JP2646585B2 - Linear component detector - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a linear component detection device that detects a linear component from an original image.

[Conventional technology]

Conventionally, in order to recognize an object, the object is urged as an image, and information such as features of the image is extracted to recognize the object. Therefore, in general, an object is photographed by photographing means such as a television camera, and the image is electrically processed to extract information.

As such electrical processing, the contour of an object in an image may be extracted. As a method of extracting the contour, a method using a mapping based on a Hough transformation or a line segment is extracted by a spherical mapping. There is something.

In the Hough transform, as shown in FIG. 14 (A), a point (xi, yj) on the image plane is converted to ρ = xi cos θ + yj sin θ (1) on the parameter plane as shown in FIG. 14 (B). ) Is converted to a sine wave (mapping function) represented by

When sine waves corresponding to points on a straight line on the image plane are plotted on the parameter plane and a histogram is created, all sine waves intersect at two points. The coordinates (θ ₀ , ρ ₀ ) of this point correspond to the length ρ ₀ and the rotation angle θ ₀ of a vertical foot lowered from the origin to the straight line formed by each point, and represent one straight line. . Accordingly, information on a straight line formed by each of the outer points can be extracted.

By the way, there are the following devices for extracting a contour by such a mapping, that is, a line segment. This is
As shown in FIG. 15, an image input unit 1 such as a television camera,
An edge detecting unit 2 for extracting a contour of an object from image information sent from the image input unit 1 and generating a contour point address
And a speed adjustment buffer 3 for temporarily storing the contour point address, and a function ρ = xi cos θ + yj
a function value calculation unit 4 for calculating sinθ and performing Hough transform;
Buffer 5 for temporarily storing function value ρ, and histogram creating unit 6 for creating a two-dimensional histogram according to function value ρ
A histogram memory 7 for writing a histogram,
And a host computer 10. The host computer 10 includes a function value calculation unit 4, a histogram creation unit 6, a control unit 11 that controls the histogram memory 7, and the buffers 3 and 5, a peak detection unit 12 that detects peak points on the two-dimensional histogram. A determination unit 13 for determining whether a detected peak point is larger than a predetermined threshold, and a contour point which is a basis of a mapping function curve passing through the peak point when the peak point is larger than the threshold. And a line segment output unit 14 that outputs that is a linear component.

On the other hand, by detecting a linear component using such a device, it is possible to approximate a broken line in a region including a curve ("Image Recognition Theory" by Makoto Nagao, published by Corona Co., Ltd.). The basic concept of this broken line approximation will be described with reference to FIG.

First, from the point P ₁ on the periphery of the area to the farthest point P ₂ on the periphery
Draw a line. Next, points P ₃ and P ₄ on the farthest periphery from the line
Find If the distances from the straight lines to the points P ₃ and P ₄ are larger than a certain value d, the straight lines P ₁ P ₃ and P ₂ P ₃ (or P ₁ P ₄ and P ₂ P ₄ ) are drawn. Next, for each of these two straight lines, the point farthest from the straight line is obtained, and the distance to the straight line is calculated. (The points are found on both sides of the straight line.
It's easy to tell which point to take.) If this distance is larger than d, a straight line is drawn again from both ends of the straight line to that point, and the approximation by one straight line is replaced by the broken line approximation by two straight lines. In this way, by increasing the degree of approximation while bisecting the fold line until the distance between the straight line and the farthest point on the periphery is sufficiently small, sufficient approximation of the periphery of the region is obtained by the fold line. Will be obtained. The results will depend, of course, on the starting point. It is better to take the point of maximum curvature as this starting point.

Here Determination of curvature maximum points, from an arbitrary point P _i on the curve as shown in FIG. 17, P _ik point separated by k or to the left and right
Draw a string between and P _{i + k} and measure how far P _i is from this string. When this distance went by moving the point P _i, find the location P _i such that maximum. Then it determines that P _i is a point of curvature maximum when the distance at this point is sufficiently large.

[Problems to be solved by the invention]

By the way, although the conventional linear component detection device can detect a linear component in an image, it has a problem that it cannot detect information on the longest line segment. Thus, for example, can not be set the position of the optimal starting point P _i when performing piecewise linear approximation of the curve as described above, the accuracy of the polygonal line approximation is reduced.

Accordingly, it is an object of the present invention to detect the longest linear component from the detected linear components and output the data.

[Means for solving the problem]

In order to achieve the above object and solve the problems of the prior art, the present invention provides a mapping function calculating means for calculating a mapping function value with respect to the contour point coordinates of an image, and a histogram according to the calculated mapping function value. In a straight-line component detection device including a histogram creation unit to be created, and a line segment extraction unit to extract line segment information indicated by contour point coordinates from a peak point on the histogram, the straight line component detection device passes through a passing peak point of the generated mapping function group. A predetermined mask edge is set, and the mapping function group detects the largest continuous component among the continuous components passing through the mask edge, and determines the largest continuous component among the linear components represented by the peak points. It comprises a longest line segment extracting means for extracting as a function, and a line segment data extracting means for extracting position data of the line segment from the extracted mapping function.

〔Example〕

FIG. 1 shows an example of a straight-line component detecting apparatus according to the present invention. As in the conventional apparatus, an image input unit 21 such as a television camera and an image outline sent from the image input unit 21 are used to extract the contour of an object. An edge detector 22 for extracting and generating a contour point address, a speed adjusting buffer 23 for temporarily storing the contour point address, and a function ρ = xi c
a function value calculation unit 24 for calculating osθ + yj sinθ to perform a Hough transform, a buffer 25 for temporarily storing a function value ρ, a histogram creation unit 26 for creating a two-dimensional histogram according to the function value ρ, and a histogram memory 27 for writing a histogram And a host computer 30. Further, the host computer 30 includes a function value calculation unit 24, a histogram creation unit 26, a control unit 31 that controls the histogram memory 27 and the buffers 23 and 25, a peak detection unit 32 that detects a peak point on the two-dimensional histogram, A peak forming edge point extraction unit 33 for extracting an edge point such that the mapping function passes through the peak point, and a mapping function group obtained by performing a Hough transform on the extracted edge point to analyze the range of existence of the line segment. A line segment range calculating unit 34 to calculate, a line segment component point extracting unit 35 to extract edge points forming the line segment, and a determining unit 36 to determine whether the extracted line segment is longer than a predetermined length to be detected. And a line segment data output unit 37 that outputs line segment data based on the determination result. In this embodiment, the longest line segment extraction unit according to the present invention is realized by a peak forming edge point extraction unit 33 in the host computer 30, and the line segment data extraction unit according to the present invention includes a line segment range calculation unit 34 and a line segment This is realized by the constituent point extracting unit 35.

 Next, the operation of the present apparatus will be described.

The edge detection unit 22 performs a contour point detection process and a thinning process on the image signal A (i, j) sent from the image input unit 21, and extracts an edge composing point with a width of 1. The extracted edge constituent point information is stored in the buffer 23 as B (i, j) as shown in FIG. Here, a value of “1” is given to an edge composing point, and a value of “0” is given to other points. The function value calculator 24 performs Hough transform on the edge component point information B (i, j) obtained in this manner. This processing flow is illustrated in FIG. Note that the symbols shown in FIG.
Are taken as shown in FIG. 3, and θ is taken in steps of 1 ° within a range of −90 ° ≦ θ ≦ 180 ° as shown in FIG. Further, ρ is assumed to be in the range of 1 ≦ ρ ≦ ρmax (ρmax is the maximum value set in the histogram array H (θ, ρ)).

The function value calculation unit 24 stores such a processing result in the buffer 25
, And the histogram creating unit 26 sequentially inputs the function values from the buffer 25 and outputs the two-dimensional histogram H (θ, θ,
ρ) is created and written to the histogram memory 27.

Next, the peak detecting unit 32 detects a peak point on the two-dimensional histogram. This processing flow is illustrated in FIG. In FIG. 6, how to take θ and ρ is the same as in the case of the processing flow shown in FIG. By this processing, a peak point (maximum frequency δ) on the histogram is detected, θ giving the maximum frequency to X, ρ giving the maximum frequency to Y
Is extracted. This (X, Y) represents the parameter of the longest linear component in the B (i, j). Here, the minimum value δ _TH of the detection line is set in advance, and the maximum frequency δ <δ _TH
At the time of, all the processing is ended.

Next, the peak forming edge point extracting unit 33 extracts an edge point from B (i, j) such that the mapping function passes through the peak point (X, Y). This processing flow is as shown in FIG.
By this processing, an edge point on a straight line having a parameter of (θ, ρ) = (X, Y) is placed on the array C as C (i, j) =
Stored as 1.

Next, the line segment range calculation unit 34 performs a Hough transform on the previously generated C (i, j), and extracts a mapping function group that passes through (θ, ρ) = (X, Y). This is realized by processing H ′ (θ, ρ) prepared as a two-dimensional histogram array in the same flow as the processing flow shown in FIG.
Then, the line segment range calculation unit 34 calculates the existing range of the line segment by analyzing the generated mapping function group H ′ (θ, ρ) as follows. This is shown in FIG. First, when an edge image as shown in FIG. 8 (I) is Hough-transformed, a mapping function pattern as shown in FIG. 8 (II) is obtained. All the mapping function patterns pass through (θ, ρ) = (X, Y). When this mapping function pattern is enlarged, it becomes as shown in FIG. 3 (III), and when a horizontal mask EDGE (θ) is set in the ρ = Y portion, it becomes as shown in A, B, and C portions of FIG. 4 (IV). , Each line segment appears as one continuous component. Here, the largest continuous component (part A in this case) is detected, and θ addresses (α, β) at both ends are calculated. Then As shown in FIG. 5 (V), the existence range of the longest component in C (i, j) is calculated. Here, θmin and θmax are the values θ that give the extreme values of the mapping function group of the part A in FIG.
An address. This processing flow is illustrated in FIG. In FIG. 9, steps (9-1) and (9)
In steps (2-4) and (9-3), a process of giving "1" to the position where the mapping function exists and giving "0" to the other positions is performed, and in steps (9-4) and (9-5), The mask is being set. Also, the code POINT in step (9-10)
(L) = 0 is the left end position of the continuous component, and the code KOSU
(L) = KOSU (l) +1 represents the number of constituent points of the continuous component.

Next, the line segment constituent point extracting unit 35 extracts the edge points forming the line segment from C (i, j). This means that from all points where C (i, j) = 1, This is realized by extracting points. This processing flow is illustrated in FIG.

Then, the determination unit 36 determines the presence of the extracted line segment. This compares the number of edge points Z of the extracted line segment with the preset minimum line segment length Z _TH , Is determined. If it is determined that Z ≧ Z _TH , the first
Move from step (10-6) to (10-7) in FIG.
_If it is determined to be _TH, it is determined that there is no line segment, and the entire process ends. When Z ≧ Z _TH , the line segment data output unit 37 outputs the line segment data obtained as described above. Here, the peak points on the histogram (X and Y detected by the peak detection unit 32), the line segment existence range (the line segment range calculation unit 3
Output θmin, θmax obtained in 4). With these data, it is possible to reconstruct the line segment position as shown in FIG.

FIG. 12 shows a second embodiment of the straight-line component detecting apparatus according to the present invention, in which a new histogram is created by removing the mapping function based on the line segment for which the extraction processing has been completed. A function removing unit 38 is provided. The mapping function removing unit 38 removes a mapping function that is already unnecessary after the extraction process is completed according to a processing flow as shown in FIG. 13, to prevent the occurrence of a false peak point on the histogram, This enables stable detection of short linear components. If such processing is repeatedly executed until δ <δ _TH as described in the processing shown in FIG. 6 or until Z <Z _TH as shown in FIG. 10, B (i, j)
It is possible to accurately detect a linear component having a length longer than a certain length. By using the line segment data detected by such processing, it is possible to create a line drawing of an object, perform modeling, and the like.

In the above embodiment, the Hough transform is performed with the origin at the upper left of the image, but the origin may be arbitrarily selected.

〔The invention's effect〕

As described above, the linear component detection device according to the present invention sets a predetermined mask edge that passes through the passing peak point of the generated mapping function group, and sets the mapping function group of a continuous component that passes through the mask edge. And the longest line segment extracting means for extracting the largest continuous component among the straight line components represented by the peak point as a mapping function relating to the longest line segment, and extracting the longest line segment from the extracted mapping function. Since the line segment data extracting means for extracting the position data is provided, the longest line segment among the linear components included in the edge image is detected, the position data is output, and the line segment can be reconstructed. . Therefore, for example, when approximating a curved line, it is possible to detect the position of the longest straight line component in the image area, so that it is possible to stably find an optimal starting point. There is an effect that the approximation accuracy is further improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of a linear component detecting device according to the present invention, FIG. 2 is a diagram showing edge constituent point information, and FIG.
5 is a diagram for explaining how to take the signs θ and ρ used in FIG. 5, FIG. 4 is a diagram for explaining the range of the sign θ used in FIG. 5, and FIG. 5 is a function value calculation shown in FIG. 6 is a flowchart showing an operation example of the peak detection unit shown in FIG. 1, and FIG. 7 is a flowchart showing an operation example of the peak forming edge point extraction unit according to the embodiment of the present invention. FIG. 8 is a diagram illustrating a processing principle in a line segment range calculating unit according to the embodiment of the present invention. FIG. 9 is a flowchart showing an operation example of the line segment range calculating unit according to the embodiment of the present invention. FIG. 11 is a flowchart showing an operation example of a line segment constituent point extraction unit, a determination unit, and a line segment data output unit according to the embodiment of the present invention. FIG. 11 is a diagram showing an example of line segment reconstruction.
FIG. 13 is a block diagram showing a second embodiment of the present invention, FIG. 13 is a flowchart showing an operation example of the mapping function removing unit shown in FIG. 12, FIG. 14 is a diagram for explaining the principle of Hough transform, FIG. 15 is a block diagram showing an example of a conventional linear component detection device, and FIG.
FIG. 16 and FIG. 17 are diagrams for explaining a technique of approximating a broken line of a curve. 32 peak detecting section 33 peak forming edge point extracting section (longest line segment extracting means) 34 line segment range calculating section (line segment data extracting means) 35 line segment component point extracting section (line segment data) Extraction means) 37 ... Line segment data output unit

Claims (1)

(57) [Claims] 1. A mapping function calculating means for calculating a mapping function value with respect to a contour point coordinate of an image; a histogram generating means for forming a histogram according to the calculated mapping function value; A linear component detection device comprising: line segment extraction means for extracting line segment information indicated by point coordinates; a predetermined mask edge passing through a passing peak point of the generated mapping function group is set; Detecting the largest continuous component among the continuous components passing through the edge,
A longest line segment extracting means for extracting this as a mapping function relating to the longest line segment among the straight line components represented by the peak points, and a line segment data extracting means for extracting position data of the line segment from the extracted mapping function And a linear component detecting device.