JP2004038669A - Image processing method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To enhance speed and reduce a used capacity of a memory by improving a label processing method. <P>SOLUTION: The image processing method has a step of attaching a temporary label to a continuous picture element group of the same type in two-dimensional image data, and a connecting process step of connecting adjacent continuous picture element groups as the same label when continuous picture element groups with different temporary labels are adjacent. In the step of attaching the temporary labels, the temporary labels are sequentially attached to continuous picture element groups of the same type along a picture element arranged direction and pieces of data of a temporary label or no label corresponding to each picture element are stored in a labeling memory, but when attaching a temporary label to a predetermined picture element, the temporary label is determined by using at least either one of temporary label data of the image data one picture element prior to the predetermined picture element or temporary label data of the labeling memory of an adjacent picture element one line prior to the predetermined picture element. By configuring the labeling memory in a cache memory 11, processing speed of a processor 10 is enhanced. <P>COPYRIGHT: (C)2004,JPO

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image processing method for image data arranged two-dimensionally, and more particularly to a method for processing image data characterized by a process of connecting a group of consecutive pixels of the same kind. The present invention is used, for example, when automatically mounting an electronic component on a printed circuit board or the like, imaging the electronic component, correcting the tilt or position of the electronic component by image processing, and accurately positioning the electronic component. .
[0002]
[Prior art]
2. Description of the Related Art Conventionally, the entire imaged area including the background of an object is temporarily stored in an image memory as two-dimensional image data as binary image data. In order to perform such processing as image recognition and the like, generally, the upper left pixel of the entire image is sequentially scanned from the diagonally opposite lower right pixel, and background pixel data usually represented by "0" is skipped. , "1" is searched for pixel data of the object. Then, a temporary label is given (labeled) to the pixel. If the pixel data represented by “1” is continuous, the same temporary label is adopted for the pixel data. Furthermore, when the temporary labels already assigned are combined, the temporary labels are arranged, that is, relabeling is performed after all the scans are completed. After relabeling, scanning is performed again, and cumulative addition is performed for each label to determine the area and the center of gravity.
[0003]
For example, Japanese Patent Application Laid-Open No. 9-91426 describes a method of obtaining an area and a start address coordinate without performing relabeling. In the case of the start address coordinate, the processing time can be shortened as compared with the center of gravity. The error increases in accuracy, and the search area in the pattern matching described later must be expanded, resulting in a reduction in the overall processing time. Also, depending on the figure, it is expected that the connection process between the labels will be complicated when the number of merged pairs increases.
[0004]
In pattern matching for high-precision positioning, the shape of the part to be recognized is stored in advance as a standard pattern, a correlation operation is performed between the standard pattern and the recognized image, and the position of the object with the position where the correlation is maximum is determined. The position is being detected. In order to obtain the correlation while gradually shifting the search area set with some allowance in consideration of the amount of movement of the pattern to be searched, the number of searches increases and the processing time increases as the search area expands.
[0005]
[Problems to be solved by the invention]
However, in the label processing by the above-described image processing method, the data subjected to the label processing must be stored, and the data amount is a bit corresponding to the number of labels to be prepared (8 bits when the upper limit of the number of labels is 255). Is required, a larger memory is required as compared with the binarized data, and the number of bytes to be read out increases, so that the processing time also increases.
[0006]
Further, in pattern matching, it takes a long time to calculate the correlation while shifting the search area little by little. If the processing is speeded up using a reduced image or the like, an erroneous detection or a decrease in accuracy occurs.
[0007]
In view of the above, the present invention provides a method of processing image data that improves the label processing method to increase the speed and reduces the memory usage, and narrows the range of a search area in pattern matching. It is an object of the present invention to provide an image processing method which is fast and free from erroneous detection.
[0008]
Other objects and novel features of the present invention will be clarified in embodiments described later.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, the invention according to claim 1 of the present application includes a step of assigning a temporary label to a group of continuous pixels of the same type in image data arranged two-dimensionally, A link processing step of linking as the same label when the groups are adjacent to each other,
In the step of attaching the temporary label, temporary labels are sequentially attached to the same kind of continuous pixel group of the image data along the pixel array direction, and the corresponding temporary label or data without a label is stored in the labeling link memory. When assigning a temporary label to a predetermined pixel, the temporary label of the pixel is determined by using at least one of the temporary label data of one pixel before the image data and the temporary label data of the labeling link memory of an adjacent pixel one line before. Is to decide.
[0010]
In the image processing method according to the second aspect of the present invention, in the first aspect, the provisional label of a predetermined pixel is determined by searching for a temporary label of an adjacent pixel in a predetermined search order and selecting a temporary label detected first. It is characterized by performing by doing.
[0011]
An image processing method according to a third aspect of the present invention uses the labeling table according to the second aspect, which has a function of registering the number of pixels for each temporary label and information of one adjacent label for one temporary label. In the detection of the adjacent label, the first adjacent label detected in a different search order from that at the time of provisional labeling is registered in the labeling table, and after the provisional labeling and the detection of the adjacent label are completed, the temporary labels adjacent to each other are detected by the labeling table. The connection processing step is performed by adding the number of pixels.
[0012]
An image processing method according to a fourth aspect of the present invention is the image processing method according to the third aspect, wherein the cumulative coordinate value in the line / pixel direction is registered in the labeling table in addition to the number of pixels for each temporary label, The area and the center of gravity of the connected labels are obtained by adding the cumulative coordinate values of the temporary labels.
[0013]
An image processing method according to a fifth aspect of the present invention is the image processing method according to the fourth aspect, wherein an area and a center of gravity of each closed figure of the master image serving as a reference are obtained and each closed figure of the image of the DUT to be positioned is positioned. And calculating the angle and the amount of displacement from the comparison of the area and the center of gravity of each closed figure between the obtained master image and the image of the device under test.
[0014]
An image processing method according to a sixth aspect of the present invention is the image processing method according to the fifth aspect, wherein the object to be measured is positioned by using pattern matching together in a search area narrowed based on the angle and the displacement amount. And
[0015]
An image processing method according to a seventh aspect of the present invention is the image processing method according to the first, second, third, fourth, fifth or sixth aspect, wherein the labeling link memory is arranged in a cache memory.
[0016]
BEST MODE FOR CARRYING OUT THE INVENTION
Hereinafter, embodiments of an image processing method according to the present invention will be described with reference to the drawings.
[0017]
FIG. 1 shows the configuration of an apparatus used in the image processing method according to the present invention, and FIG. 2 shows a two-dimensional array of binary images (10 pixels in the horizontal X coordinate direction and 10 pixels in the vertical Y coordinate direction in the illustrated example). FIG. 9 is an explanatory diagram showing that (10 pixels are arranged).
[0018]
In FIG. 1, an analog image signal 2 captured by an ITV camera 1 such as a CCD camera is digitized by an A / D converter 3 in an image processing device 8, output as multi-valued data 5, and output to a FIFO (fast (In-first-out) memory 7. The multi-valued data 5 is binarized by the binarization circuit 4, converted into gray-level two-dimensional image binarized data 6, and similarly stored in the FIFO memory 7. The processor 10 reads out the image data signal 9 from the FIFO memory 7 at the same time as the reading, stores the multi-valued data in the working memory 12, and reads the binary data and performs a labeling process described below (labels the same kind of continuous pixel group as a continuous pixel group). Processing). At the time of labeling, a labeling link memory in which labeling data is stored is created (placed) in the cache memory 11, and the result of labeling is stored in the working memory 12. Note that the cache memory 11 may be included in the processor 10 or may be arranged outside the processor. The cache memory 11 temporarily stores cache data from the working memory 12 and the like.
[0019]
The processor 10 reads the binary image data two-dimensionally arranged in the X and Y directions illustrated in FIG. 2 of the FIFO memory 7, and displays “1” pixels indicated in white along the pixel array lines in the X direction. A temporary label is assigned to a continuous pixel group of data.
[0020]
Hereinafter, a temporary labeling step in consideration of such connectivity will be described. FIG. 3 is a schematic diagram of a method of processing image data, and FIG. 4 is an explanatory diagram showing a two-dimensional shape array obtained by temporarily labeling the example of FIG. The provisional labeling is sequentially shifted in the X coordinate direction (pixel arrangement direction) from the first pixel of the first line, and is sequentially shifted to the next line when the line ends.
[0021]
In the example illustrated in FIG. 3, a labeling process of a pixel corresponding to the fourth pixel on the fifth line is illustrated. A method of attaching a temporary label will be described with reference to this and a flowchart of the temporary labeling shown in FIG.
[0022]
Assuming that the horizontal coordinate of the processing address is m and the vertical coordinate is n, the data of the (m, n) coordinate {in the example, (4, 5)} of the binary image data is read. Now, since the area and the center of gravity of the white part in FIG. 2 are to be obtained, if the read data is “0” (gray part), the label data one pixel before the same line is read from the (m− 1) After writing to the address, the label data one pixel before the line data being processed in FIG. 3 is rewritten to 0 (the processing of (5) in FIG. 3 is performed). Then, at the time of data processing of (m + 1, n), it is used as label data of one pixel before. Further, thereafter, the data is written to the labeling link memory and is used as one-line preceding label data in the processing of (m-1, n + 1), (m, n + 1), (m + 1, n + 1).
[0023]
In the case of "1" (white part)
(1) Label data one pixel before
(2) Label data at address (m-1) of the labeling link memory (where the label data of the previous line is stored)
(3) Label data at address m in the labeling link memory
(4) Label data at address (m + 1) in the labeling link memory
If the label is other than "0" and the label is other than "0", the processing is interrupted and the read label value is processed as the label value of the coordinate (m, n) being processed, and the labeling table is processed. If the above four data are all “0”, the coordinates (m, n) of the new label (the label starts from 1 and is incremented by one each time a new pixel group is detected) The processing of the labeling table is performed as the label value of.
[0024]
The above is the provisional labeling. However, since the outside of the array must be referred to at the head line and at the head and tail of the line, all of the parts are "0" in this embodiment. Data). When the temporary labeling is completed, a two-dimensional shape array as shown in FIG. 4 is created. Actually, since the labeling data overwrites the labeling link memory, the labeling data is not developed as shown in FIG. 4, but is shown to show the result of the temporary label.
[0025]
It should be noted that the data in the cache memory is updated to another data if the access is not performed for a while, and it is necessary to read the access data again from the working memory 12 when the access is made again, but it is like the label data in the labeling link memory. The labeling link memory is resident in the cache memory so that the data that is repeatedly read at a frequency is not updated. In addition, the data one line before, the line data being processed, and the data after labeling shown in the lower part of the labeling link memory in FIG. 3 are also stored in the cache memory or the primary storage means in the processor 10.
[0026]
In the following, a detailed flowchart of the labeling table process in FIG. 5 is shown in FIG. A method of rewriting a label and processing a labeling table will be described with reference to FIG. When the temporary label value of the coordinates being processed in the labeling process (temporary labeling process) in FIG. 5 is determined, as shown in FIG.
(1) Write the label data one pixel before to the address (m-1) of the labeling link memory;
(2) rewrite the label data one pixel before to a new label “r”;
Memory operation is performed in the following procedure. After the memory operation is completed, the labeling table is operated next.
[0027]
The labeling table has an array shape as shown in FIG. 8, and the number of labels is taken in a row (numerical values in the vertical direction, 1, 2, 3, 4, 5 in the figure), and a column ( In the (horizontal direction), four items of one adjacent label, two pieces (number of pixels), three cumulative horizontal coordinate values, and four cumulative vertical coordinate values are prepared. For example, r in the matrix (r, 2) in FIG. 6 indicates the label value, that is, the r-th row in FIG. 8, and 2 indicates the number item in the second column. All the labeling tables in FIG. 8 are cleared before the processing starts. Here, the cumulative horizontal coordinate value is the sum of the horizontal coordinate values of each pixel of the continuous pixel group to which the label is attached, and the cumulative vertical coordinate value is the sum of the vertical coordinate values of each pixel.
[0028]
In the processing of the labeling table, the term of the number of rows (the number of pixels) corresponding to the number of labels is incremented by 1 (one is added to the matrix (r, 2) of the labeling table as shown in FIG. 6), and the accumulated horizontal coordinates The horizontal coordinate value and the vertical coordinate value of the coordinate being processed are added to the term of the accumulated vertical coordinate (m is added to the matrix (r, 3) of the labeling table as shown in FIG. , 4) with n). In the adjacent label section,
(1) Data at address (m-1) of the labeling link memory
(2) Data at address m in the labeling link memory
(3) Data at address (m + 1) in the labeling link memory
If there is a label other than "0" or "r" and the label is not "0" or "r", the processing is interrupted and the read label value is written. If the read label is "0" or "r", the process ends without rewriting the term of the adjacent label. What is important here is the order of reading from the labeling link memory. In the present embodiment, the connection in the horizontal direction is mainly performed at the time of provisional labeling (the connection in the X coordinate direction is given priority), and the connection in the vertical direction is performed at the time of adjacent labeling. We are connecting. That is, prioritizing the connection in a different direction between the time of provisional labeling and the time of adjacent labeling facilitates the subsequent connection operation.
[0029]
FIG. 8A shows the result of the labeling table obtained by performing the temporary labeling of the two-dimensional image data of FIG.
[0030]
In the following, regarding the connection processing step of connecting adjacent temporary labels using the labeling table obtained as a result of the temporary labeling step, an example of the flowchart of the connection processing of adjacent labels in FIG. 7 and an example of the labeling table in FIG. This will be described with reference to FIG.
[0031]
FIG. 8A shows the state of the labeling table after the provisional labeling step is completed. As shown in the flowchart of FIG. 7, the connection processing step starts processing from the first row of the labeling table, and ends when the item of the area (the number of pixels) becomes “0” (state where there is no label). The processing content is first read from the adjacent label item of the row {if the row being processed is “i”, read from the matrix (i, 1) of the labeling table}, and if the numerical value is “0”, the next step is Move to the processing of the row. If the read numerical value is other than “0”, the numerical value of the item of the numerical value p (the line “i” being processed, the label p adjacent to the line), the cumulative horizontal coordinate and the cumulative vertical coordinate are processed. The value is added to the value of the i-th row, which is a row, and rewritten to the value (the processing of a (p, j) = a (i, j) + a (p, j) in FIG. 7 is performed, and the addition result is p rows Where a (p, j) represents the value of the row p and the column j of FIG. 8). In the i-th processing line, the items of the area, the accumulated horizontal coordinate, and the accumulated vertical coordinate are rewritten to “0” (the processing of a (i, j) = 0 in FIG. 7 is performed).
[0032]
FIG. 8B shows the result of performing the above connection processing for FIG. 8A. The number of rows “0” is deleted from the obtained connection processing result, and the number is defined as an area for each row,
Barycenter horizontal coordinate value = cumulative horizontal coordinate value / area
Barycenter vertical coordinate value = cumulative vertical coordinate value / area
The barycentric coordinates of each closed figure are obtained according to the following equation. FIG. 9 shows the result of calculating the center of gravity for FIG. 8B.
[0033]
When the provisional labeling step and the connection processing step are completed, a final labeled two-dimensional shape array as shown in FIG. 10 is created. Actually, since the linking process is performed in the labeling table, the memory is not expanded as shown in FIG. 10, but is shown to show the result of the linking process.
[0034]
Next, a pattern matching method using the area and the center of gravity of a closed figure with a label will be described. When the binarized image of FIG. 2 is registered as a standard pattern of a master image serving as a reference, and pattern matching is performed with the image of the DUT to be positioned, the temporary labeling step and the connection processing step are used when the standard pattern is registered. Find the area and center of gravity, and register them together with their numerical values. Furthermore, in the case of pattern matching with an angle, a rotation standard pattern at each angle is registered in advance by using an affine transformation (movement and rotation processing by coordinate transformation of image information) for each closed figure or the entire image. Processing time can be reduced.
[0035]
When pattern matching is performed on an example of a two-dimensional array of binary image data (an example of an image of an object to be positioned) obtained by moving (including rotating) FIG. 2 as shown in FIG. By the attaching step, a result as shown in FIG. 12 is obtained on the labeling table. When the connection processing and the calculation of the area and the center of gravity are further performed on the result of FIG. 12, a result as shown in FIG. 13 is obtained. FIG. 14 shows a two-dimensional shape array obtained by performing temporary labeling on FIG. In practice, the labeling data overwrites the labeling link memory, so the memory is not developed as shown in FIG. 14, but is shown to show the result of the temporary label.
[0036]
FIG. 15 shows the area of the standard pattern and the area and the center of gravity data of the recognition image of the device under test sorted by area. The length and angle between the centers of gravity of the closed figures 1 and 2 (correctly, it is necessary to consider the aspect ratio, but here the aspect ratio is set to 1: 1 for simplicity) is calculated. The calculation of the length is used when it is impossible to determine only the area due to the positional relationship of the closed figures (in this example, there is only two closed figures, so it does not function, but if there are many closed figures, Based on the information on the area and the length between the centers of gravity, it is conceivable to make correspondence between the standard pattern and the closed figure of each recognized image.)
[0037]
In the present embodiment, the relative angle between the closed figures 1 and 2 is obtained for the standard pattern and the recognition image, and the angle difference is the rotation angle from the standard pattern to the recognition image. This is illustrated in FIG. In the closed figure 1, the center of gravity 13 (3, 5.47) of the standard pattern has moved to the center of gravity 15 (8.5, 5) of the recognized image. It has moved from (7.55, 5.45) to the center of gravity 16 (3.75, 5.65) of the recognized image. The rotation angle is 359.7 ° between the closed patterns 1-2 in the standard pattern, whereas it is 187.7 ° in the case of the recognition image, and the rotation angle between the standard pattern and the recognition image is The difference is 188.0 °. Here, when the standard pattern is rotated and superimposed on the recognition image, when the counterclockwise direction is plus and the clockwise direction is a minus angle, the standard pattern angle is θx, and the recognition image angle is θy. The angle θz for rotating the pattern is
(1) When θx> θy
θz = 360− (θx−θy)
(2) When θx <θy
θz = θy−θx
It becomes.
[0038]
A certain degree of accuracy can be expected even with the displacement and rotation angle obtained from the comparison of the area and the center of gravity of the standard pattern and the recognition image. And the positioning accuracy can be improved up to sub-pixels (dimensions smaller than one pixel).
[0039]
In addition, using the displacement and the rotation angle obtained from the comparison between the area and the center of gravity of the standard pattern and the recognition image, the displacement of the center of gravity of the closed figure 1 ± 1 to 2 pixels was obtained using the displacement and the rotation angle. Affine transform the standard pattern at the angle closest to the rotation angle, perform pattern matching with the recognized image using the image created in advance as the master image, obtain the respective correlation values, obtain the displacement amount by normalized correlation, and then In the same manner, the amount of displacement is obtained for the closed figure 2 and the rotation angle is calculated again from the obtained amount of displacement between the closed figures 1 and 2, thereby obtaining a more accurate rotation angle.
[0040]
In the case of normal rotation pattern matching, if the search area is an area of ± 5 pixels (the search area is determined by the maximum displacement of the standard pattern), the standard pattern needs to be moved 11 × 11 = 121 times for calculation. There is. If n patterns of rotated images are prepared in consideration of the rotation (the number of times is determined by the error due to the angle range to be obtained and the angle deviation), 121 × n more calculations are required. On the other hand, when the processing method according to the present embodiment is used, the search area can be limited to ± 1 to 2 pixels, so that the number of movements can be suppressed to 9 to 25 times, and the rotation can be performed in advance with the approximate rotation angle. Is known only for one rotated image, so that the number of movements does not increase.
[0041]
Therefore, in a case where the search area is wider or a case where it is required to obtain the rotation accuracy with high accuracy, the processing method according to the present embodiment further increases the effectiveness.
[0042]
According to this embodiment, the following effects can be obtained.
[0043]
(1) In the temporary labeling step, of the image data arranged two-dimensionally as shown in FIG. 2, a temporary label is sequentially assigned to the same kind of continuous pixel group along the pixel arrangement direction, and the pixels are stored in the labeling link memory. Each time a corresponding temporary label or data without a temporary label is stored, when a predetermined pixel is assigned a temporary label, as shown in FIG. 3, the labeling link of an adjacent pixel one pixel before and one line before the image data as shown in FIG. The temporary label of the pixel is determined using the label ring data (temporary label data) of the memory.
[0044]
Therefore, the memory capacity is reduced by processing the temporary label information for each pixel while overwriting the labeling link memory for one line, and the labeling link memory is stored in the cache memory 11 in the processor 10 of FIG. And the number of accesses to the labeling data (temporary label data) stored in the labeling link memory can be reduced, and the speed can also be increased in this regard.
[0045]
(2) Using a labeling table having a function of registering the number of pixels for each of the temporary labels and the information (connection information) of one adjacent label for one temporary label, the pixel arrangement direction of the image data (X coordinate ), A temporary label is sequentially attached to the same kind of continuous pixel group along the direction (the search order is prioritized in the X coordinate direction). Is registered in the labeling table, and after the provisional labeling and the detection of the adjacent label are completed, the connection processing step is performed by adding the number of pixels of the temporary labels adjacent to each other by the labeling table. By adding the number of pixels between adjacent temporary labels in this connection processing, the area that is a continuous band of the same kind of pixels is obtained.
[0046]
As described above, since the information of the adjacent label can be registered on the labeling table, the relabeling becomes unnecessary, and the speed can be improved and the memory capacity can be reduced.
[0047]
(3) The cumulative coordinate values in the line / pixel direction are registered in the labeling table in addition to the number of pixels for each temporary label, and the cumulative coordinate values of adjacent temporary labels are added in the connection processing step, thereby re-labeling. The area and the center of gravity of each connected label can be obtained without processing the data. As described above, by performing the concatenation and the center-of-gravity processing in parallel, the number of memory accesses can be reduced and the speed can be further increased.
[0048]
(4) In the above-described provisional labeling step and the link processing step, the area and the center of gravity of each closed figure of the master image serving as a reference are obtained, and the area and the center of gravity of each closed figure of the image of the DUT to be positioned are similarly calculated. The angle and the displacement can be calculated at high speed by comparing the area and the center of gravity of each closed figure between the obtained master image and the image of the device under test. By calculating the positional relationship and the rotation angle in this way, it is possible to narrow the search area when performing pattern matching.
[0049]
(5) In order to position the device under test using pattern matching together in a narrowed search area based on the angle and displacement of each closed figure in the master image and the image of the device under test, the number of searches is The reduction can increase the image processing speed as a whole. In addition, since the search area is narrowed, a high processing speed can be maintained without using a search with a reduced image, so that harmful detection due to the reduced image or erroneous detection even when a similar image exists in the vicinity may occur. Absent.
[0050]
Although the embodiments of the present invention have been described above, it will be obvious to those skilled in the art that the present invention is not limited to the embodiments and various modifications and changes can be made within the scope of the claims.
[0051]
【The invention's effect】
As described above, according to the image processing method of the present invention, when a temporary label is assigned to a continuous pixel group of the same type in image data arranged two-dimensionally, the same type of continuous pixel of the image data is used. Temporary labels are sequentially assigned to the groups along the pixel array direction, and the corresponding temporary labels or data without temporary labels are stored in the labeling link memory for each pixel. The temporary label of the pixel is determined by using at least one of the temporary label data of one pixel before and the temporary label data of the labeling link memory of the adjacent pixel one line before. By limiting the temporary label data after labeling to a one-line labeling link memory and arranging it in a cache memory, a small memory capacity is required and high-speed labeling processing can be realized.
[0052]
In addition, in the linking process, when an item of the adjacent label, the number of pixels, and the cumulative coordinate value is added to the labeling table and the process of obtaining the area and the center of gravity is performed simultaneously with the process of connecting the adjacent labels, without accessing the labeling data again. High-speed area and center-of-gravity processing is possible by processing within the labeling table.
[0053]
Further, using the processing method for obtaining the area and the center of gravity at a high speed, the displacement amount and the rotation angle between the reference master image and the image of the device under test are calculated, and the search area for pattern matching is calculated based on the data. By narrowing and processing with a fixed rotation angle, a method of processing image data for performing high-speed and high-accuracy positioning can be realized.
[Brief description of the drawings]
FIG. 1 is a block diagram illustrating an apparatus configuration used in an embodiment of an image processing method according to the present invention.
FIG. 2 is an explanatory diagram showing an example of a two-dimensional shape array of binary image data.
FIG. 3 is a schematic diagram of a process of temporarily labeling image data according to the embodiment;
FIG. 4 is an explanatory diagram illustrating a two-dimensional shape array in which an example of the binary image data in FIG. 2 is provisionally labeled.
FIG. 5 is a flowchart of a temporary labeling step (temporary labeling process) in the embodiment.
FIG. 6 is a detailed flowchart of a labeling table process in the flowchart of FIG. 5;
FIG. 7 is a flowchart of a linking step of adjacent labels according to the embodiment;
FIG. 8 is an explanatory diagram illustrating an example of a labeling table according to the embodiment;
FIG. 9 is an explanatory diagram showing the calculation results of the area and the center of gravity obtained by the connection processing using the labeling table of FIG. 8;
FIG. 10 is an explanatory diagram showing a two-dimensional shape array after final labeling processing (after connection processing) of the example of the binary image data of FIG. 2;
11 is an explanatory diagram showing an example of a two-dimensional shape array of binary image data obtained by moving the binary image data of FIG. 2;
FIG. 12 is an explanatory diagram illustrating an example of a labeling table in the case of the binary image data in FIG. 11;
FIG. 13 is an explanatory diagram showing the results of the area and the center of gravity obtained from the labeling table of FIG. 12;
FIG. 14 is an explanatory diagram showing a two-dimensional shape array obtained by temporarily labeling the example of the binary image data shown in FIG. 11;
FIG. 15 is an explanatory diagram showing an example of a movement amount and a rotation angle calculation result in a comparison between the binary image data of FIGS. 2 and 11;
[Explanation of symbols]
1 camera
2 Image signal
3 A / D converter
4 Binarization circuit
5 Multi-valued data
6 Binary data
7 FIFO memory
8 Image processing device
9 Image data signal
10 processor
11 Cache memory
12 Working memory
13 Coordinates of barycenter of closed figure 1 of standard pattern
14 Coordinates of barycenter of closed figure 2 of standard pattern
15 Coordinates of center of gravity of closed figure 1 of recognition figure
16 Coordinates of barycenter of closed figure 2 of recognition figure

Claims (7)

A step of assigning a temporary label to a group of continuous pixels of the same type in the image data arranged in a two-dimensional manner, and a linking process of linking as a same label when a group of continuous pixels with different temporary labels is adjacent to each other And an image processing method comprising:
In the step of attaching the temporary label, temporary labels are sequentially attached to the same kind of continuous pixel group of the image data along the pixel array direction, and the corresponding temporary label or data without a label is stored in the labeling link memory. When assigning a temporary label to a predetermined pixel, the temporary label of the pixel is determined by using at least one of the temporary label data of one pixel before the image data and the temporary label data of the labeling link memory of an adjacent pixel one line before. An image processing method characterized in that: 2. The image processing method according to claim 1, wherein the determination of the temporary label of the predetermined pixel is performed by searching for a temporary label of an adjacent pixel in a predetermined search order and selecting a temporary label detected first. A labeling table having a function of registering the number of pixels for each temporary label and the information of one adjacent label for one temporary label is used. 3. The image according to claim 2, wherein the detected adjacent label is registered in the labeling table, and after the provisional labeling and the detection of the adjacent label are completed, the connection processing step is performed by adding the number of pixels of the adjacent temporary labels by the labeling table. Processing method. By registering the cumulative coordinate values in the line / pixel direction in addition to the number of pixels for each temporary label in the labeling table, and adding the cumulative coordinate values of adjacent temporary labels in the connection processing step, the label after connection is obtained. 4. The image processing method according to claim 3, wherein an area and a center of gravity of the image are obtained. 5. The image processing method according to claim 4, wherein an area and a center of gravity of each closed figure of a master image serving as a reference are obtained, and an area and a center of gravity of each closed figure of an image of an object to be positioned are obtained. An image processing method comprising calculating an angle and an amount of displacement from a comparison between the area and the center of gravity of each closed figure between the image and the image of the device under test. 6. The image processing method according to claim 5, wherein the object to be measured is positioned by using pattern matching together in a search area narrowed based on the angle and the displacement amount. 7. The image processing method according to claim 1, wherein said labeling link memory is arranged in a cache memory.

Images