JPH0139151B2 - - Google Patents

Description

[Detailed Description of the Invention] [Technical Field] The present invention relates to a shape recognition device that uses a visual sensor to calculate numerical data, that is, feature quantities, determined from the shape of an object such as the length, area, and center of gravity of the object. It is something.

[Prior Art] The configuration and operating principle of a conventional shape recognition device will be explained based on the drawings. FIG. 1 is its block diagram. A is an image pickup tube as a visual sensor that optically detects objects to be measured X and Y, B is a camera interface that converts the video signal from image pickup tube A into a signal suitable for further processing, and C is a camera interface. A/D that digitizes information from B
D is an image processor consisting of a control unit that synchronizes signals with the converter and other connected components, and D is a central processing unit (hereinafter simply referred to as CPU) that controls this system and executes various calculations. ),
E is a CRT that displays image data from the image processor C on a cathode tray tube (hereinafter simply referred to as CRT) F based on the timing from the CPUD.
A controller, G is a random access memory (hereinafter simply referred to as RAM) that temporarily stores the image data from the image processor C converted into run-length data suitable for post-processing by the CPUD, and H is a random access memory (hereinafter simply referred to as RAM) of the system described above. Read-only memory (hereinafter simply referred to as ROM) that stores programs for executing functions. In the system described above, the image data of the object to be measured is converted to run-length data suitable for calculating the feature values of the object, and the next step of processing is performed based on this run-length data. That's what I do. The flow of information processing up to calculating feature amounts from this image data will be explained based on FIG. FIG. 2A shows image data from the image processor C, for example binary image data, and FIG. 2B shows run length data and its storage method according to the data.
The CPUD converts the image data matrix shown in Figure A into run length data and stores it in the RAMG. As shown in Figure 2A, the objects to be measured X and Y are distinguished in the image data by space and their contours, and are expressed as binary information, either in space or inside the contours of the objects. . At this time, the shape recognition resolution of this device is determined by whether the measurement range is subdivided into several parts to determine whether it is a space or an object. Run-length data treats the continuation of the same type of data as information, for example, Figure 2A.
The run length data of the image data matrix is expressed as shown in FIG. 2B. The figure shows a case in which the image data shown in FIG. 2A is treated as one matrix and is converted into run-length data in descending order of rows. The amount of information held as one run length data is in what row of the image data matrix is the information, and what column is the youngest column in which the image data is displayed (hereinafter simply referred to as the starting point). Three types (a _o , b _o , c _o in Figure 2 B) of how many columns the image data continues to be displayed (hereinafter simply referred to as the end point)
This is the information. As is well known, since the object to be measured is captured by the image pickup tube A by scanning with an electron beam, video signals are serially input to the image processor C in accordance with the order of horizontal scanning lines.
The image processor C binarizes this series of video signals as either a signal indicating a space or a signal indicating an object, and converts it into a digital signal that can be easily processed by the CPUD. Based on the binarized image data for each horizontal scanning line, the CPUD starts timing from the scanning start time of each horizontal scanning line, measures and stores the starting and ending points of the signal indicating the object. . For example, in the image data of FIG. 2A, conversion to run-length data starts from the youngest row in which objects ) and end point (clock signal 16
One run length data is formed by three pieces of data (a ₁ , b ₁ , c ₁ in FIG. 2B). Similarly, if the data is converted in the ascending order of the rows and there are two or more types of image data in the same row, they are stored in the memory in the descending order of the starting point and ending point (in Figure 2B, relationship between a ₂ , b ₂ , c ₂ and a ₃ , b ₃ , c ₃ ). When the CPUD completes the search for the entire image data matrix as described above, the processing for that one screen ends. The run length data row shown in FIG. 2B created in this way is used as data for the next connectivity analysis, and various feature quantities of the objects to be measured X and Y are calculated. This completes a series of operations.

However, the shape recognition device described above has a slow processing speed from run-length data to the next stage of connectivity analysis because the data is serially continuous, and is not suitable for real-time processing where the time to process one screen is limited. The adaptability of these materials was low.
In addition, using a computer with high-level functions with large processing power for this purpose directly leads to an increase in cost and equipment labor, which is in some ways contradictory to the purpose of popularizing shape recognition devices. Therefore, there has been a strong desire for a shape recognition device that does not require such equipment and has a high processing speed.

[Object of the Invention] The present invention was made in response to the above-mentioned needs, and an object of the present invention is to provide a shape recognition device that is capable of efficiently processing various data and has a high processing speed.

[Configuration of the invention] The configuration of the present invention for achieving the above object is described in the third section.
This will be explained based on the diagram.

a visual sensor that detects the shape of an object to be measured; an image processing means that converts a shape signal corresponding to the object from the visual sensor into an image data matrix; run-length data generation means for converting into run-length data prioritized from the end; run-length data storage means comprising a plurality of blocks assigned scanning orders and priorities; run-length data distribution means for distributing and storing run-length data for each block assigned a corresponding scanning order and priority in the run-length data storage means; The gist of the present invention is a shape recognition device characterized by comprising a run length data processing means for performing analysis and calculating a feature quantity recognized from the shape of the object to be measured.

[Example] [Configuration of Example] Hereinafter, an example of the present invention will be described in detail with reference to the drawings.

FIG. 4 shows a block diagram of the entire configuration of this embodiment. First, its structure will be briefly explained, and then its structure and operation will be explained in detail while illustrating the flow of information based on that structure.

1 is an image pickup tube; 2 is a camera interface that formats the analog signal from the image pickup tube 1 to be suitable for subsequent processing; and 3 is a camera interface 2.
4 is a CRT controller that causes the CRT 5 to display visually based on the output from the signal converter 3; 6 is the signal converter 3;
7 is a horizontal synchronization separator that separates the horizontal synchronization signal of the image data from the signal converter 3, and 8 is a clock counter that generates a pulse for each division of image data. In accordance with the input of the horizontal synchronizing signal separated by the separator 7, it is reset and starts counting. 9 is a field determiner that determines whether the horizontal synchronization signal from the horizontal synchronization separator 7 is an odd number or an even number when image data is sent by interlaced scanning; 10 is a field determination device; It receives a frame signal from the device and is reset for each screen.
For each frame, an address for storing two types of data, the start point and end point of image data that constitute run-length data, is generated from two inputs: the horizontal synchronization signal from the horizontal synchronization separator 7 and the pulse from the delimiter pulse generator 6. 11 is an address switching device which stores the run length data in the run length data storage section 15, which will be described later, based on the field information from the field determiner 9 and the address generated by the address generation counter 10. In addition to specifying the control circuit 1 described later
6, and addresses the run-length data stored in the run-length data storage section 15 as described above based on the address designation signal. 12 is a limit circuit which outputs a gate closing command to the gate circuit 13 when a predetermined number or more of run length data is generated in one field. The gate circuit 13 receives pulses from the delimiter pulse generator 6, and outputs the input as is to the latch circuit 14 if the gate closing command signal is not input. 1
4 inputs the clock signal from the clock counter 8, and this input signal is sent to the gate circuit 13.
The latch circuit 15 outputs the pulse to the data bus of the run length data storage section 15 at the same time as the input pulse from the latch circuit 1 to the address specified by the address switch 11 as described above.
A run-length data storage section 16 consisting of a known random access memory that stores the information specified in 4 controls each of the above-mentioned components as described above, and is stored in the run-length storage unit group 15. This is a control circuit that performs connectivity analysis based on run length data. That is,
To roughly describe each of the above components in terms of their operation, they include the image pickup tube 1 as a visual sensor, the camera interface 2 and signal converter 3 as image processing means, and the clock counter 8 and latch as run length data generation means. circuit 14, run-length data storage section 15 as run-length data storage means, delimiter pulse generator 6 as run-length data distribution means, horizontal sync separator 7, field determiner 9, address generation counter 10, and address switch 11 , and a control circuit 16 which corresponds to a run-length data processing means and controls each of the above-mentioned parts.

[Operation of each component] Next, the flow of signals between the components described above will be illustrated, and the operation thereof will be described in detail. First, FIG. 5A is a diagram showing the process of converting the shapes of the objects to be measured X and Y into electrical signals using the image pickup tube 1. As shown in the figure, the image pickup tube 1 scans the image of the object to be measured with a single electron beam in the horizontal and vertical directions of the drawing, and outputs an analog signal corresponding to the brightness, darkness, and color of the image of the object to be measured at that time. . Electron beam scanning is
The sequence shown in the diagram is 1-1' → 2-2' → 3-3'...The dotted line in the diagram is called the retrace line, which runs from the end point of the scanning line to the next scanning start point. It represents the scanning up to and is unrelated to the video signal. Therefore, the signal from the image pickup tube becomes a series of analog signals as shown in FIG. 5B. In FIG. 5B, the pulse a output for each retrace line is called a horizontal synchronizing signal, and is a signal for synchronizing the video signal.
Pulse b at the end of the 512th horizontal scan is called a vertical synchronizing signal, and together with the horizontal synchronizing signal, synchronizes the video signal. This video signal is binarized by the next-stage signal converter 3 as shown in FIG. 5C, and converted into three types of information: a signal representing space or an object, and a synchronization signal. The image data created in this way is transmitted to the horizontal sync separator 7 as shown in FIG.
It is converted as shown in FIG. 5E. Here, the pulse generation point by the delimiter pulse generator 6 is the fifth point.
As shown in Figure F, it represents the boundary between the space and the object, that is, the outline.

Through the above operations, we have shown that the video signal can be subdivided into elongated rectangles whose long sides are in the horizontal direction of the drawing (image data for each scanning line, Figure 5 F).Next, we can subdivide this elongated rectangle in the vertical direction of the drawing. The operation of obtaining the target image data matrix will be explained.

As shown in FIG. 5A, it takes a certain amount of time to scan the electron beam from left to right in the drawing (hereinafter simply referred to as horizontal scanning). Therefore, the purpose of subdivision described above can be achieved by regarding the information within this fixed time as a collection of information within a shorter predetermined time (FIG. 6). In this example, the time required for horizontal scanning is set to be the same as the number of horizontal scans for one screen.
When Tm (s), a clock counter 8 which oscillates at a time expressed by Tn = Tm/512 (s) is used, the counter is reset for each horizontal synchronization signal, and time is measured every time a new horizontal scan is performed. and has achieved its purpose. That is, the count number of the clock counter 8 is always input to the latch circuit 14, and when the pulse signal from the delimiter pulse generator 6 is input to the latch circuit 14 via the gate circuit 13, the latch circuit 14 It detects the count number input in the .
By the above operation, the video signal (Fig. 5A) is 512
Conversion to a 512 row image data matrix (FIG. 6) is accomplished.

As a second step, a method for converting the image data matrix created as described above into run-length data and storing it in the run-length data storage section 15 shown in FIG. 7A will be described.
FIG. 7A shows a large number of blocks in the run-length data storage section 15 (a ₁₁ , a ₁₂ , ... a _512,64 in the figure).
The number of rows indicates the horizontal scanning order of one screen of the image pickup tube 1, and the number of columns indicates the priority assigned to each block. In this example, the processing of run-length data will be described below with respect to one screen taken in from the image pickup tube 1 as one unit, so the number of lines in the run-length data storage section 15 shown in FIG. 7A is 512. Further, although the number of columns is arbitrary, in this example, it is set to 64 for reasons described later. On the other hand, as explained in the conventional example, run-length data is expressed by consecutive rows of elements of an image data matrix, their starting points, and ending points. When configuring , it can be considered that the memory address itself already contains information about the row of the three elements of run-length data,
It is clear from the figure that conversion to run-length data can be achieved simply by storing other information such as the starting point and ending point in the run-length data block of each run-length data storage unit 15 corresponding to each row of the image data matrix. It is. As mentioned above,
The method of storing run length data in the memory in this embodiment is as follows:
Due to the special arrangement of 5, special circuitry is required. The components that perform this operation are the delimiter pulse generator 6, horizontal sync separator 7, field determiner 9, address generation counter 10, and address switch 11 shown in FIG. First, the functions of these components will be explained. In response to the horizontal synchronizing signal from the horizontal sync separator 7, which is input to the field discriminator via the clock counter 8, the field discriminator outputs the following two types of information. One of them is to send information (hereinafter simply referred to as frame information) as to whether one screen has finished, that is, whether 512 horizontal synchronization signals have been input, to the address generation counter 10.
The seed is a signal (hereinafter simply referred to as field information) that determines whether the input horizontal synchronization signal is an odd number or an even number, and is output to the address switch 11. The frame information is used to convert the image data matrix into run length data for each screen, and is used to reset the function of the address generation counter 10. Here, the address generation counter 10 operates as described below. First, the start of the first horizontal scan of one screen is determined based on the frame information and the horizontal synchronization signal from the horizontal synchronization separator 7, and the address of the run length data block indicated by _a11 in FIG. 7A is generated. This run-length data block is composed of two addresses as shown in _FIG . From the starting point data address of , to the end point data address in a ₁₁ with the next pulse input, to the starting point data address in a ₁₂ with the next pulse input, sequentially a _1,1 , a _1,2 , a _1,3 ,... , a Count up the address until the delimiter pulse continues to be input up to _1,64 . At this time, when a horizontal synchronization signal is input from the horizontal synchronization separator 7, the address indicated by the address generation counter changes to _a21 shown in FIG. By inputting pulses, the starting point data address in _A21 , the ending point data address, the starting point data address in _A22 , and the ending point data address are sequentially changed.
Count up the addresses until the separator pulses continue to be input until a _2,1 , a _2,2 , a 2,3 , ..., a _2,64 , and repeat the above operations for each horizontal scanning line, all 512 _, for one screen. Finish the conversion.

Next, if the image data is input to the address generation counter by interlaced scanning instead of sequential scanning, the field information cannot be calculated based on the actual video signal by simply configuring the count-up according to the simple pulse input described above. Since it is not possible to specify an address for a run-length data block, the address information generated by the address generation counter is changed as appropriate depending on the field information, that is, the information on whether the horizontal scan is an odd number or an even number. It is for converting. The address switch 11 has this function.

In this way, the field information and frame information control the regular storage order of run-length data as described above. The data stored in the address designated by this method is the count number of the clock counter 8 latched by the latch circuit 14 described above, in other words, this is the count number of the image data matrix in which the delimiter pulse is generated. This indicates the column number.

The function of the limit circuit 12 and gate circuit 13 is, for example, when the number of pulses from the delimiter pulse generator for each horizontal scanning line is abnormally large and cannot be stored in the 64 units of the run-length data storage unit group, the data is This is to prevent the overflow of water. In other words, when the limit circuit 12 inputs a pulse indicating 64 run length data or more, it outputs a gate closing command to the gate circuit 13,
The gate circuit 13 closes the gate in response to the gate, and thereafter prohibits output of pulses from the delimiter pulse generator 6 to the latch circuit 14. As a result, the latch circuit 14 does not output its latch data to the run-length data storage unit group, thereby protecting against memory overflow and the like. In addition, the number of run-length data for each horizontal scanning line is 64 in this embodiment, and the number of divisions of the horizontal scanning number (the number of columns of the image data matrix).
is set to 512 by the frequency Tn of the clock counter 8. Therefore, the above limit circuit operates only when objects and spaces are displayed in less than 8 columns (512 ÷ 64 = 8 columns), but in reality, with a small amount of information of about 8 columns, the shape of the object is It is difficult to specify the shape recognition device, and in such a case, sufficient recognition operation cannot be expected as a shape recognition device.
Furthermore, in such cases, there is often a problem with the setting of the threshold value for digitization by the signal converter 3, so it is necessary to check the system function, so the configuration of the limit circuit 12 is changed. is adopted.

[Connectivity Analysis of Example] Next, a connectivity analysis is performed using the run-length data stored in the run-length data storage unit 15 as described above, and various feature quantities of the target object to be measured are determined. The operation of each part regarding calculation will be explained.

First, Figure 8 shows the main flowchart of connectivity analysis, and Figure 9 shows the flowchart of connectivity analysis for each horizontal scanning line of the image data matrix.We will briefly explain the function of each step in these two flowcharts. explain.

In FIG. 8, step 100 is a step for performing initial settings, and a step for setting buffers, etc. that will be used for connectivity analysis thereafter, and step 200 is a step for determining whether connectivity analysis has been performed for all horizontal scanning lines in one screen. The judgment step 300 is a judgment step of whether or not the content of the run length data is zero. Step 400 is a step for performing connectivity analysis for each horizontal scan, the details of which are shown in FIG.
0 is for updating the scan line of the image data matrix for which connectivity analysis is performed in step 400. Step 600 also increments the number of scan lines on which connectivity analysis is performed.

FIG. 9 is a flowchart that further details step 400 in FIG. ), 402 is a step for determining whether or not there is connectivity, and 403 is a step for storing label numbers and priority numbers of the run length data storage unit 15 in a buffer, which will be described later, when there is connectivity. Step 404 counts the number of run-length data blocks with connectivity. 405 is the steps from 401 to 404 above.
Step 406 determines whether all run length data in the BK buffer has been detected. 406 is a step to determine whether connectivity has been detected by performing steps 401 to 405. Next step 407 is to assign a new label number. Step 408 is a step of determining whether one or more pieces of connectivity exist.
9 is an addition step to the feature amount buffer, which will be described later;
410 is a step for updating the RL buffer and feature buffer, which will be described later. Step 41
1 is a step for updating the run length data block for which connectivity analysis is performed in steps 401 to 410, and 412 is a step for determining whether the data updated in step 411 is zero.

The operation of each step in the flowchart of connectivity analysis has been briefly explained above, and below, detailed explanation will be given for each step regarding the actual content setting and updating.

First, various buffers used to perform the connectivity analysis according to the flowcharts of FIGS. 8 and 9 will be described. Fig. 10 shows the various buffers. Fig. 10A shows the BK buffer, Fig. 10B shows the RL buffer, Fig. 10C shows the priority order buffer, and Fig. 10D shows the label buffer and the first buffer.
Figure 0E shows the feature amount buffers. 10th
As shown in the figure, the BK buffer in FIG. It also consists of 64 prioritized run-length data blocks. Similarly to the blocks in the run-length data storage section 15, each of these blocks is configured to be able to store information about the start point and end point, which is run-length data. 10th
The RL buffer in FIG. B is a buffer consisting of label number storage units that correspond one-to-one to each run-length data block of the BK buffer.
Here, the label number is a number virtually assigned to each piece of run length data, and is like a temporary name of the object to be measured, which is assigned as described later when performing connectivity analysis. The priority buffer shown in FIG. 10C is a buffer that stores the priority number of the block that stores the run-length data when connectivity analysis is performed and run-length data with connectivity exists in the BK buffer. be. Also, the first
The label buffer shown in FIG. 0D has a storage area that corresponds one-to-one with the priority buffer, and is a buffer that stores the label number assigned to the run length data stored in the priority buffer. As shown in the figure, the feature buffer shown in FIG. 10E stores various types of information in units of label numbers. As will be described later, data subjected to connectivity analysis is assigned a label number having a pseudonym meaning, and each label number has a feature amount buffer shown in FIG. 10E. When there are multiple pieces of data with the same label number, the amount of information already stored in the feature buffer is updated based on the next newly assigned data with the same label number. After all, it can be said that it is a buffer that calculates the feature amount for each label number.

Using the above five types of buffers and the run-length data stored in the run-length data storage unit 15, processing is performed based on the flowcharts shown in FIGS. 8 and 9 to calculate the desired feature quantity. Do it. Here, as an example, a case will be described in which connectivity analysis of the image data matrix and its run length data shown in FIG. 11 is performed based on the flowchart described above. 11A and 11B respectively show the storage format of the image data matrix of the object to be measured, . . , and its run length data. The method of generating and storing run-length data has already been described in detail, but the Arabic numerals 1' to 31' in the figure indicate the divisions when dividing image data into run-length data, and 1'' to 31''. indicates run length data corresponding to that classification.
Further, the contents of the data 1'' to 31'' are, for example, run length data 1'' of the image data of section 1' having two types: the column number x which is the starting point, and the column number y which is the end point.

The run length data thus created is processed as follows by the processing means shown in the flowcharts of FIGS. 8 and 9.

First, step 100 shown in FIG. 8 is executed, the contents of all buffers shown in FIG. 10 are cleared, and preparations are made for execution of future processing. Next, in step 200, it is determined whether future connectivity analysis performed for each horizontal scan of the image data matrix has been performed for all scan lines. That is, since the number of scanning lines in this embodiment is "512", which is determined based on the number of horizontal scanning lines of the image pickup tube 1 shown in FIG. When compared, if N>512 is true, the process moves to step 700 and the series of operations ends. Also N＞
If 512 is false, it is determined that the connectivity analysis for one screen has not yet been completed, and execution moves to the next step 300.

In step 300, a run length data string Y (11th scan) corresponding to the data of the Xth scan (FIG. 11) of the currently referenced image data matrix is
) is called into the control circuit 16, and it is determined whether the data is zero or not. If it is zero, the process moves to step 600, and if it is not zero, the process moves to step 400, where connectivity analysis is performed based on the data value.

At step 600, a variable N set to the number of rows of the currently referenced image data matrix is incremented, and the process proceeds to step 200 described above.

On the other hand, when the connectivity analysis is completed in step 400, step 500 is executed, the contents of the BK buffer described above are updated to the contents of the run length data string Y that is currently being referenced, and the process moves to step 600, and the same process is performed thereafter. The action is performed repeatedly.

The above is the process executed in the main routine, and the connectivity analysis shown in step 400 will be explained in more detail with reference to FIGS. 9 to 11.

First, when execution moves to step 400 of the main routine, step 401 shown in FIG. 9 is executed, and the contents of the BK buffer are read into the control circuit 16. Here, if the currently referenced run-length data string is P shown in FIG. 11B, there is no image data before that as shown in FIG. 11A, so all run-length data blocks have zero data. Therefore, BK
I also remember the contents of Batsuhua as zero.

Next, step 402 is executed, and the run length data indicated by the priority level "1" in the P column is sent to the control circuit 1.
6 reads and checks the connectivity with the previous BK buffer.

Here, to check the connectivity, as shown in Figure 12, the starting point (x _o ) and ending point (y _o ) stored in the run length data block of the P column currently being referred to.
is calculated by the first equation below using the starting point (x _o-1 ) and ending point (y _o-1 ) in the run length data block of the Z column that was referenced earlier and is currently stored in the BK buffer. It is to judge.

x _o < y _o-1 and x _o-1 < y _o (1) In other words, the one that satisfies the first equation has an overlapping part of two run length data as shown in Fig. 12, and this overlapping part is A predetermined number (1 in this example)
It is set to . ), then it is assumed that the run length data (x _{o -1} , y _{o -1} ) of the Z column and the run length data (x _o , yo ₎ of the P column have connectivity.

However, as shown in FIG. 11, everything in the BK buffer is now zero, so there is no part that satisfies the first equation, so connectivity is not recognized, and the process moves to step 405. do.

In step 405, it is determined whether the above-described connectivity analysis has been performed on all run length data in the BK buffer shown in FIG. 10A.
In this example, since all the data in the BK buffer is zero, the process moves to the next step 406.

Step 406 is the same as Step 10 of the previous initial setting.
This shows the process of determining whether the contents of the contact number counter, which was set to zero in 0, is zero, but since the data in the BK buffer is zero, there is no change in the contents of the counter. Step 407
is executed, and the latest label number is assigned to the currently referenced run-length data. In other words, the latest label number "1" is added to the run length data "1".
is tentatively attached. Next step 409
is executed, various feature quantities calculated by the label number and run length data "1" given in step 407 to the feature quantity buffer shown in FIG. Reset the counter.

The connectivity analysis of one run-length data is completed above, and then in step 411, the priority of the run-length data string P being referenced is incremented, and the run-lengths stored in the block with the priority are incremented. Load data.

In the following step 412, it is determined whether or not the read run length data is zero. If it is not zero, the process returns to step 401 and the same processing as described above is performed. In this example, the run length data of priority number "2" in column P is zero, so the above-described connectivity analysis process for each row is completed and the process returns to step 500 of the main routine. Step 5
At step 00, the contents of the BK buffer are updated from the data in the Z column to the currently referenced data in the P column, and the process moves to step 200.

Next, the process moves on to the Q column in the run length data storage section 15. In this process as well, steps 200 and 300 are the same as in the case of the P column, so they will be omitted and the explanation will start from step 401.

First, the contents of the BK buffer read in step 401 are the run length data of line P, ie, "1", by executing step 500.
The data to be referenced is "2". Then, in the next step 402, the connectivity regarding the run length data of "1" and "2" is determined. In this case, as is clear from the first equation, it is determined that there is connectivity. In other words, it is determined that the run length data "1" and "2" represent the same object to be measured, and the process moves to step 403.

In step 403, the priority ("1") and label of the run length data (in this case, data of "1") for which connectivity has been recognized are entered in the priority buffer and label buffer shown in FIG. 10C and D. A number (“1”) is stored respectively.

Next, in step 404, since connectivity is recognized, the connectivity counter is incremented to 1.
Set to . In the next step 405, it is determined whether all connectivity analysis has been performed for the data in the BK buffer ("1"), but in this example, since only this "1" has been performed, the step The process moves to 406. In step 406, it is determined whether the contents of the previously set concatenated number counter are zero or not, and since it is not zero, step 408 is executed.

In step 408, it is similarly determined whether the connected number counter is 1 or not.

Step 409 is a step for storing the data in the feature buffer as described above. In this case, since the run length data "2" has connectivity with "1", it is the same as "1" and is stored in the feature buffer. Step 403
The label number "1" stored in the label buffer is added to the label number "1" that is already stored in the buffer with the label number "1" of the feature quantity buffer.
The content of is updated based on the data content of this "2", and the feature obtained from the two data "1" and "2" is updated with the label number "1" of the feature buffer. The content shall be within the buffer. Thereafter, the connected number counter, which was set to 1 as described above, is reset and the process moves to step 411. Thereafter, steps 411 to 600 and steps 200 to 300
The steps up to this point will be omitted since they repeat the same operations as described above.

As described above, a series of operations of this connectivity analysis are performed for each column corresponding to one scanning line of run-length data, and in step 200, all columns of run-length data, in this example, " Step 200 when the connectivity analysis is completed for "512"
From there, execution moves to step 700, and the present connectivity analysis ends. In the case of this example shown in FIG. 11, it is clear that the process is performed up to row U of the run-length data according to exactly the same procedure as explained above, and the final analysis is performed by the procedure of this connectivity analysis up to row U. Four types of buffers with label numbers ``1'', ``2'', ``3'', and ``4'' are used for feature quantity buffers, which are standard data, and each feature quantity buffer has a label number as shown in Figure 13. It can be seen that the feature amount calculated from the run length data is stored.

Finally, as can be seen from Fig. 13, the label number, which can be called a temporary name, attached to the object to be measured is different from the one given to the object to be measured when the object has a complicated shape like the object shown in Fig. 11A. Length data columns S to U
As far as we can see up to the line, it is determined as if two objects exist, and in the connectivity analysis procedure, one object has two or more label numbers (label number "2" in this example). ，
The processing procedure when "3") is assigned will be described.

In this example, it is only when the connectivity analysis is performed on the V row of the run length data shown in FIG. 11B that it can be determined that the label numbers "2" and "3" are the same object. Here, detailed explanations of the steps that have already been described and explained will be omitted.

First, when the V row of run length data is referenced, the BK buffer and
The contents of the RL buffer store the data immediately before the V row, that is, the U row, and have changed to the states shown in FIGS. 14A and 14B. Therefore, the run length data ``16'' with the priority level ``1'' in the V row is referred to, and as the first stage of this connectivity analysis, in step 402, the leftmost run length data 12' of the BK buffer is referred to. When the connectivity of
03, the priority of the run length data "12" is stored in the priority buffer and the label number is stored in the label buffer as shown in FIG. 15A.

Next, step 404 is executed, and the concatenation number counter is set to "1". As shown in Figure 14, the contents of the BK buffer include run length data up to priority level 4, so the next step 40 is performed.
5 is executed, the execution returns to step 401, and the connectivity between "13" in the BK buffer and the reference data "16" is checked in the same way as before, and the contents of each buffer are shown in FIG. B, and the connected number counter is set to "2".

Perform the same operations as above (steps 401 to 405).
This is performed for the data “14″ and “15″” in the BK buffer, but as is clear from the first equation, these three types of data are judged to have no connectivity with the reference data “16″, The contents of the priority buffer and label buffer remain unchanged as shown in FIG. 15B, and since the concatenation number counter is set to "2", step 410 is executed next. Here, label numbers are sorted and feature quantities are calculated as described below. First, from the contents of the label buffers, it is determined that the various feature quantities stored in the feature quantity buffers with label numbers 2 and 3 are related to the same object, and among the feature quantity buffers shown in FIG.
1 newly updated by combining the feature values and run length data 16′ of label numbers “2” and “3”
Assume that there are two features.

Next, the updated feature quantity is stored in the feature quantity buffer indicated by the smallest label number in the label buffer (FIG. 16). The subsequent processing is the same as described above.

As explained above, according to this connectivity analysis,
No matter how complex the shape, the various feature quantities of the object to be measured, which is the final goal, are stored in the feature quantity buffer for each object, and the content is updated each time, making accurate shape recognition possible. It turns out that it is possible.

[Effects of the Embodiment] As described above in detail regarding its configuration, operation, and processing process, this embodiment uses the run length data storage unit 15 as shown in FIG. Data is regularly stored in the storage unit 15 as described above. Therefore, the procedure for performing connectivity analysis based on the run length data stored in the storage unit 15 can be performed using only extremely simple processing steps as shown in the flowcharts of FIGS. calculation can be accomplished in a short time.

In other words, as explained using the example of the flow chart shown in FIGS. 8 and 9, when performing connectivity analysis, the run-length data to be referenced and the run-length data to be compared for the presence or absence of connectivity are , it is possible to calculate the feature amount by simply executing it on the currently referenced run length data stored in the BK buffer and the run length data for which connectivity analysis was performed immediately before, so the absolute processing The procedure is simplified and therefore requires less time for the process.

[Effects of the Invention] As described in detail above, the shape recognition device according to the present invention stores run-length data that is prioritized and scan-ranked in advance according to an image data matrix when storing run-length data. A block is provided, and run length data is regularly stored in the block. Therefore, the purpose of the connectivity analysis procedure is achieved by simply repeating the same routine for each piece of run length data with the same scan ranking. In other words, when performing connectivity analysis, it is only necessary to check the connectivity analysis between the minimum necessary run length data, and since the same routine is used repeatedly for the connectivity analysis, the analysis procedure can be simplified. It will be done. Therefore, it is possible to easily calculate the feature amount, which is the ultimate goal, and reduce the required time, and it is possible to provide a shape recognition device suitable for real-time processing and the like where the time for calculating the feature amount is limited. This is advantageous in terms of cost and maintenance compared to conventional systems in which processing was performed by a computer with higher-level functions in order to improve analysis speed, and it is encouraging the spread of shape recognition devices. It also has secondary effects such as encouraging the use of the technology and expanding its field of use.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a conventional shape recognition device, FIG. 2 is a diagram showing a method for generating run length data, FIG. 3 is a configuration diagram of the present invention, and FIG. 4 is a configuration of an embodiment of the present invention. 5A to 5F are explanatory diagrams of the operation, FIG. 6 is a schematic diagram showing the method of generating the image data matrix, FIGS. 7A and 7B are schematic diagrams showing the run length data storage section, and FIG. Figure 8 is a flowchart showing the main routine of the connectivity analysis, Figure 9 is a flowchart of the connectivity analysis for each line, and Figures 10A-
E is a schematic diagram showing the various buffers, FIG. 11 is a schematic diagram showing an example of the image data and run length data configuration, FIG. 12 is a diagram explaining connectivity, and FIG. 13 is an example of the content of the feature buffer. A schematic diagram, FIG. 14 is a schematic diagram showing an example of the contents of the BK buffer and RL buffer, FIGS. 15 A and B are schematic diagrams showing changes in the priority buffer and label buffer,
FIG. 16 shows schematic diagrams showing changes in the feature amount buffer. 1, A... Image pickup tube, 15... Run length data storage unit, 16... Control circuit, 7... Horizontal sync separator, 6
...Separator pulse generator, 9...Field judger,
10...Address generation counter, 11...Address switch, 14...Latch circuit.

Claims (1)

[Scope of Claims] 1. A visual sensor that detects the shape of an object to be measured; an image processing means that converts a shape signal corresponding to the object from the visual sensor into an image data matrix; and the image data matrix. run-length data generating means for converting the data into run-length data prioritized from the scanning start end for each scan; run-length data storage means comprising a plurality of blocks assigned scan orders and priorities; run-length data distribution means for distributing and storing the run-length data generated by the length-data generation means for each block assigned a corresponding scanning order and priority in the run-length data storage means; A shape recognition device comprising: run length data processing means for performing connectivity analysis based on the data and calculating a feature amount to be recognized from the shape of the object to be measured.