JPH0346045B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method and an apparatus for detecting the positional coordinates of corner points of an object to be measured represented by a grayscale image and the inclination of its posture at high speed.

The method and device of the present invention can be applied to, for example, a positional deviation of a workpiece to be worked on by a robot from a teaching point,
It can be applied to a device that can quickly measure posture deviations, correct the robot's teaching points in real time, and allow the robot to continue working normally.

[Prior Art] Most of the mechanical parts that robots work on have corner points. By detecting the positional coordinates of this corner point and the inclination of the edge line connected to the corner point, it is possible to calculate the positional deviation and posture deviation of the entire object.

One method for detecting corner points has traditionally been to use a large-scale computer to process a gradation image in which one screen consists of 256 x 256 pixels and the gradation level of each pixel is digitized into 256 classes (8-bit data). A method has been devised to do so. For example, edge lines are extracted by first-order differentiation and second-order differentiation, and edge lines and corners are extracted by mask operators and template matching.

However, although the use of a large-scale computer allows complex and sophisticated image processing, it is essentially a sequential arithmetic processing method, which has the disadvantage of requiring a large amount of time and increasing costs.

In order to compensate for this, a method has also been considered in which the video signal from the television camera is binarized by setting an appropriate threshold, the data is compressed, and then the contour is traced to search for corner points. However, this method uses 2
It is difficult to set a threshold for converting data into values, and even if you make great efforts to adjust lighting, backgrounds, etc., it is often impossible to process the data due to disturbances caused by various noises.

For these reasons, the above-mentioned method has not been applicable to the field of control, such as robot control, which requires real-time control, low cost, and high reliability.

[Problems to be Solved by the Invention] There is no method or device for detecting the position, orientation, etc. of a corner point of an object that is sufficient to satisfy the high speed and high reliability applicable to the field of control.

The present invention has been made to improve the above-mentioned drawbacks, and an object of the present invention is to detect the position and orientation of a corner point of an object to be measured at high speed and with high precision using simple steps.

[Means for Solving the Problems] The first invention relates to a corner point detection method, and the structure of the invention is as follows.

A method for processing a grayscale image of a workpiece obtained by projecting an image of the workpiece onto a light-receiving surface of a photoelectric converter and sampling the grayscale level of a video signal in accordance with the plane coordinates of the light-receiving surface. In this step, a virtual line is set on the grayscale image, a brightness gradient of the grayscale image at a point on the virtual line is calculated along the virtual line, and a point where the absolute value of the brightness gradient is greater than or equal to a predetermined value is determined. is detected as an edge point on the edge line of the object to be measured, and starting from the edge point, the direction of the edge line at that point is determined from the brightness gradient of the point on the edge line, and the direction of the edge line along the edge line is determined. Sequentially executes a process of calculating the next point on the edge line separated by an interval, moves along the edge line, and determines that the absolute value of the brightness gradient is greater than or equal to a predetermined value, and the direction of the brightness gradient is greater than or equal to a predetermined value. A method for detecting a corner point of an image is characterized in that a changing point is detected as a corner point of an edge line of the object to be measured.

FIG. 2 schematically shows a grayscale image 1 of the object to be measured. The grayscale image 1 is an image obtained by sampling the video signal output from the photoelectric converter corresponding to the plane coordinates on the light receiving surface and digitizing the grayscale level.

2 is an image of the object to be measured (hereinafter simply "object to be measured")
). The present invention detects the coordinates of the corner point A of the object to be measured 2. A virtual line L that crosses the fixed object 2 is set on the grayscale image 1.
The virtual line L connects the edge line 21 of the object to be measured 2 and the edge point B.
are intersecting.

First, while moving the moving point P on the virtual point L along the virtual line L, the brightness gradient of the grayscale image 1 at the moving point P is determined. Here, the brightness gradient is a vector quantity, and is a gradient of brightness of an image with respect to plane coordinates. The method of calculating the brightness gradient is arbitrary, but for example, including the moving point P, 2×
It can be calculated by Robert's operator, Sobel's operator, etc. using the data of each pixel divided into a mesh of 2, 3×3, 4×4, etc.

It is known that the absolute value of the brightness gradient becomes large on the edge line 21, and by detecting that the absolute value of the brightness gradient at the moving point P has become larger than a predetermined value, the moving point is moved to the edge line. It can be determined that it is edge point B on 21.

When the position coordinates of edge point B are detected in this manner, the direction in which edge line 21 exists can be estimated from the direction of the brightness gradient at edge point B. That is, since the edge line 21 extends in a direction perpendicular to the brightness gradient (vector), the edge line 21 exists in the direction of the vector obtained by rotating the brightness gradient by ±90 degrees. In this manner, the direction of the edge line 21 at edge point B is determined, and in that direction, a point Q1 set apart from edge point B by a minute distance is selected, and the brightness gradient at that point Q1 is determined. If the absolute value of the brightness gradient is greater than or equal to a predetermined value, then the direction of the edge line 21 is estimated from the brightness gradient at that point Q1, and furthermore, the direction of the edge line 21 is estimated at a point Q2 located at the next minute interval on the edge line 21. Select.

In this way, points Q1, Q2, ..., Qn
It is possible to move along the edge line 21 by selecting . By repeating these operations one after another, corner point A is reached. In the vicinity of the corner point, the direction estimated from the brightness gradient Vn at point Qn on the edge line 21 and the direction estimated from the brightness gradient Vn+ ₁ at point Qn+ ₁ are significantly different. That is, it is possible to detect a large change in the brightness gradient and determine the point on the edge line 21 at that time to be the corner point A of the object to be measured 2.

By the above operations, the corner point A of the object to be measured 2 is
The position coordinates of can be detected. Therefore, the inclination of the edge line 21 can be determined from the position coordinates of the edge point B and the corner point A, and the attitude of the object to be measured can be determined from the inclination of the edge line 21.

Next, the second invention will be explained.

FIG. 1 is a block diagram showing the concept of the device of the invention.

The configuration of the device of the present invention is as follows.

The image of the object to be measured is projected onto the light-receiving surface of the photoelectric converter, and the gray level image of the object to be measured is obtained by sampling the gray level of the video signal in accordance with the plane coordinates of the light-receiving surface. an image input device for storing; a differentiator that inputs data from the image input device and calculates a brightness gradient at an arbitrary point on the grayscale image; and a differentiator that inputs the output of the differentiator to calculate the absolute value of the brightness gradient. an absolute value calculating device for calculating the direction of the brightness gradient by inputting the output of the differentiating device; and a direction calculating device for calculating the direction of the brightness gradient by inputting the output of the differentiating device; an edge point detection device that detects a point at which the output of the absolute value calculation device is larger than a predetermined value as an edge point on the edge line of the object to be measured; A cycle in which the direction of the edge line at that point is determined from the brightness gradient of a point on the edge line, and the plane coordinates of the next point separated by a minute interval along the edge line are given to the differentiator. are executed sequentially to move on the edge line, and when the output of the absolute value calculation device is greater than a predetermined value and the output of the direction calculation device changes by more than a predetermined value, that point is detected as a corner point of the edge line. A corner point detection device includes a corner detection device.

The image input device 3 is a device that stores a grayscale image 1 regarding plane coordinates. For example, a device that extracts a video signal using an electron tube imaging device such as a television camera, or a solid-state imaging device such as a CCD or BBD, samples the video signal in accordance with plane coordinates, digitizes it, and stores it as a grayscale image 1. be. That is, the image input device 3 may be equipped with video detection means such as a television camera, or may be a device that inputs data of the grayscale image 1 detected by another device and stores it in an accessible manner. . The point is that the device is capable of storing at least the grayscale image 1.

A differentiator 4 is connected to the image input device 3, and the differentiator 4 takes in the data of the grayscale image 1 from the image input device 3 using the signal S1 from the edge point detection device, and uses the signal S1 to input the data of the grayscale image 1. Find the brightness gradient at the identified point. The configuration of this differentiator 4 may be arbitrary as long as the brightness gradient at that point can be determined. For example, if a differential operator is used as explained in the method invention, a product-sum calculator that executes the operation may be used. Can be configured.

The absolute value calculation device 5 and the direction calculation device 6 are connected to the differentiator 4, and are devices that calculate the absolute value and direction of the brightness gradient determined by the differentiator 4, respectively.

The edge point detection device 7 supplies the plane coordinates of a point on the virtual line L to the differentiator 4 as a signal S1 along the virtual line L taken on the grayscale image 1, and sends a signal S2 from the absolute value calculation device 5. is input, the plane coordinates of the point when the absolute value of the brightness gradient becomes larger than a predetermined value are stored, and the edge point B is detected.

The corner detection device 8 inputs the signals S2 and S3 from the absolute value calculation device 5 and the direction calculation device 6, respectively, and also inputs the signal S4 indicating the brightness gradient determined by the differentiator 4, and changes the signal S3. edge point B while detecting the amount of change until the amount of change becomes large.
This device detects the corner point A by sequentially applying the coordinates of points on the edge line as a signal S5 to the differentiator 4 starting from . Although the signal S4 in the figure is input from the differentiator 4, since the signals S2 and S3 include information on the brightness gradient, the direction of the edge line 21 may be calculated from the signals S2 and S3. The corner detection device 8 sequentially supplies the coordinates of the next point in the calculated direction, that is, in the direction of the edge line 21, to the differentiator 4 as a signal S5. When the change in the value of the signal S3 from the direction calculating device 6 is larger than a predetermined value, the coordinates of that point are stored and this point is detected as the corner point A of the object to be measured 2.

If the edge line 21 is a straight line, the magnitude of change in the signal S3 may be determined based on the direction of the brightness gradient at the edge point. Also, adjacent 2
The determination may be made based on the magnitude of the change in direction of brightness gradient at a point.

The above methods and devices can be accomplished by computers, digital circuits, analog circuits, and combinations thereof, but each of the above devices may include connecting the above functional devices in a pipeline structure to improve the speed of corner point detection. is desirable. Especially differentiator 4
In order to obtain each component of the brightness gradient at the same time, it is preferable that each component calculation section of the differentiator 4 is processed in parallel. Further, it is preferable that the brightness gradient determined by the differentiator 4 is simultaneously input in parallel to an absolute value calculation device 5 and a direction calculation device 6, so that the absolute value and direction of the brightness gradient can be calculated in parallel.

In this case, since a buffer register or the like is not provided between each processing unit to temporarily store sequential data, the image data is not synchronized with the system clock within each processing unit. Since the image is processed as it flows, high-speed image processing within one calculation cycle is possible.

Furthermore, in order to perform parallel processing between the differentiator 4 and the absolute value arithmetic device 5 or the direction arithmetic device 6, it is necessary to The signals S1 and S5 from the detection device 8 may be predicted from previous data and output in advance.

In the present invention, in order to prevent the corner search from being disturbed by noise, uneven illumination, etc. in the image, a first-order differential is introduced into the space of the grayscale image to remove the effects of uneven illumination. Also, the scale of the spatial first-order differential operator is 3×3, 4×4, 5×
5 or the like, and a differential operator capable of smoothing noise in the image can be employed.

[Examples] Hereinafter, the present invention will be described in detail based on specific examples.

The detection principle of this embodiment will be explained with reference to FIG. The same parts as in FIG. 2 are given the same numbers. The entire gray scale image of the object 2 including the corner point A is taken as f(x, y), and two arbitrary points P1 and P2 are set in this image. The search starts from point P1 along the straight line L connecting P1 and P2 in the direction of vector r. Spatial first-order differential values fx and fy for the grayscale image f(x, y) at each point
A brightness gradient consisting of is calculated. Here space 1
The order differential values fx and fy are the x and y components of the brightness gradient. Then vector absolute value √ ₂ +
Calculate fy ² and compare it with the threshold TH. If the vector absolute value √ ² + ² is larger, it is determined to be edge point B, and its coordinates are detected as edge point coordinates (xe, ye). do.

Next, the spatial first derivative value (fxe,
directional gradient θe = ATAN(fye/fxe)
, and in the clockwise search (in the case of Figure 3), θe+
Determine the 90° corner point search direction t. In counterclockwise search, the search direction is θe−90°. A search point Q is determined along the search direction, the slope of the brightness gradient on that point is determined, and the next search point is determined. In this way, the search is performed sequentially, and the vector absolute value √ ² + ² of the spatial first derivative of the search point is greater than or equal to the threshold TH, and the direction of the brightness gradient of the search point θ =
ATAN(fy/fx) is detected, and if |θ−θe| is larger than the threshold TG, the search point is set to corner point A.
The coordinate value of A is determined as the corner position coordinate value (xc,
yc). Further, the corner posture can be detected as the inclination of a straight line connecting the coordinate values (xe, ye) of edge point B and the coordinate values (xc, yc) of corner point A.

FIG. 4 is a block diagram showing the configuration of an apparatus according to an embodiment of the present invention.

As the image input device 3, an image memory 31 is provided that stores a grayscale image 1 obtained by sampling a video signal from a television camera (not shown). The differentiator 4 includes product-sum calculators 41 and 42 and an address generator 90 for determining each component of the brightness gradient.
An address control unit 90 configured as a part of
Consists of 0. The absolute value calculation device 5 and the direction calculation device 6 are an absolute value calculation device 50 and a direction calculation device 6, respectively.
Consists of 0. The edge point detection device 7 includes an edge/corner determiner 91 and an address generator 90.
and an edge position coordinate value detector 71,
The corner detection device 8 includes an edge/corner determiner 9
1, an address generator 90, and a corner position coordinate value detector 81. Further, the outputs of the edge position coordinate value detector 71 and the corner position coordinate value detector 81 are input to the robot control device 100, and the robot control device 100 detects the object to be measured from the corner point coordinates and the slope of the edge line. Detects deviations in the position and posture of the robot and corrects the robot's teaching points.

An input image signal f is captured in the image memory 31 in advance. The address generator 90 contains the (x, y) coordinate values of the set points P1 and P2, and a clockwise or counterclockwise search that specifies whether the corner search is to be performed in either direction of the edge line 21 using + or -. A designated signal RLPS is given. For example, when the scale of the product-sum operation is 3×3, this address generator 90 generates sub-scanning address signals S7, (xi, yi) that drive the product-sum calculation units 41 and 42 by (-1, -1).
(0, -1) (1, -1) (-1, 0) (0, 0) (1
,
0) (-1, 1) (0, 1) (1, 1) is repeated in the order of (x-1,
y-1), (x, y-1) (x+1, y-1) (x-
1, y) (x, y) (x+1, y) (x-1, y+
1) Repeatedly output in the order of (x, y-1) (x+1, y+1). This image memory address signal S
6 (X, Y), image memory 3
The image signals S8 and f(X, Y) outputted from 1 are converted by product-sum calculators 41 and 42 into spatial first-order differential values fx and fy in the x direction and y direction, respectively, and are further converted into vector absolute values. The arithmetic unit 50 generates a vector absolute value signal S2 of the first spatial differential value, F=√ ² +
fy ² and the direction calculator 60 convert it into a direction signal S3 of spatial first-order differential value, θ=ATAN(fy/fx).

Initially, the address generator 90 starts along the edge point search direction r from the set point P1 to the set point P2.
In order to access the data of each pixel, image memory address signals S6, (X, Y) are sequentially output. When the vector absolute value signal S2 of the spatial first-order differential value becomes larger than a preset threshold value TH, the edge/corner determiner 91 determines that
A trigger signal TGE is output to the address generator 90 and the edge position coordinate value detector 71. In response to this trigger signal TGE, the address generator 90 switches from a mode of searching on a straight line connecting set points P1 and P2 to a mode of searching for a corner point, and sends an edge address signal to the edge position coordinate value detector 71. S9, output (x, y). The edge position coordinate detector 71 detects this edge address signal S9.
(x, y) are latched at the timing of the trigger signal TGE, detected as edge position coordinate values (xe, ye), and outputted to the robot control device 100.

Next, the address generator 90 uses the spatial first-order differential values fx and fy in the x and y directions and the clockwise and counterclockwise search designation signals RLPS to extract pixel data from the image along the corner point search direction t. Outputs memory address signal S6, (X, Y). The edge/corner determiner 91 uses a vector absolute value signal S2 of the first spatial differential value to a preset threshold value.
TH, and when the absolute value of the difference between the brightness gradient direction θe of edge point B and the brightness gradient direction θ of the current point becomes larger than a preset threshold TG, a trigger signal TGC is generated. Output.
The corner position coordinate detector 81 is connected to the address generator 9
Corner address signal S9 from 0 (x, y)
is latched at the timing of the trigger signal TGC, the point is detected as the corner position coordinate value (xc, yc), and the value is sent to the robot controller 100.
Output to.

In FIG. 4, the image memory 31 is an ordinary random access memory, and the edge position coordinate value detector 71 and the corner position coordinate value detector 81 can be constructed from ordinary latch circuits. Furthermore, the absolute value calculator 50 and the direction calculator 60 are of the ordinary EPROM type.
Can be configured with memory. That is, the absolute value and direction of the brightness gradient are stored in the EPROM address specified by the spatial first-order differential values fx and fy in the x and y directions, and a corresponding function table is created, and fx, If fy is accessed as an address signal, the corresponding absolute value and direction will be obtained on the data line.

FIG. 5 is a block diagram showing a detailed circuit configuration of address generator 90 of FIG. 4. Referring to FIG.

The address generator 90 includes an address control section 900 for generating an address signal S6 in the image memory 31, and a main scanning section 910 for sequentially generating coordinates of points on the line along the virtual line L or the edge line. Consists of.

The address control unit 900 includes a clock oscillator 901 which serves as a synchronizing signal for accessing the image memory 31.
, a frequency divider 902 that divides the output signal of the clock oscillator 901, a sub-scanning circuit 903 that outputs a control signal for differential operation in synchronization with the output signal of the clock oscillator 901, and The adder 904 generates an address signal for fetching data from the image memory 31.

The main scanning unit 910 inputs the coordinates of two points P1 and P2 at both ends, and outputs the coordinate values of the points on the virtual line L along the virtual line L. and a corner direction generator 915 which inputs the brightness gradient and a signal RLPS indicating the search direction and outputs a signal indicating the direction of existence of the next point on the edge line.
and a latch circuit 914 that stores the coordinates of the current search point.
and an adder 913 which generates the coordinates of the next point on the edge line based on the output of the latch circuit 914 and the output of the corner direction generator 915.

In FIG. 5, a clock oscillator 90 outputs a basic clock signal CLK1 in this embodiment to a frequency divider 902. For example, if the scale of the product-sum operation is 3×3, this frequency divider 902 is 1/9, 4×
The basic clock signal CLK1 is divided by 1/16 if it is 4, or 1/25 if it is 5×5, and the product-sum operation cycle clock signal CLK2 is output. At the beginning of the search operation, the set points P1 and P2 cause the straight line generator 911 in the main scanning section 910 to generate address signals for points on the virtual line along the edge point search direction. This address signal is transmitted to the switch 91.
2 to the input address signal S of the adder 904.
1, S5 (x, y), and the basic clock signal
The sub-scanning address signal S7 (xi, yi) output from the sub-scanning circuit 903 driven by CLK1 is added by an adder 904 and converted into an image memory address signal S6 (X, Y) in edge point search mode. Ru.

Next, after detecting edge point B, the trigger signal TGE
This causes the switch 912 to switch, thereby switching the address generator 90 from the edge search mode to the corner search mode. In this corner search mode, the latch circuit 914 connects the adder 900 to the adder 900 at the timing of the output clock signal CLK2 of the frequency divider 902.
The input address signals S1, S5 (x, y) to the latches are latched. The signal is output to one input terminal of adder 913 as an address signal (x, y). A corner direction generator 915 emits a signal indicating a small amount of movement, and an adder 913 outputs a corner direction address signal (-Δx, Δy) for clockwise rotation, and a corner direction address signal (Δx, -Δy) for counterclockwise rotation. output to the other input terminal. Therefore, adder 913
Address signals (x±Δx, y±Δy) indicating the coordinates of the next point added by are passed through the switch 912 and input address signals S1, S5 (x,
y), and this signal is further added to the sub-scanning address signal (xi, yi) to become the image memory address signal S6, (X, Y) in the corner point search mode.

FIG. 6 is a block diagram showing a detailed circuit configuration of corner direction generator 915 of FIG. 5. 6th
In the figure, the spatial first-order differential values in the x and y directions
fx and fy are the SIN function calculator 930 and the
Input to COS function calculator 931. the result,
x component Δx (Δx=
fy/F) and the y component Δy (Δy=fx/F). The output signals of the arithmetic units 930 and 931 are further polarized by polarity converters 932 and 933 according to the clockwise and counterclockwise search designation signals RLPS,
For clockwise search, corner direction address signal (-
Δx, Δy), or corner direction address signals (Δx, −Δy) in the case of a counterclockwise search.

FIG. 7 is a block diagram showing the detailed circuit configuration of the product-sum calculators 41 and 42 of FIG. 4. In FIG. 7, a spatial first-order differential operator W in the x direction is set in advance in the product-sum calculation coefficient memory 410. Sub-scanning address signal S7, (xi, yi)
Therefore, the product-sum calculation coefficient W (xi, yi) read from the product-sum calculation coefficient memory 410 is transferred to the image memory 3 by the image memory address signal S6, (X, Y).
The image signal f(X, Y) read from 1 is multiplied by the multiplier 411, and its output is sent to the adder 41.
input to one input terminal of 2. This adder 41
The output of 2 is latched by the latch circuit 413 at the timing of the basic clock signal CLK1. Its output is input to the other input terminal of the adder 412, and the product-sum operation cycle clock signal
It is latched by the latch circuit 414 at the timing of CLK2. Eventually, after one calculation cycle corresponding to one cycle period of the clock signal CLK2, one product-sum calculation is completed, and the first-order differential value fx in the x-direction space is output.

In FIG. 7, the y-component product-sum calculation unit 42 of FIG. Orientation space first derivative
The operation is exactly the same as that of the product-sum calculator 41 except that fy.

FIG. 8 shows the edge/corner determiner 91 of FIG.
FIG. 2 is a block diagram showing a detailed circuit configuration of the circuit.

The edge/corner determiner 91 includes a comparator 950 that compares the magnitude of the absolute value S2 with a threshold value TH;
A trigger circuit 951 that holds the result and outputs a trigger signal, a latch circuit 952 that inputs the direction signal S3 and stores the value at edge point B, and a latch circuit 952 that inputs the direction signal S3 and stores the value at edge point B. an adder 953 that calculates the difference between the adder 953, an absolute value circuit 954 that calculates the absolute value of its output, a comparator 955 that compares its output with the threshold TG, a gate circuit 956 that inputs its output, and a gate circuit 956 that holds its output. and a trigger circuit 957 that outputs a trigger signal.

In Figure 8, the vector absolute value F of the spatial first derivative value and the preset threshold value TH
are compared by a comparator 950, and if the vector absolute value F is greater than the threshold value TH, a comparison output EDGE of level "1" is obtained, and if it is smaller, a comparison output EDGE of level "0" is obtained.
This comparison output is EDGE of D type and clock CK of trigger circuit 951 consisting of flip-flop.
The coordinate values of edge point B (xe, ye) are input to the terminal.
A trigger signal TGE for latching is output from the trigger circuit 951. At this time, the direction signal θ of the spatial first-order differential value inputted to the latch circuit 952 at the timing of the trigger signal TGE is latched, and becomes the direction signal θe of the edge point B. Adder 95
The direction signal θ at the time of corner search is input to the positive input terminal of No. 3, and the direction signal θe of the edge point is input to the negative input terminal.
is input, the direction difference signal θ-θe is obtained, which is converted into an absolute value by the absolute value circuit 954, and the absolute value signal of the direction difference |θ-θe| is sent to one of the comparators 955 at the next stage. Input to input terminal. Comparator 9
The other input terminal of 55 has a threshold value set in advance.
Since TG is set, the comparator 955 outputs the comparison output DIFG, which has a level of "1" if the direction difference absolute value signal |θ-θe| is greater than the threshold value TG, and "0" if it is smaller, to the AND gate 956. Output to one input terminal of Since the comparison output EDGE is input to the other input terminal of the AND gate 956,
When the level of both input terminals becomes “1”, that is, the vector absolute value signal S2 of the spatial first-order differential value
is larger than a preset threshold TH, and the absolute value signal of the difference between the direction signal θe of the spatial first-order differential value of the edge point and the current direction signal θ |θ
−θe| is the preset threshold TG
When it becomes larger, the output level of AND gate 956 also becomes "1". This "1" level signal is input to the clock CK terminal of a trigger circuit 957 consisting of a D-type flip-flop in the next stage, and is used to latch the coordinate values (xc, yc) of corner point A. This becomes the trigger signal TGC. In the initial state of operation, the two trigger circuits 951 and 957 are
It is reset by the reset pulse RESET, and the trigger signals TGE and TGC are each “0”.
held at the level.

The device of this embodiment has the above configuration and operation.

According to the apparatus of this embodiment, since it is possible to process gray scale images, it is also possible to detect corner points inside the image, which was impossible with existing apparatuses based on binarized image processing. In addition, the hardware includes two product-sum calculators that calculate the spatial first-order differential value and two dedicated calculators that calculate the vector absolute value and direction value of the spatial first-order differential value, and each calculation stage is pipelined. The connection made it possible to perform image processing at high speed. Furthermore, in the past, corner points were searched inefficiently by performing line-sequential scanning and tracing contour lines, but with this method,
Using the fact that the direction value calculated from the next differential value is 90 degrees off from the next corner search direction, it is fed back to the corner search address calculation, making it possible to find corner points extremely efficiently. . Since these hardware components are composed of relatively simple elements, the image processing device can be made compact and inexpensive.

FIG. 9 shows the edge/corner determiner 91 of FIG.
Fig. 10 is a block diagram showing the configuration of an example in which the processing is realized by a computer, and Fig. 10 is a flowchart showing the processing. In this embodiment, the threshold value TH and the threshold value TG are set in advance in the computer 960 by the console 961. Absolute value calculator 5
The vector absolute value signal S2, F of the spatial first-order differential value output from 0 and the direction signal S3, θ of the spatial first-order differential value output from the direction calculator 60 are calculated by the computer 9.
60. The embodiment is the same as the embodiment shown in FIG. 4, except that the trigger signal TGE is output if it is determined to be an edge point, and the trigger signal TGC is output if it is determined to be a corner point. Next, the operation flow of the computer 960 shown in FIG. 10 will be explained.

The computer 960 operates in interrupt mode, and in response to an interrupt signal from the hardware side, in step 200,
Read the vector absolute value and direction θ. Next, in step 202, the threshold value TH set by the console 961 and the vector absolute value F are compared, and NO
If so, the operation ends immediately, and if YES, the process moves to step 204. In step 204, it is determined whether the interrupt is the first time, and if YES
If so, the trigger signal TGE representing the edge point is outputted in step 206, and the current direction θ is stored as the edge point direction θe in step 208, and the operation ends. If NO in step 204, the absolute value of the difference between the current direction and the edge point direction θe |θ−θe| is compared with the threshold value TG set by the console 961 in step 210, and if NO, the operation is started immediately. If YES, the process moves to step 211 and a trigger signal representing the corner point is output.
Outputs TGC and ends the operation.

With the above-described operation of the computer, corner positions and orientations can be searched in the same way as in the embodiment shown in FIG. You can easily find the optimal judgment conditions by changing them freely. Since the image targeted by the present invention is a complex grayscale image, the threshold value TH and TG are set in advance.
In many cases, it is not possible to set a fixed value.
This embodiment can solve the above-mentioned difficulties.

[Effects of the Invention] The present invention detects points on the edge line of the object to be measured represented by a grayscale image based on the brightness gradient, moves on the edge line using the brightness gradient, and moves in the direction of the brightness gradient. A method and apparatus for detecting corner points by change. Therefore, it is resistant to noise because it uses the brightness gradient, and the corner point detection speed is high because it uses a simple method of sequentially estimating the edge line using the brightness gradient and moving on the edge line. . Furthermore, since points on the edge line are estimated sequentially, movement on the edge line is reliable and corner points can be detected with high accuracy.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the concept of the apparatus of the present invention. FIG. 2 is an explanatory diagram illustrating the corner point detection principle of the present invention. FIG. 3 is an explanatory diagram illustrating the corner point detection principle of a device according to a specific embodiment of the present invention. FIG. 4 is a block diagram showing the configuration of the apparatus of the embodiment. Figures 5 and 6
7 and 8 are block diagrams showing detailed configurations of the address generator, corner direction generator, product-sum calculator, and edge corner determiner, respectively, of the same device. FIG. 9 is a block diagram showing the configuration of a portion of an embodiment in which the edge corner determination device of the same apparatus is configured by a computer, and FIG. 10 is a flowchart showing the processing of the computer. 2...Object to be measured, 21...Edge line, B...Edge point, A
...Corner point, L...Virtual line, V...Brightness gradient.

Claims (1)

[Claims] 1. A measured object obtained by projecting an image of the measured object onto the light-receiving surface of a photoelectric converter and sampling the gray level of the video signal in accordance with the plane coordinates of the light-receiving surface. In the grayscale image processing method, a virtual line is set on the grayscale image, a brightness gradient vector of the grayscale image at a point on the virtual line is calculated along the virtual line, and the absolute value of the brightness gradient vector is calculated. Detecting a point whose value is a predetermined value or more as an edge point on the edge line of the object to be measured, and starting from the edge point, detecting the brightness gradient at a point where the absolute value of the brightness gradient vector is a predetermined value or more. A process of estimating the direction of the edge line at that point from the direction of the vector, and calculating the next point that is spaced from the point in the estimated direction by a minute interval and where the absolute value of the brightness gradient vector is greater than or equal to a predetermined value. are sequentially executed to move along the estimated direction, and locate a point at which the absolute value of the brightness gradient vector is a predetermined value or more and the direction of the brightness gradient vector changes by a predetermined value or more as an edge line of the object to be measured. A method for detecting a corner point of an image, characterized in that the corner point is detected as a corner point of an image. 2 A grayscale image regarding the plane coordinates of the measured object obtained by projecting the image of the measured object onto the light-receiving surface of the photoelectric converter and sampling the gray level of the video signal corresponding to the plane coordinates of the light-receiving surface. an image input device that stores data, a differentiator that inputs data from the image input device and calculates a brightness gradient vector at an arbitrary point on the grayscale image, and a differentiator that inputs the output of the differentiator to calculate the brightness gradient vector. an absolute value calculation device that calculates an absolute value; a direction calculation device that inputs the output of the differentiator to calculate the direction of the brightness gradient vector; and a direction calculation device that calculates the plane coordinates of a point along an arbitrarily set straight line on the grayscale image. an edge point detection device that sequentially supplies the information to the differentiator to detect a point at which the output of the absolute value calculation device is larger than a predetermined value as an edge point on the edge line of the object to be measured; and after detecting the edge point. , starting from the edge point, estimating the direction of the edge line at that point from the brightness gradient vector of the point on the edge line, and moving away from the point at that point by a minute interval along the estimated direction, and The output of the absolute value calculation device is moved along the estimated direction by sequentially executing a cycle of giving the plane coordinates of the next point at which the output of the absolute value calculation device is larger than a predetermined value to the differentiator. is a predetermined value or more, and when the output of the direction calculation device changes by a predetermined value or more, a corner point detecting device detects that point as a corner point of the edge line. 3. The differentiator includes an address control unit that sequentially takes in data of the grayscale image of plane coordinates taken in a mesh shape from the input device, including the input coordinates of the point, and calculates each component of the brightness gradient. 3. The corner point detection device according to claim 2, further comprising a product-sum calculator that performs a product-sum calculation on the data according to a predetermined operator. 4. The differentiator, the absolute value arithmetic device, and the direction arithmetic device are connected in a pipeline, and the arithmetic processing is performed within a cycle in which the coordinates of the point are provided to the differentiator. The corner point detection device according to item 2. 5. The product-sum calculator has a calculator that independently calculates the x component and y component of the brightness gradient, and each component calculator is connected to the image input device in parallel, and simultaneously calculates each component in parallel. The corner point detection device according to claim 3, characterized in that the corner point detection device calculates the following. 6. The absolute value calculator and the direction calculator input each component of the brightness gradient as an address signal,
3. The corner point detection device according to claim 2, further comprising a storage device that stores absolute values and directions corresponding to each component as data and is accessed by the address signal.