JPS61294306A - Detector for rotation deviation - Google Patents

Abstract

PURPOSE:To detect the rotation deviation of an input pattern by collating input information, which is related to input pattern constituting picture elements included in a ring-shaped window, with standard information. CONSTITUTION:The image of an object 2 to be recognized is picked up by a TV camera 1 and is subjected to the binarization processing by a binarization circuit 4 to generate an input pattern having the rotation deviation of an angle. This input pattern is sent to a black picture element detecting circuit 5 to detect black picture elements included in the window, and the contents of a counter 10 are taken in at every black picture element detection time by an adder 8 to calculate a horizontal load sum. A CPU 16 stores this horizontal load sum in a RAM 17. The CPU collates stored data in the RAM 17 with standard data to detect an extent of rotation deviation.

Description

[Detailed Description of the Invention] <Technical Field of the Invention> This invention obtains an input pattern by imaging a stationary or moving object to be recognized, and compares this input pattern with a standard pattern to identify the object to be recognized. In relation to visual recognition technology for subsequent recognition, in particular, the present invention provides a novel rotational deviation detection device that rapidly detects the amount of rotational deviation of an input pattern with respect to a standard pattern prior to pattern matching.

<Summary of the Invention> The present invention sets a ring-shaped window for an input pattern, generates information regarding the pattern constituent pixels included in the window, and then creates a standard system that similarly generates this information using the window. The amount of rotational deviation of the input pattern is detected by comparing it with information, and this makes it possible to detect the amount of rotational deviation of the input pattern at high speed at video rates.

<Background of the Invention> In general, a two-dimensional visual recognition device recognizes an object by superimposing an input pattern and a standard pattern on an image, detecting the degree of superimposition and matching between the two patterns.

Therefore, when performing this pattern matching, it is necessary to accurately align both patterns. Conventionally, an XY stage or the like is used to position the object to be recognized at a predetermined stopping position, and then this is imaged with a television camera to form the input pattern. This input pattern is then matched against a standard pattern. However, in the case of this type/method, a mechanism for positioning the object to be recognized is required, which has the disadvantage that the overall structure of the visual recognition device becomes complicated, and pattern matching takes time equal to the time required for positioning operation. .

Therefore, in recent years, the amount of positional deviation and rotational deviation of the input pattern with respect to the standard pattern is detected as data, the deviation between the input patterns is corrected based on this detected data, and then pattern matching for object recognition is performed. A method was developed.

However, in the past, in order to specifically detect rotational deviation of an input pattern, the contour of the input pattern was first extracted, and then the moment of the input pattern was determined from this contour data.
This moment is used to calculate the amount of rotational deviation of the input pattern, which requires a significant amount of calculation time, making it difficult to achieve high-speed processing.

<Object of the Invention> The present invention provides a rotational deviation detection device that is capable of detecting the amount of rotational deviation of an input pattern with respect to a standard pattern at high speed, thereby enabling quick and easy object recognition even if the input pattern is tilted. The purpose is to

<Structure and Effects of the Invention> In order to achieve the above object, the rotational deviation detection device of the present invention includes a pattern generation means for binarizing an object image to generate a standard pattern and an input pattern, and a pattern generation means for generating a standard pattern and an input pattern. A window setting means for setting a ring-shaped window, and generating standard information regarding pattern constituent pixels included in the window for each standard pattern having a rotational shift of a different angle, and generating standard information for each angle for each standard pattern! standard information recording means for recording information; and a determination unit that generates input information regarding pattern constituent pixels included in a window for each input pattern, and compares this input information with the standard information to determine the amount of rotational shift of the input pattern. We made sure to have the means to do so.

According to this invention, since it is possible to correct the rotational deviation between patterns on the data, there is no need to position the object to be recognized at a predetermined stop position, a special positioning mechanism is no longer required, and the entire device is It can be simplified, the time required for positioning operations can be saved, and the efficiency of object recognition processing can be improved.

In addition, the amount of rotational deviation of the input pattern can be detected with a video radar using a simple hardware configuration such as an adder, without the need for complex calculation processing using moments, which speeds up calculation processing and improves efficiency. realizable.

Furthermore, since the data for detecting rotational deviation is imported using a window, even if the input pattern contains image noise, its influence can be reduced, and there is no risk that the noise part may be mistakenly recognized as part of the pattern. This has the remarkable effect of achieving the object of the invention, such as being able to accurately detect and correct the amount of rotational deviation.

<Description of Embodiments> FIG. 1 shows an example of a circuit configuration of a rotational deviation detection device according to the present invention. In the illustrated example, the television camera 1 images a stationary or moving object 2, for example from above, and sends an image output (shown in FIG. 2(1)) related to interlaced scanning to the synchronization separation circuit 3. The synchronization separation circuit 3 extracts a horizontal synchronization signal HD, a vertical synchronization signal VD, an odd field signal OD (shown in FIG. 2 (2)), and a clock signal CK from the image output.
(shown in Figure 2 (4)), etc., and separate the video signal VDi.
is output to the binarization circuit 4. The binarization circuit 4 is shown in FIG.
As shown in 3), a certain threshold level TH is set for the video signal VDi, and the odd fields of the video signal VDi are converted into black and white binarized to form a binary pattern and output.

Figure 3 shows an image of a binary pattern, Figure 3 + 1) shows an elliptical standard pattern P0, Figure 3 (2)
respectively indicate the same elliptical input pattern Pi. Note that in the figure, the shaded areas are black pixels forming the pattern, and the areas other than the shaded areas are white pixels corresponding to the foreground. In this embodiment, each pattern is assembled in a pixel range of 256 bits in length and width, and in the case of the human cover turn P shown in FIG. It's happening.

A black pixel detection circuit 5 for detecting black pixels in each pattern is connected to the binarization circuit 4, and a readout side of the window memory 6 is connected to the black pixel detection circuit 5. This window memory 6 is used to set a circular ring-shaped window 7 for each pattern, as shown in FIGS. 4 to 7. Standard pattern P, (4th
The width of the window 7 is set so that it touches the outline of the window (shown in the figure) and the width of the window 7 corresponds to exactly two pixels.

Thus, the black pixel detection circuit 5 detects the black pixels (indicated by diagonal lines in FIGS. 4 to 7) included in the window 7, and sends the detection signal to the adder 8. The adder 8 and the q-buffer register 9 are used to cumulatively add the horizontal addresses of the window memory 6 for one horizontal scanning line each time a black pixel is detected, and the The added value (referred to as "horizontal weight sum") is taken into buffer register 9 and held there.

In addition, in FIG. 1, a horizontal counter 10 and a vertical counter 11
is used to address the pixel position in the window memory 6 when reading and writing from the window 7, and among these, the counted value of the horizontal counter 10 (horizontal address of the window memory 6) is given to the adder 8. be. Furthermore, the gate circuits 12 and 13 are connected to the odd field signal O.
Opening/closing is controlled by D and clock signal GK, and supplies write control signal W and read control signal R to window memory 6. Furthermore, gate circuit 14 is controlled to open and close by odd field signal OD, and supplies clock signal CK to horizontal counter IO and vertical counter 11, respectively.

The cumulative addition value (horizontal load sum) of the buffer register 9 is 110 (Input 10utp) at the end of each horizontal scan.
ut) via port 15.
Processing Unit) 16, and the CPU 16 stores the imported data contents in RAM (Ra
17 and displays it on the display unit 18, and executes a series of calculations and processes related to detection of rotational deviation of the input pattern P.

In addition, in the figure, F ROM (Programmable
The ReadOnly Memory 19 stores a series of programs for detecting and correcting rotational deviations, and the RAM 17 has an area for storing various data as well as a work area for executing various processes. Further, the gate circuit 20 is a circuit for generating an interrupt signal INT to the CPU 16.

Fig. 4 shows a standard pattern P0 with no rotational deviation, and Fig. 5 shows a standard pattern P1 given a rotational deviation of angle θ.
FIG. 6 shows a standard pattern P2 given a rotational deviation of a larger angle θ2, and FIG. 7 shows an input pattern P having a rotational deviation of a certain angle θ. In each figure, 7 indicates a window set for each pattern, and the shaded area indicates the black pixels of each pattern included in window 7. Also, histograms Ha and H shown accompanying each figure.
+, Hz, and H indicate the magnitude of the horizontal load sum for each horizontal scanning line, and the area diagram of the RAM 17 indicates the storage range (shaded area in the figure) of each horizontal load sum. According to the figure, it can be seen that as the rotational deviation of the pattern becomes thicker, the shape of the histogram changes and the data storage range of the RAM 17 expands.

Therefore, prior to object recognition processing, first, a standard model with no rotational deviation, then each standard model given a rotational deviation of a different angle, and then each standard model given a rotational deviation of a different angle, as shown in Figure 1. Using the circuit shown in the standard pattern PIIIPIIP! I... is determined and standard data for detecting rotational deviation is created.

When the standard model (with no rotational deviation) is now imaged by the television camera 1, a binarization process is executed for the first odd field of the video signal vDi, and the standard pattern P0 shown in FIG. 4 is generated. This standard pattern P0 is
The black pixel detection circuit 5 detects the black pixels included in the window 7, and the adder 8 takes in the contents of the horizontal counter 10 at each black pixel detection timing and calculates the horizontal A weighted sum is calculated. And the horizontal synchronization signal H in the odd field
An interrupt signal INT is generated to the CPU 16 at each time timing D, and the CPU 16 takes in the horizontal load sum of the buffer register 9 and stores it in the RAM 17 during this horizontal synchronization period. This calculation and import of the horizontal load sum is performed line by line across all horizontal scanning lines. In this case, if the retrieved horizontal load sum has a value other than zero, the angle θ of the RAM 17 is calculated.

The data is sequentially stored in the area corresponding to the address from the beginning (however, if the value of the horizontal weight sum is zero, the data is not stored in the RAM 17 (see FIG. 4).

5 and 6 show angles θ3. Standard patterns P, , Pt given a rotational shift of θ2 are shown. For each of these patterns pt and pg, after calculating the horizontal load sum using the same procedure as above, the respective data is RA
It is stored in the area of the address corresponding to the angle θ1° θ2 of M17.

The following further different angles θ1. θ4. ..., standard pattern Pz-Pa with rotational deviation of θ7, ...
, P,1 (not shown), the horizontal load sum is calculated using the same procedure and each data is stored in the RAM 17, thereby completing the generation of standard data for detecting rotational deviation.

Next, when the object to be recognized is imaged by a television camera l, a video signal VD! Binarization processing is performed on the first odd field of , and an input pattern P1 having a rotational shift of angle θ as shown in FIG. 7 is generated. This input pattern P! is sent to the black pixel detection circuit 5 to detect the black pixels included in the window 7, and at the same time, the adder 8 takes in the contents of the horizontal counter 10 at the black pixel detection timing and calculates the horizontal weight sum. Calculated. And at each time timing of the horizontal synchronization signal HD in the odd field,
An interrupt signal INT is issued to the CPU 16, and the CPU
16 is the buffer register 9 during this horizontal synchronization period.
The horizontal load sum is taken in and stored in a predetermined area of the RAM 17. When this calculation and import of the horizontal weight sum for each horizontal scanning line is executed over all horizontal scanning lines, the CPU 16 then stores the RAM 1 related to this input pattern P in the period of the subsequent even field.
7 is stored in each of the standard patterns P (+P, ,P
2. . . . Sequentially collates with standard data regarding p7, searches for the standard data with the highest matching degree, and detects the corresponding angle (for example, θ4) as the amount of rotational deviation of the input pattern P. After that, the CPU 16 calculates this angle θ4.
Based on this, the rotational deviation of the input pattern P is corrected, and pattern matching processing for object recognition is started using the corrected pattern.

Note that in the above embodiment, the amount of rotational shift of the input pattern is determined using the sum of horizontal loads, but instead of this sum of horizontal loads, the number of black pixels for each horizontal scanning line or the average value of the sum of horizontal loads may be used. Rotational deviation detection processing is possible.

[Brief explanation of the drawing]

Fig. 1 is a circuit block diagram of a rotational deviation detection device according to the present invention, Fig. 2 is a time chart showing signal waveforms of the circuit of Fig. 1, and Fig. 3 (11(2) shows a standard pattern and a human cover turn). 4 to 7 are diagrams showing the principle of detecting rotational deviation of an input pattern with respect to a standard pattern. 1...Television camera 4...Binarization circuit 5
...Black pixel detection circuit 6...Window memory 8...Adder 9...
...Buffer register 10...Horizontal counter 16...CPU1
7...RAM

Claims (5)

[Claims] (1) A pattern generation means for binarizing an object image to generate a standard pattern and an input pattern, a window setting means for setting a ring-shaped window for the standard pattern and the input pattern, and each pattern having a rotational shift of a different angle. Standard information storage means for generating standard information regarding pattern constituent pixels included in the window for each standard pattern and storing the standard information for each angle; A rotational deviation detection device comprising: a determination unit that generates the input information and compares the input information with the standard information to determine the amount of rotational deviation of the input pattern. (2) The rotational deviation detection device according to claim 1, wherein the pattern generation means includes a television camera that images an object, and a binarization circuit that binarizes the object image. (3) The rotational deviation detection device according to claim 1, wherein the standard information and the input information are the sum of horizontal loads of patterns obtained for each horizontal scanning line. (4) The rotational deviation detection device according to claim 1, wherein the standard information and the input information are the number of pattern constituent pixels determined for each horizontal scanning line. (5) The rotational deviation detection device according to claim 1, wherein the standard information and the input information are horizontal weighted averages of patterns.