JPS63141190A - Method for extracting isolated drawing - Google Patents

Abstract

PURPOSE:To execute a processing at high speed by using an operator including an isolated drawing and executing the scanning. CONSTITUTION:In a longitudinal direction by latches 11, 12 and 13 of the number equal to the number of picture elements of the longitudinal direction of an operator, the data for one picture element are read from a frame memory, for the part of an intermediate 12, OR is obtained by an OR gate 14 and the part is added to a latch 15 together with the 11 and 13. The output of the latch 15 is successively added to a latch 16, a register 17 and latches 18 and 19, and the output of the latches 15, 16, 18 and 19 and shows the upper part, intermediate part and lower part values respectively at the time of t+1, t, t-u-1 and t-u. The number of steps between latches 15-19 shows the number of the picture elements in the lateral direction of the operator. Respective outputs of the latches 15-19 are added to an arithmetic circuit 20, further, the value Qt of a frame at the time of (t) is added from an 8-bit register 21 and the value Qt+1 of the frame at the time of t+1 is obtained.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention extracts isolated figures (characters and isolated symbols) by separating them from other continuous lines and symbols in automatic reading of drawings. It is about the method.

(Prior Art) Recently, in the field of CAD or automatic drafting of drawings, drawings are automatically read and this data is processed in various ways to obtain desired results. at present,
Drawings that can be automatically read are written with limited conventions and are limited to binary screens with a background of 0 and recorded portions of 1.

In order to automatically read a binary drawing written in this way and further separate lines from characters or symbols, the drawing is scanned by an operator in 3 rows and 3 columns, and if a pixel of 1 is found, that pixel is used as the starting point. Generally, this is done by performing propagation processing as , determining and labeling the connected components, repeating this operation one after another to label all connected components, and then examining the properties of each connected component. It is. Here, a connected component refers to an area surrounded by continuous border lines, and the propagation process refers to a region surrounded by continuous boundaries, and the propagation process is a component that initially becomes 1 and is labeled with a certain value.
This refers to moving the center to the center pixel of the operator in row 3 and column 3 to a pixel that also takes 1 in the other 8 pixels, giving the same label, and repeating this operation one after another.

The above labeling method involves determining the first connected component, removing this connected component from the drawing, performing the same operation on the remaining drawings to determine the second connected component, and repeating this operation. It is common to determine all connected components. Since this method takes too much processing time, the applicant has disclosed in Japanese Patent Application No. 61-140285 "High Speed Labeling" a method that can complete labeling with several scans.

(Problems to be Solved by the Invention) When reading figures using the method described in the previous section, propagation processing is performed for each connected component, which requires a lot of processing time, and recognition of characters and symbols requires labeling. Since it is carried out after the completion of the process, there is a disadvantage that the processing time becomes even longer. Although the high-speed labeling proposed by the applicant can reduce the number of scans, there is a limit to the reduction in processing time since the propagation process remains the same. Next, when there are a large number of connected components, there is a problem in that a multi-valued memory is required to distinguish between them. For example, if the number of connected components is about 100, 6-bit memory is sufficient, but
If the number is around 1000, 10-bit memory will be required. Furthermore, since it is necessary to store each connected component, the required memory capacity becomes extremely large, which inevitably increases the cost of the device.

SUMMARY OF THE INVENTION An object of the present invention is to provide a method that can solve the above-mentioned problems and directly extract isolated figures by one scan, thereby completing the process in an extremely short time.

Another object of the present invention is to provide a method in which the number of memory bits is only 1 bit and the capacity of the memory used is small.

(Means for Solving the Problems) The isolated figure extraction method of the present invention scans the drawing with an operator whose size is slightly larger than the size of the isolated figure, and the sum of pixels in the frame of the operator is equal to the background value. , when the sum of pixels in the portion inside the operator is different from the value of the background, the figure inside the operator is determined to be an isolated figure.

(Function) A figure is a collection of lines, characters, and symbols, but characters are always isolated, and some symbols are isolated and some are not. Also, both have a size limit. In comparison, lines are always continuous and long.

The present invention extracts isolated figures by utilizing the properties of these general figures. To do this, create an operator of an appropriate size that includes characters or symbols, scan the figure with this operator, and if the background value is a binary image with a value of 0, all pixels on the operator's frame will be O and will be inside the frame. When there is a 1 part, this 1 part is isolated and is a character or symbol, so it is extracted.

In order to extract isolated shapes using this operator, the shapes must be written according to a certain convention,
In the present invention, ■ the thickness of the characters must be less than a predetermined size (30 to 20 mm in the case of the example), ■ they must not be written on the edge of the paper (those below 3 nua are invalid), and ■ the characters must be There are three restrictions: it must not touch other characters, symbols, or lines. The operator is generally square, but depending on the shape of the isolated figure, it may be a vertical or horizontal rectangle, or may have other shapes. The extracted figure is erased by writing its address into the stack memory, and the remaining figures are scanned in the same manner to extract the next isolated figure. Thereafter, similar operations are repeated to determine the addresses of all isolated diagrams.

The extracted isolated figures are compared with character patterns to recognize each character, and those that are not recognized as characters are treated as symbols. Character recognition may be performed using a known method.

The above-mentioned extraction of isolated figures can be performed only by software, but it is undesirable for figures with dense lines, characters, and symbols because it takes time and slows down the processing speed.
In the embodiment, it is configured by hardware.

(Example) Examples will be described in detail below with reference to the drawings.

When inputting data of an input figure, it is necessary to select an imaging device with a resolution and scanning range commensurate with the fineness and size of the figure. In the embodiment of the present invention, graphics are input using a high-resolution imaging device with 2000 pixels per scan line and 3000 scan lines. According to this device, A
In the case of a 4th edition (210 x 297 mm) drawing, if the effective range is 90%, the horizontal resolution will be approximately io lines per 1 mm, and in the case of an A3 edition (297 x 420 mm), it will be approximately 7.5 lines, so the normal It has the resolution necessary for shape recognition. Therefore, data can be input by one scan for up to A3 size. In the case of a drawing of A2 size or larger, the image may be captured in parts.

The input data is stored in a frame memory, and the isolated figure is extracted by reading the data from this frame memory. To extract an isolated figure, scanning is performed using a square operator having an appropriate number of pixels. The size of the operator varies depending on the fineness of the figure, and for A4 size drawings, three sizes are used: 30x30.25x25 and 20x20, and for A3 size, 20x20.18x18°14x14. The reason why three types of operators are used in the same drawing is that it may be difficult to separate isolated drawings with just one type of operator depending on the arrangement of lines, characters, and symbols, but the size of the figure When it is known in advance, it is also possible to select one type of operator with the optimal size and use only that operator. Moreover, the reason why the operator for the A3 size is chosen to be small is that when the figures are drawn in the same size, the number of pixels for each figure light decreases. Figure 1 conceptually illustrates the scanning situation by the operator, in which the operator scans in the horizontal direction, and after the operator reaches the right edge of the image, moves one step in the vertical direction.
Send one scan line at a time. The operator's position is represented by the coordinates of the upper left corner. Operator a is located at position (m, n), and operator b is located at position (m', n+1), which has been moved by one scanning line.

FIGS. 2A and 2B are diagrams for explaining an algorithm for detecting isolated figures by an operator. In FIG. 2(A) at time t, the values taken by the upper part 1, the lower part 2, and the middle part 3 at the right end of the operator are respectively At
, BL, and Ct. Here, 01 is the OR value of the pixel values on the vertical line excluding At and Bt, and At, Bt, C, . . . all take the value of 1 or O in the case of a binary image. If the time interval between the right end and the left end is U, the values taken by the left end 4 and 5.6 are At-, respectively.

B, -1+Ct-u. In FIG. 2 (B) at the time (t + 1), one pixel after t, the rightmost value ? , 8.9 are A I + B I + C respectively.
1, and the value of the leftmost middle 10 is CL-u+1'. and the value of the frame at time (t+1) as Q, respectively.
Q4. If the value inside the frame is RL I RL*1, the part of the frame at time (t + 1) is the value at time t, and the parts 7.8.9 and 10 are newly added. join,
4,5°6 and 3 are subtracted. For the part inside the frame, 3 is added and 10 is subtracted. Expressing this as a formula, QL-1=Qt (At-u + Bt-8"
Ct-u)+ (A, +B, +C,) ” CL-a+I Ct'−−−−・−・・−・・
(1) Rt-+ = Rc-Ct, □u+I
+C(,・−・・−・−・・(2).

FIG. 3 is a block diagram showing the operation of the isolated figure extraction circuit. Data for one pixel is read in the vertical direction from the frame memory by latches 11, 12, and 13, the number of which is equal to the number of pixels in the vertical direction of the operator. 12 part is or gate 1
Take or by 4 and latch 15 with 11.13
Add to. In this manner, scanning is performed by the operator. Let the output of No. 11 be A, the output of No. 13 be B, and the output of No. 12 be C. The output of latch 15 is the output of latch 16. register 17
．． The outputs are applied to latches 18.19 one after another, but the outputs to latches 15.16.18.19 are t+l and t+l, respectively.
t, tu+l, the upper part at the time of tu,
Indicates the middle and lower values. The number of steps between launches 15 to 19 indicates the number of pixels in the horizontal direction of the operator. The output of the latch 15 is A1+Bt, I+Ct1, s The C output of the latch 16 is Ct%, and the output of the latch 18 is CL-u.
+I, and the output of latch 19 is At-11+Bt, respectively.
-u+ct-u, so if we add these outputs to the arithmetic circuit 20 as shown in the figure and then add Q using the 8-bit register 21, we can obtain the output (112 Q).This Qt , , are fed back to the register 21 and added to the counter 21 as Qt in the calculation of the next pixel.Same as the above operation, the output of the latch 16.18 and the output of the 8-bit register 24 are added to the calculation circuit 23. By doing so, R in equation (2).

can be obtained. In the above circuit, an O value corresponding to the width of the operator is added as an initial value. Qi and kI are compared with the background value 0 by comparators 22 and 25, and the output of the comparator 22 is O, that is, the value of the frame is O, and the output of the comparator 25 is when it is not 0, that is, the inside of the frame is a value of 1 or more. In other words, when there is an isolated figure within the frame, an output is output to the AND gate 26. As described above, three types of operators are used for each size of paper, and these operators are created by changing the number of elements of the latch 12 and register 17, and each extracts an isolated figure. The above circuit is one example, and it goes without saying that various circuits having the same function can be created.

FIG. 4 is a general block diagram, and FIG. 5 is a flowchart showing the operation of this block diagram. To extract an isolated figure,
First, data is read from the frame memory 31 into the isolated diagram extraction circuit 32 mentioned above, and when an isolated diagram is extracted, the address in the upper left corner is written to the address stack RAM 33 as the operator's address, and the isolated diagram extracted by the isolated diagram deletion circuit 34 is written. The operator continues scanning from the point where the area is erased. When the end of the scan line is detected, the process advances one scan line and continues scanning by the same operator, ending when the final address of the screen is found. All circuits are controlled by the CPU 35.

After detecting the position of the isolated figure in this way, it is detected whether each isolated figure is a character or a symbol, and non-characters are treated as symbols.

In this case, a known method may be used for character recognition.

(Effects of the Invention) As explained above, in the present invention, a preferably rectangular operator that includes an isolated figure is used, and scanning is performed by a hard circuit, so that an isolated figure can be produced at an extremely high speed compared to the conventional method using propagation processing. can be extracted. The degree of speedup varies depending on the content of the drawing, but it is 1/2 to 1/10
0 speedup is possible.

Furthermore, in the past, when there were a large number of connected components, a memory with a large number of bits was required, and a memory with a large capacity was required to store the connected components as they were, but in the present invention, only the addresses of isolated diagrams are stored. Therefore, all you need is a binary memory of walking distance, and the memory capacity is approximately 1710.
It's only a matter of degree.

[Brief explanation of the drawing]

Figure 1 is an explanatory diagram showing the scanning situation by the operator, Figure 2 is a diagram to explain the algorithm for isolated figure detection by the operator, Figure 3 is a block diagram of the isolated figure extraction circuit, and Figure 4 is a comprehensive diagram. FIG. 5 is a flowchart showing the operation of this block diagram. 1.7... Upper right end of the operator 2.9... Lower right end of the operator 3.8... Middle right end of the operator 4... Upper left end of the operator 5... Lower left end of the operator 6.10. ... left end middle 11 of the operator ... upper part of the input latch 12 ... middle of the input latch 13 ... lower part of the input latch 14 ... OR gate 15.18.19 ... latch 1
7...Register 20.23...Arithmetic circuit 21, 24...8-bit register 22, 25...Comparator 26...And gate 31...Frame memory 32...Isolated figure extraction circuit 33 ... Address stack RAM 34 ... Isolated diagram erase circuit 35 ... CPU Figure 3 Ktchif < Agony

Claims (1)

[Claims] 1. The drawing is scanned by an operator whose size is slightly larger than the size of the isolated drawing, and the sum of the pixels in the part of the frame of the operator is equal to the background value, and the sum of the pixels in the part inside the operator is An isolated figure extraction method characterized in that a figure inside an operator is determined to be an isolated figure when it differs from a background value. 2. A method for extracting an isolated figure according to claim 1, wherein a binary screen is scanned by a square operator. 3. Detecting the OR value C of pixel data A at the top end of the vertical line of the square operator, pixel data B at the bottom end, and pixel data in between, and the time t and (t- u) Detecting the above c values (C_t, C_t_-_1) at (u is the time required to feed the width of the operator) and the A, B and C values at times (t+1) and (t-u+1) ( A_t_+_1, B_t_+_1
, C_t_+_1; A_t_-_u_+_1, B_t_
-_u_+_1, C_t_-_u_+_1), and the values of the operator's frame at times t and t+1 are Q_t, Q_t_+_1, and the internal values are R_t, R
When _t_+_1, the following (1) is obtained from each of the above data.
Q_t_+_1 and R_t_+_ by equation and equation (2)
1, and Q_t_+_1=Q_t-(A_t_-_u+B_t_
-_u+C_t_-_u)+(A_t_+_1+B_t
＿＋＿１＋C＿t＿＋＿１)＋C＿t＿−＿u＿＋＿１
−C_t……(1) R_t_+_1=R_t−C_t
____u_+_1……(2) From the above calculation result, Q_t
3. The isolated diagram extraction method according to claim 2, comprising the step of outputting that _+_1 matches the background value 0 and R_t_+_1 does not match 0.