JP2000155011A - Position measuring method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically determine the optimization of parameters such as the combination of preprocessing filters and the thinning ratio of matching processes without the intervention of a worker. SOLUTION: This position measuring device is provided with a means 2 acquiring a good sample image of the same kind as the image of an on-line inspection, a means 3 setting a template region used for position measurement, a means 4 setting the search area scanning a template for position measurement, and a means 5 automatically determining the strategy for stably conducting position measurement at a high speed. The position measurement strategy automatic determining means 5 optimized the parameters such as the kind and sequence of filter processing and thinning ratio in response to the sample image with the template and search area set manually. The genetic algorithm(GA) is employed for determining the optimum combination of parameters.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to position measurement using images, and more particularly, to automatically and stably measure the position of a freely set object by automatically combining strategies of preprocessing and thinning processing. The present invention relates to a position measurement method and device to be determined.

[0002]

2. Description of the Related Art One of the techniques used for a position measuring device is one using an image. In the position measurement device using this image, an object whose position is to be measured is registered in advance as a template, the template is scanned from the input image to perform template matching, and when the correlation value becomes highest, the object is measured. Measure the position. This technology is used in the manufacturing process of various industrial products such as printed boards, and the markings to be measured are called alignment marks.

In this position measurement, an alignment mark suitable for position measurement and a pattern that substitutes for the position are set in an image, and adjustment is performed so that the alignment mark set using template matching can be measured stably. Specific contents of the adjustment include a resolution of the image for high-speed processing, a pre-processing that emphasizes a feature amount in the image so that the measurement can be performed stably, and a combination of the resolution and the pre-processing. For each position measurement target, these adjustments are performed by a developer who is familiar with image processing by trial and error.

[0004] As a conventional technique of position measurement using images,
There is a method in which the position measurement processing is performed in two stages of coarse measurement and dense measurement. In other words, the position of the image is reduced and the position is roughly measured at a high speed, and the position of the second stage is measured with high accuracy using the high-resolution image. According to JP-A-8-111599, which shows an example of this method, as a method of dividing the processing into two steps, a rough position is extracted at high speed using density projection, and then the correlation value between the template and the input image is extracted. Is calculated and the position is measured with high accuracy. The flowchart of FIG. 2 shows a general procedure of position measurement by preprocessing, coarse position measurement processing, and dense position measurement processing.

[0005]

When performing position measurement using an image, it is important how fast and stable the set template can be detected in the input image. In the method in which the processing is divided into two steps, it is difficult to measure the position stably when the alignment mark is small or the pattern is complicated and the density projection cannot be stably obtained. It is also conceivable to perform position measurement by reducing the resolution of the image and reducing the amount of data.However, this method also lacks features in the image when the alignment mark is small or the pattern is complicated. Therefore, stable matching by template matching cannot be performed.

In order to perform the matching process stably, it is necessary to emphasize the features of the image so that the features are not lost even if the resolution of the image is reduced, or to adjust the degree to which the resolution is reduced so that the features are not lost. . In other words, it is necessary to change the position measurement strategy according to the situation of the image to be used,
It is necessary to adjust parameters necessary for position measurement, such as setting of a template, an area to be searched, and resolution for speeding up.

FIG. 3 shows a conventional strategy determination procedure. Conventionally, in the search strategy setting (S24), the operator has empirically set parameters such as the type of filter used for preprocessing and the reduction rate (thinning rate) of image data. Then, a plurality of samples are evaluated (S25), and if the required performance is not satisfied (NG), the search strategy is reset by feedback (S28). If the performance is not satisfied only by changing the search strategy, feedback S27, S26 and the resetting of the template and the search area are further required.

As described above, conventionally, every time a sample is evaluated, the operator resets and re-evaluates the parameters over a long period of time, which places a heavy burden on the operator. further,
The image feature enhancement process requires know-how of a skilled person, which is a difficult task for ordinary workers.

An object of the present invention is to overcome the problems of the prior art, automatically optimize parameters and the like, and determine a strategy for a high-speed and stable position measurement method without operator intervention. It is to provide a device.

[0010]

According to the position measuring method of the present invention which achieves the above object, a target image is input, a predetermined preprocessing is performed, and a search process using a template including a characteristic of a predetermined position of the target image is performed. When performing and measuring the predetermined position, in advance, for a template and a search region set based on the sample image of the target image, a combination of parameters consisting of a plurality of types and order of filter processing,
It is characterized in that selection is made so that features of the image are maintained or emphasized, and the preprocessing is performed by a combination of selected parameters when measuring the position of the target image.

The combination of parameters including the types and orders of the plurality of filter processes for the pre-processing and the thinning rates and the thinning positions in the X and Y directions of the image for the search process are determined by the characteristics of the image. It is characterized in that the search processing is selected so as to be accelerated while maintaining the same, and in the position measurement using the target image, the preprocessing and the search processing are performed by a combination of the selected parameters.

Further, the position measurement of the target image is performed in two stages. First, coarse position measurement is performed by lowering the resolution of the image to be used by applying the combination of the parameters, and the search area is narrowed according to the result. At the same time, the resolution is restored and dense position measurement is performed.

Further, the combination of the parameters is selected so that the degree of recognition by the matching process between the template and the sample image is high and the processing time is short. The recognition degree is determined depending on a difference between a correlation value at the predetermined position by the matching processing and a maximum correlation value at a position other than the predetermined position. It should be noted that the maximum correlation value at a position other than the predetermined position is obtained from a result of a matching process on all the image data when a plurality of image data are obtained from a thinning rate set for an image to be used.

The position measuring apparatus according to the present invention measures a predetermined position by inputting a target image and performing predetermined pre-processing, performing a search using a template including a feature of the predetermined position of the target image. A template area setting means for setting a template used for position measurement based on a sample image of the target image; a template search area setting means for setting a search area for scanning the template; and a template set by these means. And a parameter selection means for selecting a combination of parameters including the type and order of filter processing to be performed on the sample image as the pre-processing so that the characteristics of the image are maintained or emphasized. It is characterized by having.

Alternatively, there is provided a position measuring mechanism for inputting the target image, performing predetermined preprocessing, and performing a search process using a template including a feature of a predetermined position of the target image, and further, based on a sample image of the target image. Template area setting means for setting a template to be used for position measurement, template search area setting means for setting a search area for scanning the template, and using the template and search area set by these means as the pre-processing. The type and order of the filter processing applied to the sample image and a combination of parameters including the thinning rate and the thinning position in the X and Y directions of the sample image are selected so as to speed up the search processing while maintaining the characteristics of the image. A search strategy determination mechanism having parameter selection means for performing According to the combination of the selected parameter decision mechanism, characterized by being configured to perform search processing and pre-processing of the position measurement mechanism.

According to the present invention, when a worker sets a template region and a template search region, a pre-processing filter process enabling stable position measurement according to an image to be used is automatically selected. Can be.

Further, a combination of the resolution (thinning rate) of the image in the filter processing in the pre-processing and the search processing is automatically selected so that the characteristics required for position measurement are not lost even when the resolution is lowered. Therefore, parameter adjustment work for high-speed and stable position measurement can be significantly reduced.

That is, according to the present invention, the operator can perform high-speed and stable position measurement based on a strategy created for each target simply by setting an alignment mark or a pattern instead of the alignment mark in the apparatus.

[0019]

An embodiment of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram illustrating a processing function of a position measurement device using an image according to an embodiment. In the position measurement of the present example, a corresponding position is searched for by checking which part of the input image matches the template image registered in advance and the correlation value.
Therefore, a search strategy setting mechanism 1 for setting a template image and a search strategy, and a recording unit 6 for recording such information.
And a position measurement mechanism 7 that performs position measurement using the set strategy.

The search strategy setting mechanism 1 acquires a sample of an image used for position measurement by the sample image acquiring means 2,
The operator sets a template image using a graphic interface (GUI) for the alignment mark area of the sample image acquired by the position measurement template area setting means 3. Further, the template search area setting means 4 sets an area in the sample image for searching for a template.

Next, the position measurement strategy automatic determination means 5 evaluates the template image and the search area set by the operator, and performs preprocessing such as emphasis processing applied to the image so that position measurement can be performed quickly and stably. And speed-up method are determined automatically.

Since the processing performed by the search strategy setting mechanism 1 is performed off-line, the determined template image, search area information and search strategy information are stored in the template image area 61 and the search area information area 6 of the recording means 6, respectively.
2 and the search strategy table 63.

Position measurement mechanism 7 for online measurement
Searches the template image stored in the recording unit 6 and the input image of the search area using the determined search strategy. Therefore, the input image to be actually measured is obtained by the image obtaining means 8, and the position measurement processing means 9 searches for the position of the template in the input image according to the search strategy. The result output means 10 displays the measurement result on a display and transmits the result to a position control device or the like which needs the result.

Hereinafter, an example in which the present invention is applied to a printed circuit board inspection apparatus will be described. FIG. 4 shows a position measuring mechanism of the printed circuit board inspection device. First, a substrate 49 of a sample (non-defective product) to be subjected to position measurement is placed on the stage 43. An image of the sample is acquired by the camera 41 and transmitted to the position measurement device 42. In accordance with the inspection data, the printed board inspection apparatus controls the position of the camera 41 with respect to the target component mounted on the board, and inspects the component shape and size.
In the following, a description will be given of an example of measuring the position of an alignment mark on a printed circuit board.

The position measuring device 42 is composed of a computer device such as a personal computer (PC), and has the above-described search strategy setting mechanism 1
, A recording unit 6 and a position measuring mechanism 7 to execute a series of processing. First, an operator specifies a template 46 including an alignment mark 46 and a search area 48 for searching for an alignment mark with respect to an image 45 of a sample substrate 49 displayed on a display 44.
Based on this setting, a search strategy is automatically determined as described later.

Next, the workpieces to be inspected are successively placed on the stage 43, and the online processing of the substrate inspection is started. The position measuring mechanism 7 measures the position of the alignment mark 46 using the image of the workpiece by the camera 41 and the search strategy stored in the recording means 6. The measured position information is used for correcting the position of a mounted component on the printed board or the printed board itself. In other words, the position is corrected by operating the stage 43 using the measured position information of the alignment mark.

Next, the processing contents of the position measuring mechanism 7 will be described. As shown in FIG. 2, first, an image of a work to be used for position measurement is obtained (S11). In the position measurement of the present example, coarse position measurement is performed to speed up the processing (S14), and the position is limited to a narrow range, and dense position measurement is performed for high-accuracy position measurement without reducing image data. Perform processing (S1
5). Thereby, highly accurate position measurement can be performed at high speed.

In the coarse position measurement processing in step S14, image data used for position measurement is reduced, and the amount of data used for processing is reduced to speed up the position measurement processing. At this time, if the image data is simply reduced, features in the template and the input image will be lost, and the recognition rate will be reduced, and stable position measurement will be difficult.

Therefore, in this embodiment, a plurality of filtering processes are performed on the image to be used by preprocessing 1 to preprocessing n (S12, S13), and features in the template and the input image are emphasized or exaggerated. As a result, even if the image data is reduced, it is possible to prevent the feature in the template from being lost.
In the present embodiment, a combination of filter processing for pre-processing,
Automatically set the thinning rate for speeding up.

This embodiment automates the setting of the search strategy (S24) and the sample evaluation (S25) in FIG. 3, and uses the genetic algorithm GA (neuro / neuro) to determine the optimal combination of a plurality of parameters of the search strategy. Experiments and Applications of Genetic Methods: Interface, February 1992). Hereinafter, the automation method in the present embodiment will be described in detail.

FIG. 5 is an explanatory diagram of parameters of search strategy and strategy management. Filter processing performed as preprocessing includes a smoothing filter, an edge enhancement filter, a maximum value filter, a minimum value filter, and an intermediate value filter. Each filter has a plurality of processes depending on parameters such as a window size and a weight coefficient. FIG. 6 shows an example of the filter. There is a filter A as a smoothing filter and filters B and C as an edge emphasis filter.

For this reason, the number of combinations of the parameters for the filtering process becomes enormous. In the present embodiment, 15 types of filter processing including no processing are used as parameters of the search strategy, and as shown in the strategy management table of FIG.
A combination of five stages of filter processing is created. Of course, the type of filter and the combination of the pre-processing can be appropriately changed according to the target image.

FIG. 7 illustrates the operation of a plurality of filtering processes. For example, an image 72 is generated by applying a minimum value filter to a template image 71 including an alignment mark. In the image 72, the dark part (black) of the alignment mark is enlarged and emphasized by the minimum value filter processing. As a result, the characteristics of the alignment mark are emphasized, and the resolution can be reduced (decimated) for speeding up.

Alternatively, for the template image 71,
A vertical edge enhancement filter process is performed, and the resulting image 7
3 is subjected to the maximum value filter processing to obtain the image 74
Generate By emphasizing and expanding only the vertical component of the alignment mark, it is possible to obtain a thinned template image without losing features. As a result, the resolution of the template image used for matching can be further reduced,
Even when the template image is more complicated, stable matching processing can be performed.

FIG. 8 is an explanatory diagram of the thinning rate which is a parameter of the search strategy. For simplicity, the description will be made using 8 × 8 pixels of the template image and 14 × 14 pixels of the input image. The thinning rate is set independently in the X and Y directions (FIG. 5). In the illustrated example, the thinning rate in the X direction =
2. The thinning rate in the Y direction = 3. At this time, data used for template matching is indicated by black circle pixels.

When the thinning rate is increased (when the resolution is reduced), the amount of data used for the matching processing is reduced, so that high-speed processing can be performed. However, if the thinning rate is increased, the features of the image tend to be missing, and an appropriate correlation value between the template and the input image cannot be calculated. Therefore, it is necessary to adjust the thinning rate depending on the image to be processed.

In this embodiment, the combination of the filter processing in the pre-processing and the thinning rate in the matching processing are automatically adjusted, and the optimal combination of the parameters that can perform high-speed and stable position measurement is determined as a search strategy. , Stored in the strategy management table 63. The number of combinations of the filter processing becomes enormous, and the optimum combination changes depending on the target template image, the input image to be searched, and the resolution. Therefore, the automatic setting of the search strategy is considered as an optimal combination problem of the preprocessing and the thinning rate, and the optimization is performed using the GA. GA is one of the optimization techniques that considers a combination of parameters to be optimized as a gene and randomly generates a gene to find a combination that provides good results.

FIG. 9 is an explanatory diagram of a gene in this example. (A) is the gene structure, the thinning rate is X direction, Y
The direction is set by 4 bits (0 to 15), and the filter processing is set by 5 bits in each of 4 bits.
(B) shows an example of a gene, and the thinning rate in the X direction = 5 (010
1), Y direction = 3 (0011). Filters for pre-processing include filter B (0010), filter A (0001), filter D
(0100), filter F (0110), and no processing (0000). (C) shows 16 types of gene mapping values (4 bits) of filters A to O including no processing. Each filter in (b) is set by this gene mapping value.

FIG. 10 shows GA processing in this embodiment. First, an initial population of genes is generated (S31).
The genes of the initial population have a wide variety of n genes by random numbers. Taking FIG. 9B as an example, n bits are randomly generated using random numbers for a bit array of a 28-bit gene.

Next, n child genes are created by crossover (S32). As shown in FIG. 11, a cut surface is randomly determined from two genes, and the two genes are crossed to generate a child gene. For example, a cutting plane is set at the 10th bit by random numbers, and 10 bits or less of two parent genes are completely replaced to generate two child genes.

Next, n genes are generated from the generated 2n genes by mutation (S33). FIG.
As shown in FIG. 2, the mutation performs bit inversion with a certain probability or more for each bit. For example, 1
For the third bit, a mutation value is generated by random numbers, and if the mutation value is equal to or greater than a certain value, bit inversion is performed. Thereby, the gene can be minutely changed.

With respect to the 3n genes as described above, the degree of suitability indicating the superiority or inferiority of each gene is calculated (S34). Then, by the selection, the genes with low fitness are extinguished, and the number of genes to be left in the next generation is set to n (S35). A series of processing from S32 to S35 corresponds to one generation. It is determined whether or not any of the n genes of one generation satisfies the optimization termination condition (S
36) If the termination condition is not satisfied, the process is repeated from the crossing of S32. The termination condition is satisfied when there is a gene whose fitness exceeds a predetermined threshold value or when the number of repetitions exceeds a specified number (for example, for 30 generations).

In the present embodiment, the methods of crossover, mutation, and selection are performed as described above, but any method can be used as long as the same action can be realized. That is, it is only necessary to be able to generate a new gene from an old gene (crossover), cause a change in the gene with a certain probability (mutation), and keep the number of genes constant for each generation (selection).

Next, the calculation of the degree of conformity in step S34 will be described. In the present embodiment, the degree of matching is multiplied by each other in consideration of the two factors of the degree of recognition and the processing speed, depending on the quality of either element. Furthermore, each weighting factor m (≧
0), n (≧ 0). Weight coefficient m,
n is determined by a basic strategy of placing importance on the accuracy of the position measurement and the processing time, and can usually be handled almost fixedly. The fitness of the present embodiment is defined as in Equation 1.

[0045]

## EQU00001 ## Fitness = (m-th power of recognition) .times. (N-th power of processing speed) Since the processing speed is inversely proportional to the calculation time of the correlation value, the thinning rate, the size of the template or the search area, Depends on preprocessing time. On the other hand, the degree of recognition depends on “easiness of finding a correct answer position” and “difficulty of incorrect recognition”, and each element depends on a correlation value between the correct answer position and the incorrect answer position. The recognition degree in the present embodiment is defined as in Expression 2.

[0046]

## EQU2 ## Recognition level = (easy to find correct position) × (difficult to recognize) Easy to find correct position = correlation value at correct position−threshold 1 Difficulty to recognize = correct position Correlation value−maximum correlation value at incorrect answer position−threshold 2 where “easy to find correct answer position”
In addition, the value of 0 or less of “difficulty of misrecognition” is 0. In order to determine the degree of recognition of Expression 2, mapping processing between the template and the input image is performed, and a “correlation value at a correct answer position” and a “maximum correlation value at an incorrect answer position” are obtained. This mapping process uses the thinning rate of genetic information and preprocessing (combination of filtering processes).

FIG. 13 shows a conceptual diagram of the correlation value distribution obtained by the mapping process. The peak 61 having the highest correlation value is the “correlation value at the correct answer position”, and the peak 62 having the next highest correlation value is the “maximum correlation value at the incorrect answer position”. From Equations (1) and (2), the better the fitness, the shorter the calculation time of the correlation value,
Also, the larger the correlation value 61 and the smaller the correlation value 62, the larger.

In this embodiment, since the degree of conformity is defined as described above, it is possible to freely set the priority of the processing speed or the recognition rate by adjusting the combination of each element. Can also.

Further, in order to determine a strategy for performing template matching stably, it is necessary to evaluate a data set used for template matching processing for all sets.

FIG. 14 shows the relationship between the thinning rate and the fitness. As illustrated, when the thinning rate in the X direction is set to 2 and the thinning rate in the Y direction is set to 3 for the input image, the data groups used in the matching process are six types (a) to (f). . When it is necessary to realize stable matching processing for all of these six types of data, the matching is calculated by performing all six types of matching processing and calculating the fitness by applying the equation (1). That is, as the highest correlation value 62 at the incorrect answer position, the highest value of the incorrect answer correlation values is selected from the six types of data thinned out from the input image. However, it is not an essential requirement to consider all data groups as described above when calculating the fitness.

In this example, the degree of conformity was calculated by the formulas shown in Equations 1 and 2. However, each element of the degree of recognition was weighted, and the “correlation value of the correct position” and the “correlation value of the incorrect position” Various modifications are possible, such as obtaining the degree of recognition from the difference between the “maximum correlation values”. Furthermore, a completely different calculation formula may be used as long as the degree of fitness having the same properties can be calculated.

In this embodiment, GA is adopted as a method of optimizing a parameter combination. However, if a method of optimizing a parameter of a certain evaluation function (fitness) is used, another method such as a steepest descent method may be used. Is also good.

As described above, in the embodiment of the present invention, it is possible to automatically set the optimal position measurement strategy for the image to be used. The strategy is optimized so as to be faster and more stable, and the adjustment work conventionally performed by the worker can be performed simply and efficiently.

Each unit of the position measuring device shown in FIG. 4 may be realized by a plurality of physical units as shown in FIG.
Alternatively, functions by a plurality of means may be realized by being combined into one physical means. Incidentally, in this embodiment, a general PC (operating frequency of about 200 MHz) is used and realized by one physical means. To find the optimal position measurement strategy using this PC, a template image (60 ×
When using the search image (512 × 440 pixels) and the search image (512 × 440 pixels), it takes only about 500 seconds. From the point of view of the working time that was adjusted by trial and error by a conventional manual operation, it is equivalent to nothing.

[0055]

According to the position measurement of the present invention, it is possible to automatically determine in advance the optimum combination of the filter processing capable of maintaining or enhancing the characteristics of the image to be used. The work time required for the pre-processing of the image can be greatly reduced, and the operator does not need to be skilled because it is a simple work of setting the template area and the search area.

Further, since the optimum combination of the image resolution (thinning rate) and the optimum combination of the pre-processing filter processing and the image resolution can be automatically determined, the position measurement of the target can be performed quickly and stably. effective.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a position measuring device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a general processing procedure of position measurement using an image.

FIG. 3 is a flowchart showing a conventional search strategy determination procedure.

FIG. 4 is a schematic explanatory view in which the position measuring device of the present invention is applied to printed circuit board inspection.

FIG. 5 is an explanatory diagram showing search strategy parameters and a data configuration of a search strategy table.

FIG. 6 is an explanatory diagram showing filter coefficients of a plurality of filter processes.

FIG. 7 is an explanatory diagram showing an operation and an effect obtained by combining a plurality of filter processes.

FIG. 8 is an explanatory diagram of a thinning rate applied to a template and an input image.

FIG. 9 is an explanatory diagram showing the structure of a gene according to one example.

FIG. 10 shows a genetic algorithm (GA) according to one embodiment.
The flowchart which shows the processing procedure of.

FIG. 11 is an explanatory diagram showing processing contents (crossover) of GA.

FIG. 12 is an explanatory diagram showing processing contents (mutation) of GA.

FIG. 13 is an explanatory diagram showing a distribution of correlation values by a matching process.

FIG. 14 is an explanatory diagram showing the relationship between the thinning rate and the calculation of the degree of matching.

[Explanation of symbols]

1. Search strategy setting mechanism 2. Sample image acquisition means 3,
... Position measurement template area setting means, 4 ... Template search area setting means, 5 ... Position measurement strategy automatic determination means,
6 storage means, 7 position measurement mechanism, 8 image acquisition means,
9 Position measurement processing means, 10 Result output means, 41 Camera, 42 Position measurement device, 43 Stage, 44 Display, 45 Image, 46 Template, 47
Alignment mark, 48: Search area, 49: Printed circuit board, 63: Search strategy table.

Continued on the front page (72) Inventor Hiroshi Otsuka 7-1-1, Omika-cho, Hitachi City, Ibaraki Prefecture Inside Hitachi Research Laboratory, Hitachi, Ltd. (72) Inventor Yoshiki Kobayashi 7-1-1, Omika-cho, Hitachi City, Ibaraki Prefecture Hitachi, Ltd.Hitachi Research Laboratories (72) Inventor Eiji Shimizu 5-2-1 Omikacho, Hitachi City, Ibaraki Prefecture Hitachi Process Computer Engineering Co., Ltd. (72) Inventor Toshihiko Imai 5-chome Omikamachi, Hitachi City, Ibaraki Prefecture No. 2 No.1 Hitachi Process Computer Engineering Co., Ltd. F term (reference) 2F065 AA03 BB27 CC01 DD06 FF04 JJ03 JJ19 JJ26 QQ24 QQ31 QQ33 QQ39 QQ41 SS13 TT02 5B057 AA03 BA02 BA24 CD07 CE06 DA07 DA16 DC33

Claims (8)

[Claims] 1. A position measurement method for inputting a target image, performing predetermined preprocessing, performing a search process using a template including a feature of a predetermined position of the target image, and measuring the predetermined position, For the template and the search area set based on the sample image of the target image, a combination of parameters consisting of a plurality of types and order of filter processing is selected so that the characteristics of the image are maintained or emphasized. A position measurement method, wherein the pre-processing is performed based on a combination of selected parameters at the time of position measurement. 2. A position measuring method for measuring a predetermined position by inputting a target image, performing a predetermined pre-processing, performing a search process using a template including a feature of a predetermined position of the target image, and measuring the predetermined position. For the template and the search area set based on the sample image of the target image, the types and order of the plurality of filter processes for the preprocessing, the thinning rate of the image for the search process in the X direction and the Y direction, A combination of parameters including a thinned-out position is selected so that the search processing is speeded up while maintaining the characteristics of the image. In the position measurement using the target image, the pre-processing and the search are performed by a combination of the selected parameters. A position measurement method comprising performing processing. 3. The method according to claim 2, wherein a coarse position measurement is performed by lowering the resolution of an image to be used by applying the combination of the parameters, and as a result, the search area is narrowed and the resolution is restored. A position measurement method characterized by performing dense position measurement. 4. The position according to claim 1, wherein the combination of the parameters is selected such that the degree of recognition by the matching process between the template and the sample image is high and the processing time is short. Measurement method. 5. The position measurement method according to claim 4, wherein the degree of recognition is determined depending on a difference between a correlation value at the predetermined position by the matching processing and a maximum correlation value at positions other than the predetermined position. 6. The image processing apparatus according to claim 5, wherein the maximum correlation value at a position other than the predetermined position is a result of a matching process for all the image data when a plurality of image data are obtained from a thinning rate set for an image to be used. A position measuring method characterized by obtaining from a position. 7. A position measuring device that inputs a target image, performs predetermined preprocessing, performs a search using a template including a feature of a predetermined position of the target image, and measures the predetermined position. Template region setting means for setting a template used for position measurement based on the sample image, template search area setting means for setting a search area for scanning the template, and a template and a search area set by these means, A parameter selection unit for selecting a combination of parameters including a type and an order of filter processing to be performed on the sample image as the pre-processing so that characteristics of the image are maintained or emphasized. Position measuring device. 8. A position measuring device for inputting a target image, performing predetermined pre-processing, performing a search process using a template including a feature of a predetermined position of the target image, and measuring the predetermined position. A template area setting means for setting a template used for position measurement based on a sample image of the target image, a template search area setting means for setting a search area for scanning the template, and a template set by these means. Using a search area, the type and order of filter processing to be performed on the sample image as the pre-processing and a combination of parameters including a thinning rate and a thinning position in the X and Y directions of the sample image are maintained while maintaining the characteristics of the image. Search parameter selection means for selecting so as to speed up the search process Comprises a decision mechanism, the search according to a combination of the selected parameter strategy determination mechanism, the position measuring apparatus characterized by being configured to perform search processing and pre-processing of the position measurement mechanism.