JPH03167681A - Calculating method for characteristic parameter of object - Google Patents

Abstract

PURPOSE:To reduce the increase of data despite a complicated object shape and to process an image simply and at a high speed by handling the contour lines as the chain codes, i.e., a 1-dimensional data train when the characteristic parameters of the area, the centroid, etc., showing the characteristics of the object shape are calculated through the image processing. CONSTITUTION:The binary images obtained by photographing an object are stored in an image memory 1 for each picture element, and a processing circuit 2 consisting of a microcomputer, etc., computes the contents stored in the memory 1 to calculate the characteristic parameters. In this case, the contour lines are first tracked for calculation of the basic characteristic parameters, i.e., the moment, area, centroid, attitude angle, and circumscribed rectangle respectively. Then both furthest and nearest points are detected and the directions set from the centroids of both points are calculated. Then the derivative characteristic parameters, i.e., the moment ratio, the area rate, the aspect ratio, etc., are calculated. Furthermore the total area, the total circumferential length, etc., are calculated. A binary image to be stored in the memory 1 consists of the picture elements of logic '0' and '1'.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a method for calculating characteristic parameters such as area and center of gravity representing the characteristics of the shape of an object by image processing.

Conventional technology Conventionally, when calculating feature parameters such as the area and center of gravity of an object based on a binary image of the object,
The number of pixels in which logic rQJ or logic "1" for each pixel stored in the image memory is continuous in the X direction (TV screen raster scanning direction), that is, the run length, is determined by hardware or software in the Y direction (raster scanning direction). Solid separation is performed by sequentially obtaining run length data in the direction perpendicular to can be calculated in a short time when calculating the run length using hardware, but has the drawback that it requires special hardware and the equipment is expensive.Also, when calculating the run length using software, , there is a problem in that it takes a long processing time as a result of searching and finding a connected region of image data two-dimensionally.

Furthermore, the run length data is processed using the SRI algorithm (
Regarding the process processed by Connectivity Analysis, the more complex the shape of the object and the larger the number of objects, the longer the run length and
As data increases, and as a result, the number of data to be compared and processed in connectivity analysis increases exponentially.
There is a problem with the processing time. An object of the present invention is to provide a method for calculating feature parameters of an object in a short time by eliminating the drawbacks of the above-mentioned conventional techniques. Means for Solving the Problems The present invention performs solid separation of individual objects from the inclusion relationship between closed loops of the contour lines of the objects, and also obtains run length data from the contour lines represented by chain codes. This is a method for calculating feature parameters of objects that calculates feature parameters. Effect According to the present invention, the closed loop inclusion rW of the contour image
Since solid separation is performed based on a one-dimensional search process of examining the J coefficient, even if the shape of the object is complex and the number of objects is large, the processing required for solid separation will increase even though the number of contour data will increase. It only takes a short time.

In addition, in the process of obtaining run length data, the chain code values of the points before and after each point on the contour are checked to determine the start and end points of a run (a series of connected areas connected in the raster scanning direction). The run length is calculated from this. This method is a one-dimensional search, and even when implemented using software, the processing time is very short, and special hardware is not required, so the device can be implemented at low cost.

Embodiment FIG. 1 is a flowchart showing the overall steps of an embodiment of the present invention, and FIG. 2 is a block diagram showing the configuration for performing the calculation processing. The image memory 1 stores, for example, a binary image for each pixel obtained by imaging an object, and the processing circuit 2 implemented by a microcomputer or the like processes the contents stored in the image memory 1. Compute the feature parameters. In calculating this feature parameter, in steps 01 to n3 of FIG.
In addition to calculating the center of gravity, attitude angle, and circumscribed rectangle, the farthest point and nearest point are detected, and the directions from the center of gravity are calculated. In step n8, derived feature parameters such as moment ratio, area ratio, aspect ratio, etc. are calculated, the total amount of holes in the object is calculated, and the total area, total perimeter, etc. are calculated. First, we will explain the contour tracing method. Figure 3 shows a binary image stored in image memory 1 representing a bright object in a dark background. In this figure 3,
A black circle indicates a logic "0" pixel, as shown in FIG. 4 (1), and a white circle indicates a logic "1" pixel, as shown in FIG. 4 (2). Search for the outline of the object in this image using the following steps (1) to (5). (1) From the upper left to the lower right of the binary image shown in Figure 3, in this order from left to right in the X direction and from top to bottom in the Y direction, each pixel is ”, that is, the point where the logic changes from dark to logical “1”, that is, bright. (2> If the point surrounded by the rectangular frame 30 in Fig. 3 is detected as a candidate on the contour line and the contour line has not been traced yet, this point does not have tracked information called a marker, Determined as the starting point for contour tracking.For each pixel,
As mentioned above, information indicating whether tracking has been completed is also stored. (3) As shown in Figure 5, the current pixel on the contour line is
Assuming that the pixel before moving to this pixel Q is P, the search for a pixel moving from the pixel Q is processed according to FIG. That is, among the eight neighboring points of pixel Q, pixel P3 in the direction of pixel P→pixel Q and four pixels PI, P2; P5, P4 on the left and right in that direction are
Check in the order of P5 and select the first logical "1" pixel found. When pixel Q is the starting point, pixel P1 is selected as the pixel to the left of the starting point. If a searched marker is attached to this pixel P1, this contour search is discontinued. (4) In this way, the contour line is traced, and when the contour line returns to the original starting point, the contour tracing is terminated. Figure 6 shows the tracking results. When tracing the contour line, the pixel on the left hand is logical "1".
In this embodiment, the outline of the object is traced counterclockwise. In FIG. 6, the contour line is indicated by reference numeral 31. Such a procedure is similarly carried out for tracing the contour of the hole. The tracing of the hole contour is done clockwise, as shown in FIG. 7, and the hole contour is indicated by reference numeral 32. The starting point in Figures 6 and 7 is indicated by a double circle. (5) When contour tracing is completed in this manner, traced information called a marker is attached to pixels on the contour lines 31 and 32. Therefore, even if a candidate point on the contour line 31 or 32, for example, the point indicated by the broken line frame 33 in FIG. is stored, so no further tracking is performed.

This contour line is recorded and stored using a chain code in which the starting point S1 shown by a double circle in FIG. 6 and the direction of movement from there are sequentially represented by numbers 0 to 7 shown in FIG. For example, the chain code of the contour line 31 starting from the starting point S1 of the 6th [2I is r5456644
6..."

Advantages (a1) to (a3) of contour tracing according to steps (1) to 5) are enumerated as follows.

(a1> This is common whether the contour represents an object or a hole. (a2> The structure is simple and processing is fast. (a3) In contour tracing, image information is processed in one dimension. Since it is used (linearly), even if the size and area of the object becomes N times larger and the number of pixels increases, the time required for tracking will only be multiplied by the square root of N.

In order to reduce the time required to detect the starting point of a contour line, in one embodiment of the present invention, the starting point is detected by hardware. Therefore, the size 3 shown in FIG.
A so-called mask operator for X3 pixels is employed. , each pixel A to ■ of this mask operator represents the active value of each pixel, that is, logic "0" or logic "1". When the following first or second equation is satisfied, the pixel corresponding to pixel E is selected as a candidate for the starting point of the contour line, and the X-
Store the position in the Y coordinate system.

A-B・C-D-E = 1 (truth value)
...(1) A. B-C-D-E=
1 (truth value)...
<2> In detecting the starting point of the contour line using this hardware configuration, for example, the points in the rectangular frame shown in FIG. 10 are candidates for the starting point. For each candidate point, perform the contour tracing procedure described above. Since each pixel is configured to avoid double tracking using a marker, is the actual starting point of the outline a square? Of each pixel in , only the candidate points are double circles. These starting points include a starting point S1 on the outer circumference of the object and a starting point S2 on the outline of the hole. In this type of elliptical command, the starting point for contour tracing is detected by the hardware configuration, so
This method has the advantage that the detection speed can be significantly improved compared to a two-dimensional search in which a starting point is detected by executing a computer program using software, which will be described below.

In order to detect the starting point of a contour line, a procedure for detecting the starting point using software instead of a hardware configuration will be described as another embodiment of the present invention.This procedure is, so to speak, a thinning method. ), the logical value of each pixel is checked at predetermined intervals m (for example, m-4 in this embodiment) in the X and Y directions of one image shown in FIG.
The pixels whose logical values are to be examined are indicated by square frames in FIG. In the X direction, check the logical values of pixels at regular intervals m from the left to the right in FIG. 11, and find out that the logical values are
If an interval is found where the logic changes from logic "r" to logic "1", check the point at which the logic changes from logic "0" to logic "1" pixel by pixel within this section. For example, if pixel 34 has logic "0"
, and when the next checked pixel 35 is logic "1", the logic value changes from the logic value ro of the pixel 34 to the logic value r1 for each lit element along the arrow 36 in the X direction. I will check whether it is true or not. As a result, point 37 indicated by a double circle changes from logic "O" to logic "1" between pixels 34 and 35.
This is the first point that changes to pixel 37, and this pixel 37 is detected as a candidate for the starting point. After that, the contour line is traced from the starting point 37. Such software detects the starting point of the contour line by thinning out the pixels at regular intervals m, so the number of vM image data to be inspected can be significantly reduced and processing speed can be increased. ．． After the contour tracing procedure described above, the second step is to distinguish between objects and holes and determine the inclusion relationship, as described below. After the contour line is detected, the first one of the pixels to be ovalized is
As shown in Fig. 2, by searching for the uppermost and leftmost point, that is, the starting point SIS2 in Figs. 6I2l, 7 and 10, and examining the pixel directly above it, It can be determined whether a contour represents an object or a hole. For example, in FIG. 6, the contour line 31 is determined to be the contour line 31 of an object because the pixel 38 directly above the starting point S1 is logic "0". Furthermore, the starting point S2 in FIG. 7 is the pixel 39 directly above it.
Since the logic rlJ is true, the contour line 32 is determined to be the contour line of a hole. The outline 31 of the object is O - L O
OP (Object Loop), the latter as H-
It is called LOOP (Hole Loop) and is managed by assigning a loop number.

On the other hand, if we consider the inclusion relationship, the contour line can be defined as the following (
b1) to (b4) are possible.

(b1) Contour line of an object that does not belong to other contour lines (Mast
er Object Loop, abbreviation M-0-LOOP
).

(b2) The contour line of the hole belonging to the above contour line M-0-LOOP (Master Hole Loop, abbreviated as M-H
-Loop>. (b3) The contour line of the object belonging to the above contour line M-H-LOOP (Slave Object Loop, abbreviated as S
-OLOOP).

(b4) Outline i1 of the hole belonging to the above outline S-0-LOOP (Slave Role Loop, abbreviated as S-
H-Loop>, FIG. 12 shows four such contour lines.

(bl) above. ~(b4) will be called the contour line, that is, the attribute of the loop. This attribute is in the following thousand order <6
) to <8>.

(6) Assign loop numbers to the contour lines in order starting from the one at the top left.

(7) Set the attribute of the contour line at the top left to M10-LO
Mark as OP. (8) Starting point S of the contour line in Figure 13 in order of loop number
Focusing on point 3, search for other contour lines to the left from S3. If the contour line exists, the loop number should be smaller than that of the contour line of current interest, and the attribute of the inclusion rWt relation of the contour line has already been determined. The attributes are determined as shown in Table 1 from the type of contour line currently being focused on and the attributes of the left contour line.Here, attributes with this mark belong to the object to which the left contour line belongs. shall be taken as a thing. Also, attributes marked with ** are
It shall belong to the left contour line.

Thirdly, a procedure for converting a chain code into run length data according to the present invention will be explained.

In the present invention, the contour tracing procedure does not use the conventional concept of run length at all, so it is easy to detect objects with complex shapes, and data corresponding to run length can be immediately obtained from contour information. It can be calculated and found. For example, Fig. 14 shows a sequence of points on the traced contour line, and in the figure, the horizontal direction where the square frame part is,
In other words, it corresponds to the run length in the X direction.
Reference mark S indicates the starting point of the contour line. Now, focusing on three arbitrary points on the contour line, as shown in FIG. -1,P. ,P. ．． ,. Point Pl. ■l
Let the chain code from to point P1 be dirl, and point P,
From P. . If the chain code to 1 is dir2, then
As shown in Table 2 from the relationship between dirl and dir2,
Uniquely, the role of point P on the run length (i.e., r
ole) is determined.

(Margin below) 2nd table dirl Here, S: Start point of run length E: End point of run length T: Start and end point of run length X: Midway point on run length Table 3 shows the roles that any arbitrary point can take on the run length, and the chain codes dirl and dir2 in Figures 16 (1) to 16 (9) and their roles are shown in Table 3. , find the X-Y coordinates of all pixels that make up the contour line, and calculate the loop number (i.e. m a r k )
, and the role on the run length (namely, role), create the table shown in Figure 17. For example, in FIG. 14, on the line 40 whose Y coordinate is y,
The points on the contour line are R1 and R2, and their X coordinates are xi, x
2, in Fig. 17, the X coordinate and its roles S and E are stored in the contour point sequence information corresponding to the Y coordinate, mark is information for managing the closed loop, and ro1e is the start and end of the run. , and other information.

In the table shown in FIG. 17, the contour point sequence information is arranged in ascending order, that is, the X coordinates are arranged in order from the smallest value to the largest value to facilitate use in calculations described later. The run length n on the line 40 of the Y coordinate y in FIG. 14 is given by the following third equation. n=x2−xl+
1...(3> In this way, the run length n can be easily determined. In the following explanation, the table shown in FIG. 17 will be referred to as a run length table (abbreviated as RLT). This run length table is used. For example, as shown in FIG.
1, the contour point sequence information in Fig. 17 is four points 4.
The X coordinates from 2 to 45 are stored, and their roles S, E are also stored, so that the run lengths na, nb of the object contour 46 can be easily calculated. Also, as shown in the 19th section, for example, there are an object outline 47 and a hole outline I148.49, and at this time Y
In the coordinate line 50, points 51 to 56 are stored as contour point sequence information in FIG. 17, and the roles S and E of these points 51 to 56 are stored. , ne can be calculated.

The present invention has the excellent advantage that the run length can be easily determined in the relatively complicated images shown in FIGS. 18 and 19.

Fourth, we will describe the procedure for calculating the center of gravity and attitude angle, which are characteristic parameters, using the run length table. Generally, when determining the area and center of gravity of an object, the logical value is ``1''.
'', but the image information must be examined two-dimensionally (planarly). In addition, to find the attitude angle, we further calculate the moment of inertia, which includes the calculation of the squares of the X and Y coordinates. This calculation requires a large amount of time.

For example, to find the area of the object in Figure 20, calculate the number of pixels with a value of 1 for each Y coordinate, that is, the line Si (run length indicated by the dashed line in the figure), and calculate all Si It is sufficient to calculate the sum of the values, but all binary images must be examined.

However, using the run length table described above, St
Since we can immediately know the X coordinates of both endpoints, ni = xe
i −xsi + 1 (4)
(xei, xsi are the X coordinates of both end points of Si) Fermentation = Σn
i...(5) The area can be easily determined by i.

Similarly, the moments around the origin of the above object can be expressed by the following equations using the above ni, so these are also the first moments about the Y axis ( The geometric moment> is, The second moment of inertia about the Y axis is The first geometric moment about the X axis is =Σ(ni-xsi+ni (ni-1)/2)1
・ ...(8> Quadratic moment about the X axis (moment of inertia) (Σxj)2=E (xs i"+(xs i+1)
”+- ・+(xs i+ni-1)21IJ
1 = Σ(ni-xsi2+2xsini (nii 1〉/2 + ni(ni 1) (2ni-1)/6) ... (9> The product moment of inertia is calculated by the equations 6 to 10. The center of gravity G (g
...(11>gy=(value of formula 9)/area
...(12〉θ=
-(90-θ1) A≧OB〈0...
(13) θ=-θ1A≧OB≧0...
(14) θ - θI A<O B≧O
...(15) θ=90−θI A<O
B<O...(16) However, A1 (value of formula 10 (value of formula 8) x (■■ of formula 6)
■/Area...〈18〉B-((Kinichi of the 9th formula (value of the 7th formula) I X (1-1/Area) ・
...(19) tan2-θ1=2A/B
- (20).

The area can be found from the fifth equation above. Fifth, a procedure for calculating a circumscribed rectangle using a chain code will be described. Using the attitude angles mentioned above, the rectangle circumscribing the object will be described with reference to Figure 21. θ in FIG. 21 represents the attitude angle.

Here, we assume that the long axis direction of the circumscribed rectangle coincides with the direction of the object's attitude. At this time, the coordinate system x-y
In the coordinate system X-Y, which is rotated by θ, the range (the broken line area in the figure) where the points forming the outline of the object exist is a circumscribed rectangle.

Now, the coordinate pi in the x-y coordinate system of the point that ellipses the contour line
(xi, yi>, the coordinates of the X-Y coordinate system are Pi (Xi,
Yi), the coordinate transformation formula is as follows. Xi=xicosθ+yisinθ
−(21) Yi=−xsinθ+yicosθ
・-<'n> However, i=1, n From these values, At, Bl. Cl. The coordinates of Di are calculated as follows. Al (Xmin, Ymax) Bl (Xmin, Ymin) CI (Xmax, Ymin) DI (Xmax, Ymax) However, Xmin=min (Xi, , Xn)
---(23)Xmax=max(Xi,
, Xn) --- (24)Y
min=min (Yl, , Yn)
・= (25) Ymax=max (Yl
, , Yn) - (26) Also, the major axis length Zaxis and minor axis length saxis of the object are as follows.

lax i s=Xmax-Xmin
- (27) saxi s=Ymax-Ym
i n − (28) On the other hand, since points on the contour are represented by chain codes, a certain point Pi
(xi, yi) and the next point Pi+1 (xi+1, yi
+1) has the following relationship.

xi+ΔX→xi+1 ・
...(29〉yi+Δy→yi+1
...(3o) However, the chain code and Δ
The correspondence between X and Δy is shown in Table 4.

(Leaving space below) Table 4 Therefore, after finding the first point using equations 21 and 22, subsequent points can be found sequentially using the following equations, speeding up the processing. X+. +=Xi+Δxco
sθ+Δysinθ. ．． ．． (31)Y.
．． ．． =Yi−Δxsinθ+Δycosθ
- (32) Sixth, use the chain code to calculate the farthest point and nearest point.

The point on the outline of an object that is farthest from the center of gravity is the farthest point! The point where k is also close is called the nearest point. here,
A method for obtaining these will be described using the chain code obtained as described above.

FIG. 22 shows the concept of a method for calculating the farthest point A2 and the nearest point B2. Any point on the contour line is defined as Pi(xi, yi)(
i = 1, n), and the center of gravity G (gxgy). A and B are points that each satisfy the following formula.

dmax-min (di, , dn)
−(33) dmin=max (di,, d
n) −(34) However, di−(xi−gx>”+(yi−gy)2・=(35
) Performing the above calculation for all points on the contour line requires processing time and is not practical, but since points on the contour line are represented by chain codes, the calculation can be simplified using the method described below. can. Now, a certain point pi on the contour line
(xi, yi) and the next point p +x (X Ix,
3' t-+ ) is known to have the following relationship. xi+Δx−+X+u+
+ (36) yi+Δy→3' +++
...<37) However, the correspondence between the chain code and ΔX, Δy in Equations 36 and 37 is as shown in Table 4 above. Therefore, if we calculate Equation 38 for the first point, we get
The following points can be found sequentially using Equation 39. di” = (xi-gx>2+(yL-gy)2
...(38>(f++1" = <X+u
gχ) ” + <yI−+ gy) ”'=X++1
'2X++t ・gx+ gx2 +'/ 141'
2yt-+ 'g3' +gy2-(xi+Δx)
2-2(xi+Δx) gx+gx'+(yi+Δy)
"-2(yi+Δy) gy+gy"=x i 2-2
x i - gx+gx2+y i ”-2y i-g
y+gy”+2x i・ΔX+Δx2−2Δx−gx+
2y i・Δy+Δy2−2Δy−gy=di”+2x
− iΔx+2y−iΔy+(Δx2−2Δx−gx−
2Δy'gy+Δy'')...(39) Also, n=1.
By calculating the square table for 511 in advance, it is possible to speed up the iterative calculation of Equation 38. Seventh, the derived feature parameters are calculated based on the basic feature parameters regarding the object and the hole calculated as described above. The basic feature parameters are area, center of gravity, perimeter, moment of inertia, attitude angle, major axis length, minor axis length, farthest point, nearest point, etc., and the derived feature parameters are:
These include ratio calculations such as the moment ratio, area ratio, and aspect ratio of objects, and total calculations such as the total area and total perimeter of holes in objects.9 Effects of the Invention As described above, according to the present invention, the contour line can be It is treated as a chain code, which is a one-dimensional data string, so even if the shape of the object becomes complex, the amount of data increases little, which simplifies the process of calculating feature parameters and speeds it up. Become.

[Brief explanation of the drawing]

Fig. 1 is a flowchart showing the procedure of an embodiment of the present invention, Fig. 2 is a block diagram showing the configuration of an embodiment of the present invention, and Fig. 3 shows a bright object in a dark background for contour tracing. Figure 4 is a diagram showing the logical values of pixels, Figure 5 is a diagram showing pixels to be searched, and Section 6 is a diagram showing object tracking results.
FIG. 7 is a diagram showing the hole tracking results, FIG. 8 is a diagram showing the principle of chain code, and FIG. 9 is a diagram of the operator used to detect the starting point by the hardware ellipse.
FIG. 10 is a diagram showing a binary image showing detection of candidate points by the hardware configuration, FIG. 11 is a diagram showing detection of candidate points when detecting a starting point by software,
Fig. 12 is a diagram showing the attributes of the contour line, Fig. 13 is a diagram for explaining the rM starting point S3 of the contour line, Fig. 14 is a diagram explaining the sequence of points on the contour line, and Fig. 15 is a diagram for explaining the rM starting point S3 of the contour line. No. 3
Section 16 is a diagram for explaining the role of each point, Section 177 is a diagram showing a contour point sequence table according to the present invention, and FIG. 18 is a diagram for explaining contour point sequence information. Figure 19 is a diagram for explaining other contour point sequence information, Figure 20 is a diagram showing the procedure for calculating the area of an object, and Figure 21 is a diagram for explaining the procedure for calculating the circumscribed rectangle. FIG. 22 is a diagram showing the concept of the method of calculating the farthest point A2 and the nearest point B2. 1... Image memory 2... Processing circuit, S 1 S 2... Starting point

Claims (1)

[Claims] In a binary image of an object, a closed loop of the contour is determined by tracing the contour and recording the connection relationship of points on the contour using a chain code, and an inclusion relationship of the determined closed loop is determined. Then, for all points on the contour line, determine whether these points are the start or end of a series of connected regions (runs) connected in the raster scanning direction in the binary image. Determine whether the point is a point or any other point from the chain code relationship between the points before and after it, find the coordinate values of the start point and end point on the image, rearrange them in ascending or descending order, and then set the feature parameters. A method for calculating feature parameters using calculations.