现有技术中逐个像素扫描方法的10倍；以640*480分辨率图像为例，如果逐个像素扫描方法的检测速度为3帧/秒,那么本发明的检测速度将是30帧/秒，可以达到实时检测的目的。1. When the present invention samples the image, each interval is half of the side length of the sampling window, which improves the detection speed and does not miss the X corner point. Assuming that the radius of the paper-shaped window is r=10, the detection speed of the method of the present invention will be 10 times that of the pixel-by-pixel scanning method in the prior art; If it is 3 frames/second, then the detection speed of the present invention will be 30 frames/second, which can achieve the purpose of real-time detection.
2、本发明基于X角点的图像特征判断采样窗口内是否含有X角点，增强了算法的鲁棒性。2. The present invention judges whether the sampling window contains the X corner point based on the image feature of the X corner point, which enhances the robustness of the algorithm.
附图说明Description of drawings
图1为X角点以及检测使用的回形窗口的示意图。FIG. 1 is a schematic diagram of an X corner point and a paper window used for detection.
图2为本发明应用于视觉定位和标定的X角点检测方法的流程示意图。Fig. 2 is a schematic flowchart of the X corner point detection method applied to visual positioning and calibration according to the present invention.
具体实施方式Detailed ways
下面结合说明书附图和实施例对本发明作进一步限定，但不限于此。The present invention will be further limited below in conjunction with the accompanying drawings and embodiments, but not limited thereto.
实施例1Example 1
一种应用于视觉定位和标定的X角点检测方法，如图2所示，包括：An X corner point detection method applied to visual positioning and calibration, as shown in Figure 2, comprising:
S1：采集图像，采用回形窗口(如图1所示)对图像进行采样；设定该回形窗口取样的边长为2r个像素点，该回形窗口为正方形，则该回形窗口所取样本共含有8r-4个像素点，r小于图像中最小的X角点边长的一半；将回形窗口的所有像素点计入一个环形数据队列，回形窗口的所有像素点即样本数据P，记第i个像素点为Pi，Pi的灰度值为fi，i＝1,2...(8r-4)；S1: collect images, adopt the shape window (as shown in Figure 1) to sample the image; set the side length of the shape window sampling as 2r pixels, the shape window is a square, then the shape of the shape window Take a sample containing 8r-4 pixels in total, and r is less than half of the side length of the smallest X corner point in the image; count all the pixels of the paper window into a ring data queue, and all the pixels of the paper window are the sample data P, record the i-th pixel as P i , the gray value of P i is f i , i=1,2...(8r-4);
S2：根据X角点的图像特征，初步判断样本数据P中是否包含X角点，如果满足判断条件，则计算出X角点的亚像素级位置，否则，进入步骤S5；S2: According to the image characteristics of the X corner point, preliminarily judge whether the sample data P contains the X corner point, if the judgment condition is met, calculate the sub-pixel position of the X corner point, otherwise, go to step S5;
S3：根据步骤S2得到的X角点的亚像素级位置，进一步判断样本数据是否包含X角点，并排除重复判断的X角点；S3: According to the sub-pixel position of the X corner point obtained in step S2, further judge whether the sample data contains the X corner point, and exclude the repeatedly judged X corner point;
S4：以X角点作为回形窗口中心，重新获取样本数据，并判断数据是否满足X角点对称性条件，满足则用曲线拟合的方法计算出X角点的亚像素级位置，设置X角点重复检测标志；S4: Take the X corner point as the center of the circular window, reacquire the sample data, and judge whether the data meets the symmetry condition of the X corner point. If it is satisfied, use the curve fitting method to calculate the sub-pixel position of the X corner point, and set X Corner duplicate detection flag;
S5：使回形窗口在图像上移动获取新的样本数据，每次间隔n个像素，n∈(1,2r)，重复步骤S2到S4，检测出所有的X角点。n＝r。S5: Make the paper window move on the image to obtain new sample data, each interval is n pixels, n∈(1,2r), repeat steps S2 to S4, and detect all X corner points. n=r.
实施例2Example 2
根据实施例1所述的一种应用于视觉定位和标定的X角点检测方法，其区别在于，所述步骤S2，包括：According to an X corner point detection method applied to visual positioning and calibration described in Embodiment 1, the difference is that the step S2 includes:
S21：依次对样本数据进行灰度化；阈值可以自适应选取。S21: Grayscale the sample data in turn; the threshold can be selected adaptively.
S22：将样本数据的灰度值进行两次二值化，二值化阈值为mean±Δ，mean为样本数据灰度值的均值，Δ为阈值调节值，Δ的取值范围为20—160像素。Δ的值与整个图像的亮度有关，加上Δ作为阈值调节值，可以避免由于图像噪声的影响造成的错误判断，增强了算法的鲁棒性。计算步骤S21处理后的样本数据的阶跃次数Ns，如果Ns＝4，则执行步骤S23，否则，执行步骤S5；S22: Binarize the gray value of the sample data twice, the binarization threshold is mean±Δ, mean is the mean value of the gray value of the sample data, Δ is the threshold adjustment value, and the value range of Δ is 20-160 pixels. The value of Δ is related to the brightness of the entire image, adding Δ as the threshold adjustment value can avoid wrong judgments caused by the influence of image noise and enhance the robustness of the algorithm. Calculate the number of steps N s of the sample data processed in step S21, if N s =4, then execute step S23, otherwise, execute step S5;
S23：以样本数据灰度值的均值作为阈值，对样本数据的灰度值二值化；设定步骤S22计算得到的样本数据灰度值产生阶跃时的像素为台阶A、台阶B、台阶C、台阶D，计算这四个像素的索引值之间的距离LAB、LBC、LCD、LDA，如果LAB、LBC、LCD、LDA均小于max_T且LAB、LDC、LCD、LDA均大于min_T，max_T∈(10,15),min_T∈(5,10)，则初步判断样本数据中包含X角点，继续执行步骤S24，否则，执行步骤S5；S23: Binarize the gray value of the sample data by using the average value of the gray value of the sample data as the threshold; set the pixels when the gray value of the sample data calculated in step S22 produces a step as step A, step B, step C. Step D, calculate the distance L AB , L BC , L CD , L DA between the index values of these four pixels, if L AB , L BC , LCD , L DA are all less than max_T and L AB , L DC , L CD , L DA are all greater than min_T, max_T∈(10,15), min_T∈(5,10), then it is preliminarily judged that the sample data contains the X corner point, and continue to step S24, otherwise, go to step S5;
S24：由摄影几何及对称性原理，计算X角点的亚像素级位置L，即直线AC和BD的交点。计算公式为L＝AC×BD。包括：以台阶A、台阶B、台阶C、台阶D的像素值作为点A、B、C、D的坐标值，求得它们的三维齐次坐标，点A和点C的齐次坐标叉乘得到直线AC的齐次方程的矢量表示形式，点B和点D的齐次坐标叉乘得到直线BD的齐次方程的矢量表示形式，将表示直线AC的齐次方程的矢量和直线BD的齐次方程的矢量叉乘，得到直线AC和BD交点的齐次坐标L1，设L1的坐标为(x1,x2,x3)，则点(x1/x3,x2/x3)即为交点的二维坐标，取整后即得X角点的像素值L。步骤S2中计算得到的X角点位置是亚像素级的(精确到小数点后一位)，位置精度比较高。S24: Calculate the sub-pixel position L of the X corner point, that is, the intersection of the straight lines AC and BD, based on the principle of photographic geometry and symmetry. The calculation formula is L=AC×BD. Including: take the pixel values of steps A, B, C, and D as the coordinate values of points A, B, C, and D to obtain their three-dimensional homogeneous coordinates, and cross product the homogeneous coordinates of points A and C Obtain the vector representation of the homogeneous equation of the straight line AC, and obtain the vector representation of the homogeneous equation of the straight line BD by cross multiplying the homogeneous coordinates of point B and point D. The vector cross product of the sub-equation can get the homogeneous coordinate L1 of the intersection point of the straight line AC and BD. Let the coordinates of L1 be (x1, x2, x3), then the point (x1/x3, x2/x3) is the two-dimensional coordinate of the intersection point , and get the pixel value L of the X corner point after rounding. The position of the X corner point calculated in step S2 is at the sub-pixel level (accurate to one decimal place), and the position accuracy is relatively high.
实施例3Example 3
根据实施例1所述的一种应用于视觉定位和标定的X角点检测方法，其区别在于，所述步骤S3，包括：According to an X corner point detection method applied to visual positioning and calibration described in Embodiment 1, the difference is that the step S3 includes:
S31：判断X角点重复检测标志，如果步骤S23得到的X角点的像素值L位于不活跃区，则判定该X角点已经被检测出，则跳出本次循环，执行步骤S5；否则，执行步骤S32；S31: Determine the repeated detection flag of the X corner point, if the pixel value L of the X corner point obtained in step S23 is located in the inactive area, then it is determined that the X corner point has been detected, then jump out of this loop, and execute step S5; otherwise, Execute step S32;
S32：获取X角点的像素值L邻域像素的灰度值，所述邻域是指以X角点的像素值L为中心、以r像素为半径的范围；以该邻域灰度值的均值作为阈值将该邻域二值化，计算灰度值的阶跃次数ΔVC，如果ΔVC>min_V，继续执行步骤S4，否则，执行步骤S5；min_V＝4。S32: Obtain the grayscale value of the pixel value L of the X corner point neighborhood pixel, the neighborhood refers to the range with the pixel value L of the X corner point as the center and r pixels as the radius; the neighborhood grayscale value The mean value of is used as the threshold to binarize the neighborhood, and calculate the number of grayscale steps ΔV C , if ΔV C >min_V, proceed to step S4, otherwise, proceed to step S5; min_V=4.
实施例4Example 4
根据实施例1所述的一种应用于视觉定位和标定的X角点检测方法，其区别在于，所述步骤S4，包括：According to an X corner point detection method applied to visual positioning and calibration described in Embodiment 1, the difference is that the step S4 includes:
S41：以X角点的像素值L作为回形窗口的中心，重新获取样本序列P′；S41: Take the pixel value L of the X corner point as the center of the circular window, and reacquire the sample sequence P';
S42：以灰度值的均值作为阈值，将样本序列P′的灰度值二值化，记灰度值二值化产生阶跃时的像素为台阶A1、台阶B1、台阶C1、台阶D1，计算这四个像素索引值之间的距离L′A1B1、L′B1C1、L′C1D1、L′D1A1，如果L′A1B1＝L′C1D1且L′B1C1＝L′D1A1，继续执行步骤S42，否则，执行步骤S5；S42: Using the mean value of the gray value as the threshold, binarize the gray value of the sample sequence P′, and record the pixels when the gray value binarization produces a step as step A1, step B1, step C1, and step D1, Calculate the distance L' A1B1 , L' B1C1 , L' C1D1 , L' D1A1 between these four pixel index values, if L' A1B1 =L' C1D1 and L' B1C1 =L' D1A1 , continue to step S42, otherwise , execute step S5;
S43：用曲线拟合的方法求出台阶A1、台阶B1、台阶C1、台阶D1的一维亚像素位置A＇、B＇、C＇、D＇，包括：取样本序列P＇台阶A1附近的五个像素(台阶A1前面三个，台阶A1后面两个)，以这五个像素在样本序列P′中的索引值为x坐标，其灰度值的梯度为y坐标，进行二次曲线拟合，拟合出的曲线近似为二次抛物线，二次抛物线的极值点即为沿该梯度方向灰度变化最大的地方，即为台阶A1的一维亚像素位置A＇；以同样的方法分别求出台阶B1、台阶C1、台阶D1的一维亚像素位置B＇、C＇、D′。S43: Calculate the one-dimensional pixel positions A', B', C', and D' of steps A1, B1, C1, and D1 by curve fitting method, including: taking the sample sequence P' near the step A1 Five pixels (three in front of the step A1 and two behind the step A1), the index value of these five pixels in the sample sequence P′ is the x coordinate, and the gradient of the gray value is the y coordinate, and the quadratic curve is fitted The fitted curve is approximately a quadratic parabola, and the extreme point of the quadratic parabola is the place where the gray level changes the most along the gradient direction, which is the one-dimensional pixel position A' of the step A1; in the same way Calculate the one-dimensional pixel positions B', C', and D' of the steps B1, C1, and D1 respectively.
S44：根据步骤S43求出的台阶A1、台阶B1、台阶C1、台阶D1的一维亚像素位置A′、B′、C′、D′，以及步骤S32求出的X角点的像素值L，求出台阶A1、台阶B1、台阶C1、台阶D1的二维亚像素位置A′、B′、C′、D′；S44: According to the one-dimensional pixel positions A', B', C', D' of the step A1, step B1, step C1, and step D1 obtained in step S43, and the pixel value L of the X corner point obtained in step S32 , find the two-dimensional sub-pixel positions A', B', C', D' of the steps A1, B1, C1, and D1;
假设某台阶的一维亚像素位置为m，其对应X角点中心的像素为(x,y)，求出台阶A1、台阶B1、台阶C1、台阶D1的二维亚像素位置；台阶A1的二维亚像素位置为(x+A′-r+1,y-r+0.5)，台阶B1的二维亚像素位置为(x+r+0.5,y+B＇-3r+1)，台阶C1的二维亚像素位置为(x-C′+5r-1,y+r+0.5)，台阶D1的二维亚像素位置为(x-r+0.5,y-D′+7r-1)；Assuming that the one-dimensional pixel position of a step is m, and the pixel corresponding to the center of the X corner point is (x, y), calculate the two-dimensional sub-pixel positions of steps A1, B1, C1, and D1; The two-dimensional sub-pixel position of the step B1 is (x+r+0.5, y+B'-3r+1), the step The two-dimensional sub-pixel position of C1 is (x-C'+5r-1, y+r+0.5), and the two-dimensional sub-pixel position of step D1 is (x-r+0.5, y-D'+7r-1);
S45：按照步骤S32的方法，计算直线A′C′和B′D′的交点坐标，即X角点的像素值L的亚像素位置；S45: According to the method of step S32, calculate the intersection point coordinates of the straight lines A'C' and B'D', that is, the sub-pixel position of the pixel value L of the X corner point;
S46：计算X角点的方向信息：按逆时针方向，根据黑白变化序列得到两条边界线，包括BW(Black-to-White)线、WB(White-to-Black)线，BW线是指从黑到白跳变的边界线；WB线是指从白到黑跳变的边界线；求取BW线、WB线与水平方向的夹角θ1、θ2，即X角点的方向信息；S46: Calculate the direction information of the X corner point: in the counterclockwise direction, two boundary lines are obtained according to the sequence of black and white changes, including the BW (Black-to-White) line and the WB (White-to-Black) line. The BW line refers to The boundary line that jumps from black to white; the WB line refers to the boundary line that jumps from white to black; calculate the angles θ 1 and θ 2 between the BW line, WB line and the horizontal direction, that is, the direction information of the X corner point ;
S47：将该X角点的像素值L的邻域设为不活跃区，表示该X角点已被检测出。避免X角点重复检测。S47: Set the neighborhood of the pixel value L of the X corner point as an inactive area, indicating that the X corner point has been detected. Avoid repeated detection of X corners.
Claims (8)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810077053.1A CN108428250B (en) | 2018-01-26 | 2018-01-26 | X-corner detection method applied to visual positioning and calibration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810077053.1A CN108428250B (en) | 2018-01-26 | 2018-01-26 | X-corner detection method applied to visual positioning and calibration |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108428250A true CN108428250A (en) | 2018-08-21 |
| CN108428250B CN108428250B (en) | 2021-09-21 |
Family
ID=63156290
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810077053.1A Expired - Fee Related CN108428250B (en) | 2018-01-26 | 2018-01-26 | X-corner detection method applied to visual positioning and calibration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108428250B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111047614A (en) * | 2019-10-10 | 2020-04-21 | 南昌市微轲联信息技术有限公司 | Feature extraction-based method for extracting target corner of complex scene image |
| CN111428720A (en) * | 2020-04-14 | 2020-07-17 | 北京神工科技有限公司 | Sub-pixel level visual feature point positioning method and device based on step response matching |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1896682A (en) * | 2005-07-12 | 2007-01-17 | 北京航空航天大学 | X-shaped angular-point sub-pixel extraction |
| CN102095370A (en) * | 2010-11-22 | 2011-06-15 | 北京航空航天大学 | Detection identification method for three-X combined mark |
| CN103093451A (en) * | 2011-11-03 | 2013-05-08 | 北京理工大学 | Checkerboard intersection recognition algorithm |
| CN103345755A (en) * | 2013-07-11 | 2013-10-09 | 北京理工大学 | Chessboard angular point sub-pixel extraction method based on Harris operator |
| CN103927750A (en) * | 2014-04-18 | 2014-07-16 | 上海理工大学 | Detection method of checkboard grid image angular point sub pixel |
| CN104036516A (en) * | 2014-06-30 | 2014-09-10 | 山东科技大学 | Camera calibration checkerboard image corner detection method based on symmetry analysis |
| CN104331900A (en) * | 2014-11-25 | 2015-02-04 | 湖南科技大学 | Corner sub-pixel positioning method in CCD (charge coupled device) camera calibration |
| CN105740818A (en) * | 2016-01-29 | 2016-07-06 | 山东大学 | Artificial mark detection method applied to augmented reality |
| CN105787912A (en) * | 2014-12-18 | 2016-07-20 | 南京大目信息科技有限公司 | Classification-based step type edge sub pixel localization method |
| CN106846412A (en) * | 2017-01-23 | 2017-06-13 | 上海兴芯微电子科技有限公司 | A kind of checkerboard angle point detection process and device |
-
2018
- 2018-01-26 CN CN201810077053.1A patent/CN108428250B/en not_active Expired - Fee Related
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1896682A (en) * | 2005-07-12 | 2007-01-17 | 北京航空航天大学 | X-shaped angular-point sub-pixel extraction |
| CN102095370A (en) * | 2010-11-22 | 2011-06-15 | 北京航空航天大学 | Detection identification method for three-X combined mark |
| CN103093451A (en) * | 2011-11-03 | 2013-05-08 | 北京理工大学 | Checkerboard intersection recognition algorithm |
| CN103345755A (en) * | 2013-07-11 | 2013-10-09 | 北京理工大学 | Chessboard angular point sub-pixel extraction method based on Harris operator |
| CN103927750A (en) * | 2014-04-18 | 2014-07-16 | 上海理工大学 | Detection method of checkboard grid image angular point sub pixel |
| CN104036516A (en) * | 2014-06-30 | 2014-09-10 | 山东科技大学 | Camera calibration checkerboard image corner detection method based on symmetry analysis |
| CN104331900A (en) * | 2014-11-25 | 2015-02-04 | 湖南科技大学 | Corner sub-pixel positioning method in CCD (charge coupled device) camera calibration |
| CN105787912A (en) * | 2014-12-18 | 2016-07-20 | 南京大目信息科技有限公司 | Classification-based step type edge sub pixel localization method |
| CN105740818A (en) * | 2016-01-29 | 2016-07-06 | 山东大学 | Artificial mark detection method applied to augmented reality |
| CN106846412A (en) * | 2017-01-23 | 2017-06-13 | 上海兴芯微电子科技有限公司 | A kind of checkerboard angle point detection process and device |
Non-Patent Citations (5)
| Title |
|---|
| ABDULRAHMAN S. ALTURKI 等: ""A new X-Corner Detection for Camera Calibration Using Saddle Points"", 《RESEARCHGATE》 * |
| FUQING ZHAO 等: ""An Automated X-corner Detection Algorithm(AXDA)"", 《JOURNAL OF SOFTWARE》 * |
| WANG YAN 等: ""An X-corner Detection Algorithm Based on Checkerboard Features"", 《INTERNATIONAL CONFERENCE ON LOGISTICS ENGINEERING, MANAGEMENT AND COMPUTER SCIENCE (LEMCS 2014)》 * |
| 孟偲 等: ""具有方向特性的X角点的亚像素检测定位"", 《北京航空航天大学学报》 * |
| 马帅依凡 等: ""基于人工标记的手术导航仪"", 《山东大学学报》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111047614A (en) * | 2019-10-10 | 2020-04-21 | 南昌市微轲联信息技术有限公司 | Feature extraction-based method for extracting target corner of complex scene image |
| CN111047614B (en) * | 2019-10-10 | 2023-09-29 | 南昌市微轲联信息技术有限公司 | Feature extraction-based method for extracting target corner of complex scene image |
| CN111428720A (en) * | 2020-04-14 | 2020-07-17 | 北京神工科技有限公司 | Sub-pixel level visual feature point positioning method and device based on step response matching |
| CN111428720B (en) * | 2020-04-14 | 2023-09-26 | 北京神工科技有限公司 | Sub-pixel level visual feature point positioning method and device based on step response matching |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108428250B (en) | 2021-09-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107014294B (en) | Contact net geometric parameter detection method and system based on infrared image | |
| CN105894499B (en) | A kind of space object three-dimensional information rapid detection method based on binocular vision | |
| WO2019120011A1 (en) | Target detection method and apparatus | |
| CN108369650B (en) | Method for identifying possible characteristic points of calibration pattern | |
| CN107369159B (en) | Threshold segmentation method based on multi-factor two-dimensional gray level histogram | |
| CN115170669B (en) | Identification and positioning method and system based on edge feature point set registration and storage medium | |
| CN109741356B (en) | Subpixel edge detection method and system | |
| CN112651968B (en) | A method for detecting the deformation and pits of wooden boards based on depth information | |
| CN103198476B (en) | Image detection method of thick line type cross ring mark | |
| CN105203552A (en) | 360-degree tread image detecting system and method | |
| CN108280450A (en) | A kind of express highway pavement detection method based on lane line | |
| CN107145905A (en) | Image Recognition Detection Method for Elevator Fastening Nut Looseness | |
| CN111105452B (en) | Binocular vision-based high-low resolution fusion stereo matching method | |
| KR101589167B1 (en) | System and Method for Correcting Perspective Distortion Image Using Depth Information | |
| CN108460800B (en) | Container image positioning method and system | |
| CN107452030A (en) | Method for registering images based on contour detecting and characteristic matching | |
| CN112184765A (en) | A Vision-based Autonomous Tracking Method for Underwater Vehicles | |
| CN111354047A (en) | Camera module positioning method and system based on computer vision | |
| CN109191516B (en) | Rotation correction method and device of structured light module and readable storage medium | |
| CN102831419B (en) | Method for detecting and blurring plate number in street view image rapidly | |
| CN113313116A (en) | Vision-based accurate detection and positioning method for underwater artificial target | |
| CN107680035B (en) | Parameter calibration method and device, server and readable storage medium | |
| CN114998571B (en) | Image processing and color detection method based on fixed-size markers | |
| CN108428250A (en) | A kind of X angular-point detection methods applied to vision positioning and calibration | |
| CN111462246A (en) | Equipment calibration method of structured light measurement system |
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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210921 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |