JPS58158761A - Pattern position detecting method - Google Patents

Abstract

PURPOSE:To specify a position of the best matching picture element to one position, to make a picture element memory unnecessary, and to make a position detecting device small-sized, by taking the sum of plural best matching picture elements of a video pattern obtained by binary-coding processing. CONSTITUTION:An analog signal from a TV camera 1 is binary-coded by a binary-coding circuit 3 by a control signal from a control signal generating circuit 2, and from a cutting circuit 4, NXM-number of picture element signals are applied to a matching processing circuit 7. Also, NXM-number of reference signals corresponding to template size from a template pattern generating circuit 6 are applied to the circuit 7. By this circuit 7, AND operation is executed, and its result is counted by a counting circuit 8 and is provided to a cutting circuit 9. From this circuit 9, multi-value video signals of (m)X(n) which become one picture element corresponding to a matching binary-coded video signal are outputted. Subsequently, by an adding circuit 10, a multi-value video signal of one picture element corresponding to the video signals of (m)X(n) is outputted, is applied to a comparing circuit 11, and by the circuit 11, a position of a prescribed pattern is detected.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention detects the position by pattern matching, and detects the position when the patterns of the binary video signals to be matched are different, the template size is small, etc. Therefore, even when it is difficult to clearly detect a position from a matching result, the present invention provides a simple pattern position detection method that can perform position detection with high accuracy and has a high processing speed.

Conventionally, in position detection using pattern matching,
TV camera (analog signal from TV left camera
The signal is processed after it is converted into a binary video signal by D conversion. The way the brightness level changes at the boundary between the object and the background varies depending on the object, so the pattern represented by the binarized video signal may be inconsistent, there may be noise, the template size is small, etc. In addition, a plurality of positions with the same maximum matching degree may be obtained as a match result, or the maximum matching degree may not be obtained at the position that should originally be detected, making it difficult to detect the position clearly.

For this reason, the following can be considered.

(1) Increase the template size.

(2) Process the video signals so that they represent as much of the same pattern as possible.

However, in the case of (1) above, the scale of the apparatus becomes large and the processing time becomes long. When matching patterns of line intersections, if the line width of the template and the line width to be matched are different, the position at which the maximum matching degree is obtained will be multiple regardless of the template size, and clear position detection cannot be performed.

In (2) above, the video signal from the TV left camera is A/D converted into a multi-level video signal, imported into the image memory, all data is examined, and then the pattern to be matched is determined. All video data is transferred to the image memory. Although the device can perform various processing and accurately detect the position, the device is large and complicated, and since a large amount of signal processing is performed by software, the processing speed is slow, and the price is high, making it difficult to generalize.

The present invention eliminates such drawbacks, and will briefly describe the present invention.The present invention will be briefly described below. Matching processing is performed with the binarized reference signal, and the value of the matching processing result is used as the MXN.
1 corresponding to the binary video signals to be matched
Let m×
Calculate the sum of the values of n pixels, and use the calculated value as mXn
A multi-value video signal of one pixel corresponding to a multi-value video signal of Position detection is performed using not only the matching processing results but also the matching processing results of the pixel of interest and m×n pixels in its vicinity. In the multilevel video signal obtained from the matching processing result, pixels that show good matching always appear, and even if the pixels with the best matching are clustered on multiple sides, the system is , it is easy to specify the position of the pixel with the best matching to one position, and therefore it is possible to detect the position with high accuracy. Additionally, no image memory is required and processing can be done with simple hardware.

The present invention will be described in detail below using the embodiment shown in FIG.

In the figure, analog video signals output from a TV right camera made of a solid-state imaging device such as a COD or a tube imaging device such as a vidicon are sequentially processed based on signal commands from a control signal generation circuit 2. First, the analog video signal is A/D converted by the binarization circuit 3 to become a time-series binarized video signal for each scanning line. At this time, if the analog video signal from the TV right camera is passed through a spatial filter such as a differential filter, preprocessed, and then converted to a binary video signal, the difference in brightness and darkness that makes up the object, as well as the outline, will be small. Feature points of the image object, such as boundary lines, can be obtained. Then, the first cutout circuit 4 cuts out N pixel signals to be matched that are sequentially lined up from each of the V sequentially lined up scanning lines of the binarized video signal. 5 is an XY address counter.

On the other hand, the template pattern generation circuit 6 generates MXN binary reference signals corresponding to the template size representing the putter to be matched. 'r1+~TMN) in Figure 3). Then, in the cutout circuit 4, the cutout pixel signal is ANDed with the setting reference signal in the matching processing circuit 7, and the result is counted in the counting circuit 8, and the binarized image to be matched is One pixel (for example, 2i of bll in FIG. 2 whose position corresponds to all) corresponding to the signal (for example, a1f to aMN in FIG. 2)
An i-bit multivalued video signal of MxN is output as a time-series video signal for each scanning line. Note that if the video signal arrives at a position where even one of the MXN pixel signals to be cut out is missing, the data of the missing pixel signal is processed as 0.

Next, a second extraction circuit 9 extracts n consecutive pixel signals of each of m consecutive scanning lines from the multi-level video signal obtained by counting the AND operation results, and an adder circuit 10 extracts them. The sum of the values of the pixel signals is calculated, and as in the case of the matching processing circuit 7, a multi-value video signal of one pixel corresponding to the m x n multi-value video signals is obtained. If a video signal arrives at a position where even one of the m×n pixel signals to be extracted is missing, the data of the missing pixel signal is processed as 0. Next, in the comparison circuit 11, the initial setting of the reference is set to 0, and the pixel values of the multilevel video signal, which is the input signal, are compared sequentially.As a result, if the value of the input signal is large, the pixel value, that is, - Output the mulberry matching value and its address, store a large matching value in the maximum matching value memory 12, and newly set the large matching value in the reference of the comparison circuit 11, and output the maximum matching value address memory 13. When the signal processing is finished in this way, the maximum matching value is stored in the maximum matching value memory 12, and the address is stored in the maximum matching value address memory 13. This is the address with the maximum matching value.

The X-Y coordinates of the address with the maximum matching value are made clear by the command signal from the X-Y address counter 6, and the matching position on the image can be detected.

A specific pixel (T I 5
To detect a device, a pixel position TN on the template corresponding to the pixel position indicating the MxN matching results is used.
It can be determined from the positional relationship with Σ.

When the adder circuit 1o has a large number of m x n signals and cannot perform addition processing using digital signals in real time, the cutout circuit 9
The maximum matching value and its address can also be obtained by D/A converting the signals cut out from the above, adding them in an analog manner using an adder using a resistor, and performing the subsequent signal processing in an analog manner.

To describe the extraction circuit 4 and the matching processing circuit 7 in more detail, in FIG. 3, the extraction circuit 4 has M shift registers (S-1) corresponding to the extraction scanning lines.
-S-M), the control signal generation circuit 20 receives the binary signal from the shift register C3-M, l in a push-out manner based on the command signal, and the matching processing circuit that warps the MXN pixel signals to be matched for the first time. 7 performs an AND operation, and a counting circuit 8 counts the matching results. The frequency of the reference clock used in the control signal generation circuit 2 and j are 4.
If you use something around 0MH2, the image capture speed from the TV left camera will be faster. Real-time matching processing on the order of seconds (2 frames) can be fully performed.

The matching processing circuit 7 performs EX-O instead of AND operation.
Perform R operation (Exclusive-OR operation) processing,
If the result of matching image data 0 and 1 is taken, all M×N image data of the template will be used, which has the effect of increasing resistance to noise.

Next, an example of actual data is shown.

In the 8x8 binary image data shown in Figure 4 (this data is a template with a part), the center position of the intersection line corresponding to the position of pixel data 1 (Tss) indicated by the circle mark is set as a binary image. Detected from the signal.The position of the mark on the left end corresponds to template T11.

Further, in FIG. 6, (a) is a part of the binary image data to be matched, and the line width is larger than the pattern represented by the template. The position that we want to detect in the image data of (a) is the position of image data 1 marked with a circle in the center of the intersection line, but the matching result is the upper left end of the MXN size template (the position corresponding to T month 2). Since it is expressed as multivalued image data, we will consider the detection of this position in the future. (b) is 2 of the template in Figure 4.
This is image data of a multivalued video signal representing the result of AND processing of the valued image data 1 and the binary image data 1 of 0. There are a total of nine pixels representing the maximum matching value of 17 at the position to be detected and its vicinity. (c) is the sum of 3×3 multivalued image data resulting from the AND operation,
Further, it is image data obtained by subtracting the sum of additions from 163, and is 1 corresponding to the 3×3 multivalued video signals.
The position of the pixel was set at the center of 3×3 pixels. From this, only one position representing the maximum matching value of 0 can be found, which is the position desired to be detected.

As described above, position detection is performed by pattern matching, and the position at which the maximum matching degree is obtained is determined because the patterns represented by the binary video signals to be matched are inconsistent or the template size is small. Even if a plurality of images are obtained or erroneous matching occurs due to noise or the like, since the position is statistically detected using the degree of matching in the vicinity of the matching position of interest, the position can be detected accurately. In addition, position detection of intersection points such as boundary lines of objects that require multilevel image processing can be easily performed by performing differential processing or the like as processing before binarization.

At this time, the image pattern obtained by performing binarization processing after performing differential processing etc. tends to have broken lines or different line widths, but in the present invention, accurate position detection cannot be performed. Moreover, since no image memory is required, the device for implementing the method is simple and small, and can be realized at low cost.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of an apparatus for implementing the pattern position detection method of the present invention, FIG. 2 is a diagram showing the position of one pixel corresponding to a plurality of processed pixels, and FIG. FIG. 4 is a circuit diagram showing a specific example of the extraction circuit and matching processing circuit in the apparatus shown in FIG. FIG. 3 is a diagram showing an example of a video signal, (B) is a diagram showing an AND processing result, and (C) is a diagram showing an addition sum of 153-AND processing results. 1...TV camera, 2...Control signal generation circuit. 3...Binarization circuit, 4...Mark/cutout circuit, 6
- River: X-Y address counter, 6: Template pattern generation circuit, 7: Matching processing circuit, 8: Counting circuit, 9:... Extraction circuit, 1o...Addition circuit, 11...Comparison circuit, 12...Maximum matching value memory, 1
3. Address memory for maximum matching value. Name of agent: Patent attorney Toshio Nakao and 1 other person11
2. Country No. 381' in Figure 4 and Figure 5

Claims (2)

[Claims] (1) Analog video signal from TV camera
From the time-series binary video signal for each scanning line obtained by D conversion, N pixel signals to be matched are obtained from each of the M scanning lines arranged in i order, and the pixel signals and the above-mentioned pixel signals are obtained. MxN 2 pieces corresponding to the template size representing the pattern to be matched to the binary video signal
An AND operation or an EX-OR operation is performed with the digitized signal, and the result of the AND operation or EX-OR operation is counted by a counting circuit to match [L , and is output as a time-series video signal for each scanning line, and m consecutive multi-value video signals are outputted as a time-series video signal for each scanning line. An adder circuit calculates the sum of the values of n consecutive pixel signals of each scanning line to obtain a one-pixel multi-value video signal corresponding to the mXn multi-value video signals, and a comparator circuit calculates the sum of the values of n consecutive pixel signals of each scanning line. A pattern position detection method characterized in that the position of a predetermined pattern is detected from a signal position having an appropriate value in the multivalued video signal whose sum has been calculated. (2) After passing the analog video signal from the television camera through a spatial filter such as a differential filter, A/D conversion is performed.
The pattern position detection method according to claim (1), characterized in that the pattern position detection method is converted into a value.