A structured light corner point detection method based on slope search

Abstract

The application discloses a kind of structure light angle point detection methods based on slope search, comprising: the colorful structure light image is converted into gray image, and its edge information is enhanced then converted into binary image to obtain the connected domain R of image region segmentation m , and R m Edge contour pixel set is then solved R m Centroid coordinates C m ; according to the shape characteristics k polygon of standard structure light and C m , edge pixel set is divided into k different subsets and with C m The intersection of the diagonal of the minimum circumscribed quadrilateral contained in the minimum circumscribed quadrilateral is defined and C m The intersection point is the R m Corner point sought; by searching the same slope, the corner point coordinates of all R m Are solved, and the corner point detection of the whole structure light image is realized.The application method can quickly obtain the corner point of structure light image, and solve the problem that the existing corner point detection algorithm is prone to multiple detection, missed detection and wrong detection when detecting the corner point of structure light image in underwater imaging.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and underwater imaging technology, and particularly relates to a structured light corner point detection method based on slope search. BACKGROUND

[0002] The purpose of corner point detection is to find the interest points in an image, and the interest points are also called feature points or key points. The corner point detection has wide applications in the fields of image matching, target tracking and three-dimensional reconstruction. The corner point is the most basic interest point in an image, which can be defined as the intersection of two edges in the image or the local maximum point of curvature on the target contour. At present, common corner point detection algorithms include Harris algorithm, Fast algorithm, Surf algorithm and Shi-Tomasi algorithm.

[0003] The underwater imaging method based on structured light is an important technology for realizing real-time and cross-medium imaging, and the corner point detection of the structured light image is a very key link in the underwater imaging method based on structured light. However, the random fluctuation of the water surface can cause complex refraction distortion and motion blur of the structured light image captured by the camera, and the distortion and blur of the image seriously affect the visual observation of people and easily cause misjudgment of the image content. For this kind of image, the existing corner point detection methods cannot accurately extract the feature information of the image, and are prone to false detection, missed detection and multiple detection, which interferes with the subsequent image restoration and target recognition. SUMMARY

[0004] In view of the deficiencies of the existing corner point detection methods, the present application provides a structured light corner point detection method based on slope search, so as to ensure the accuracy of the corner point detection in the underwater imaging process and avoid the problems of multiple detection, missed detection and false detection of the corner point detection caused by the distortion and blur of the image due to the random fluctuation of the water surface.

[0005] The technical scheme adopted by the present application to solve the technical problems is as follows:

[0006] A structured light corner point detection method based on slope search, comprising the following steps:

[0007] Step 1), converting a colored structured light image into a gray-scale image, defined as J H (x); then enhancing the edge information of the image J H (x); converting the image J H (x) into a binary image

[0008] Step 2), performing region segmentation on the binary image J to obtain the connected domain of the image, and obtaining the edge contour pixel set of each connected domain R m ​ wherein m = 1, …, n, n is the maximum number of connected domains; then the centroid coordinates C of the connected domain R m are solved m ;

[0009] Step 3), according to the shape feature k polygon of the standard structured light and the centroid position of the connected domain R m , the edge pixel set is divided into k different subsets wherein k is the number of edges of the polygon in the structured light pattern;

[0010] Step 4), a minimum circumscribed quadrilateral containing the edge pixel subset is respectively constructed with the centroid coordinates C m as the vertex and the k different subsets ; then the slope λ m,r of the diagonal line passing through the centroid coordinates C m in each minimum circumscribed quadrilateral is calculated m,r ;

[0011] Step 5), the slope of the line connecting each pixel point coordinates in contained in each minimum circumscribed quadrilateral and the centroid coordinates C m is calculated, and the pixel point coordinates (x m,r , y m,r ) with the slope λ m,r is found, which is the corner point coordinates of the connected domain R m solved; the corner point coordinates of all connected domains are solved, and then the corner point detection of the entire structured light image is realized.

[0012] Advantages or beneficial effects of the present application:

[0013] (1) The present application converts the color structured light image into a gray scale image by using the channel separation technology, and enhances the edge features of the structured light image by using the principle of edge filtering, thereby solving the problems that the absorption of light by the water body will cause color deviation of the structured light image in the underwater imaging process, and the random fluctuation of the water surface will also cause distortion and blurring of the image, and the edge features of the structured light are not clear.

[0014] (2) The present application quickly obtains the corner points of the structured light image by using the slope search method, thereby solving the problems of multiple detection, missed detection and false detection of the corner points of the structured light image in the underwater imaging process. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 is the flow chart of the structured light corner point detection method based on the slope search in the embodiment;

[0016] Figures 2-5A real shooting and detection effect diagram of the structure light corner point detection method based on slope search in the embodiment; wherein,

[0017] Figure 2 A gray image to be processed;

[0018] Figure 3 A binary image;

[0019] Figure 4 A centroid calculation result of a connected domain;

[0020] Figure 5 A corner point position distribution after slope search;

[0021] Figures 6-7 A processing result diagram of a stripe grid structure light image under water wave disturbance in the embodiment; wherein,

[0022] Figure 6 A structure light image shot with distortion;

[0023] Figure 7 A corner point detection result of an image. DETAILED DESCRIPTION

[0024] The content of the present application will be described in detail below in combination with the drawings and embodiments, examples of which are shown in the drawings. The embodiments described below by referring to the drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.

[0025] Embodiment:

[0026] The structure light corner point detection method based on slope search proposed by the present application will be described below in combination with the drawings.

[0027] When imaging from underwater to air, a stripe grid structure light pattern is projected to the water surface, and a reference structure light image can be obtained when the water surface is calm.

[0028] Firstly, the image is separated by channel using the HSI color model, and the H channel image is guided filtered, and then an OSTU adaptive threshold segmentation algorithm is used to obtain a binary image; then all connected domains in the image are obtained by region segmentation of the image through a connected domain operator, and the edge pixel set and the centroid position of each connected domain are calculated; finally, the corner point coordinate information of the image is obtained by a slope search method.

[0029] As shown in Figure 1 A structure light corner point detection method based on slope search, comprising the following steps:

[0030] Step (1), the adaptive adjustable structured light pattern is projected to the water surface, the camera under the calm water surface can obtain a reference structured light image, the colored structured light image is converted into H component, S component and component according to HSI model, wherein H represents the frequency of color, which is called hue; S represents the depth of color, which is called saturation; I is the intensity or brightness of the image; the H component is defined as J H (x), as shown in Figure 2 ; then the edge information of the image J H (x) is enhanced by the guided filter algorithm, and the image J H (x) is converted into a binary image using the OSTU threshold segmentation method. As shown in Figure 3 ;

[0031] Step (2), the binary image is regionally segmented to obtain the connected domain of the image according to the connected domain operator, and the edge contour set R of each connected domain R m is obtained. Wherein m = 1, …, n, n is the maximum number of connected domains; then the centroid coordinates C m of the connected domain R m is calculated using the spatial moment function, as shown in Figure 4 ; wherein the spatial moment function is defined as follows:

[0032]

[0033] x represents the column coordinates of any pixel point in the edge contour pixel set, and y represents the row coordinates of any pixel point in the edge contour pixel set; the centroid coordinates C m can be obtained from formula (1):

[0034]

[0035] Wherein, represents the column value of the centroid C m , represents the row value of the centroid C m ;

[0036] Step (3), according to the shape feature k polygon of the standard structured light and the centroid position of the connected domain R m , the edge pixel set R is divided into k different subsets, wherein k represents the number of edges of the polygon in the structured light pattern, when the shape of the connected domain of the standard structured light is a rectangle, the value of k is 4, at this time the edge pixel subset R is divided into 4 subsets, as follows:

[0037]

[0038] in, Represents the edge pixel set Different subsets of r, r = 1, 2, ..., k;

[0039] Step (4), with centroid coordinates C m For the vertex and k distinct edge pixel subsets Construct k containing Find the minimum bounding quadrilateral and calculate the coordinates C passing through the centroid of each minimum bounding quadrilateral. m The diagonal slope λ m,r When k=4, the corresponding 4 edge pixel subsets The diagonal slope λ m,r They are defined as follows:

[0040]

[0041] in, Representing a subset column coordinates, Representing a subset The row coordinates, r = 1, 2, 3, k;

[0042] Step (5): Calculate the area contained in each minimum circumscribed quadrilateral. The coordinates of each pixel and the centroid coordinates C m Find the slope of the line, where the slope is λ. m,r pixel coordinates (x) m,r ,y m,r This point is the desired connected region R. m The corner coordinates are calculated to solve for the corner coordinates of all connected components, thereby achieving corner detection of the entire structured light image; where the slope is λ. m,r pixel coordinates (x) m,r ,y m,r The formula for calculating ) is:

[0043]

[0044] Compared with existing corner detection methods, the method of this invention performs corner detection on structured light images under water wave disturbance, such as... Figure 6 , Figure 7 As shown, the corner detection results of this type of image do not have problems of false detection, missed detection, or misdetection, thus improving the accuracy and robustness of structured light corner detection.

[0045] The above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Therefore, any changes made in accordance with the shape and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for detecting corner points of structured light based on slope search, characterized in that, comprising the steps of: Step 1), the colored structured light image is converted into a gray scale image, defined as ; then the edge information of the image is enhanced ; the image is converted into a binary image ; and ​ Step 2), the binary image is region segmented to get the connected domain of the image, and the edge contour pixel set of each connected domain is obtained ​​​​​​​ Step 3) based on shape features of the standard structured light edge and connected component centroid position, the edge pixel set is divided into different subsets where is the number of edges of the polygon in the structured light pattern;​ Step 4), in the barycentric coordinates for the vertices and different subsets of respectively construct a minimum circumscribed quadrangle containing the edge pixel subset ; then calculate the slope of the diagonal line passing through the barycentric coordinates in each minimum circumscribed quadrangle ; Step 5), calculate the area contained in each minimum bounding quadrilateral. The coordinates of each pixel and the centroid coordinates Find the slope of the line. pixel coordinates This point is the desired connected region. Corner coordinates; solve for the corner coordinates of all connected components, and thus realize corner detection of the entire structured light image; In step 5), the slope is the pixel point coordinates The calculation formula is: denotes the column coordinate of an arbitrary pixel point in the set of edge profile pixels, denotes the row coordinate of an arbitrary pixel point in the set of edge profile pixels, denotes the set of edge pixels of different subsets, .

2. The slope search based structured light corner point detection method according to claim 1, wherein, Step 2) comprises: According to the connected domain operator on binary image The connected domain of the image is obtained by region segmentation, and the edge contour set of each connected domain , , is obtained; Then the centroid coordinates of the connected domain are calculated by using the spatial moment function , wherein the spatial moment function is defined as follows: （1） , respectively represent the column coordinate and the row coordinate of any pixel point in the edge profile pixel set; the centroid coordinate can be obtained from formula (1) : （2） wherein represents the column value of the centroid , represents the row value of the centroid . 3.The slope search based structured light corner point detection method of claim 1, wherein, In step 3), when the connected domain shape of the standard structured light is a rectangle, the value of 4, at which time the edge pixel subset is divided into 4 subsets as follows: （3） wherein denotes the set of edge pixels different subsets of the set of edge pixels .

4. The slope search based structured light corner point detection method of claim 1, wherein, In step 4), when the diagonal slopes of the corresponding four edge pixel subsets are defined as follows, respectively:​ （4） wherein denotes the column coordinates of the subset denotes the column coordinates of the subset denotes the row coordinates of the subset denotes the row coordinates of the subset .

Images