JP2011123737A - Object recognition device, object recognition method, and program - Google Patents

Abstract

Object recognition accuracy is improved even when short-term and long-term luminance fluctuations occur in a part of a searched image.
A correlation value between a template image piece obtained by dividing a template image into a predetermined number and a partial image piece obtained by extracting a partial image from the searched image and dividing it into a predetermined number is obtained. The number of correlation values that are equal to or greater than the predetermined threshold is totaled to obtain a total value (S5 to S9). As a result of scanning the entire searched image, one total value that is the maximum value is extracted from the total values that are equal to or greater than the first predetermined value that is set in advance. The position of the corresponding partial image is recognized as a matching position (S15). When the total value when the matching position is recognized in S15 is equal to or smaller than the second predetermined value set to a value smaller than the predetermined number and larger than the first predetermined value, the template image is selected in S15. The partial image corresponding to the recognized matching position is replaced (S16, S17).
[Selection] Figure 3

Description

  The present invention relates to an object recognition apparatus, an object recognition method, and a program for recognizing an object similar to a template image by pattern matching.

  In the field of image processing, pattern matching is often used as a method for recognizing an object in a searched image. Pattern matching is a method in which an object to be recognized is registered in advance as a reference image called a template image, and an image in a region similar to the template image is found from the searched image. More specifically, the template image is scanned in the image to be searched, and the similarity between the template image and the partial image of the region overlapping the template image in the image to be searched is calculated at each position. As a result, the partial image in the searched image having the highest similarity is set as the matching position. A correlation value is used as an index indicating similarity, and a normalized correlation method is known as a method for obtaining the correlation value. The normalized correlation method is a method for normalizing the luminance values of both the template image and the partial image in the searched image and obtaining the correlation function value (normalized correlation value) of both, and is a uniform and linear luminance value. It has the characteristic of not being greatly affected by changes.

  As a method using such a normalized correlation method, the template image is replaced with a partial image that matches the template image in the searched image every time pattern matching is continuously performed on a plurality of searched images. A technique is known (see Patent Document 1). In addition, a method is known in which a template image and a partial image in a searched image are divided into a predetermined number and pattern matching is performed between the image divided pieces (see Patent Document 2). In this method, when the correlation value between the image segmentation pieces is lower than a predetermined value in the preparatory work, the image segmentation piece of the template image is set as a deadband region. It is not used.

JP 2006-214816 A JP-A-6-89343

  However, the above-described conventional method is strong when the luminance is uniform and linearly changed, for example, when the luminance of the illumination is uniformly deteriorated. Sensitive to external disturbances. Such partial luminance changes include short-term luminance changes and long-term luminance changes. For example, the short-term luminance change is affected by shadows caused by the operation of the device or human beings, and the long-term luminance change is a variation in deterioration among lamps in a lighting device using a plurality of lamps. There is an influence of shading generated by.

  In the conventional method of updating the template image every time template matching is performed, if a luminance change occurs in a part of the partial image of the searched image, the correlation value decreases due to the portion where the luminance change occurs, and the detected object The accuracy of recognizing is reduced. In other words, simply updating the template image each time will not only detect the detected object for short-term luminance changes, but also the error will be easily increased because the template image is updated frequently, and detection will take place over a long period of time. It becomes impossible to recognize things. Further, in the conventional method of setting the dead zone, if a luminance change occurs in a portion other than the dead zone, the correlation value is lowered due to the portion where the luminance change occurs, and the accuracy of recognizing the detected object is lowered. . That is, the detection object cannot be recognized for a short-term luminance change simply by setting the dead zone. Further, if the dead zone region is set wide with respect to the long-term luminance change, the area of the template image necessary for similarity determination is reduced, so that the detected object cannot be recognized.

  Accordingly, the present invention provides an object recognition apparatus, an object recognition method, and a program capable of improving the object recognition accuracy even when short-term and long-term luminance fluctuations occur in a part of the searched image. is there.

  The present invention provides an object recognition apparatus that scans a template image in a searched image and performs pattern matching, a template image dividing unit that divides the template image into a predetermined number of template image pieces, and the template from the searched image. A partial image having the same size as the image and extracting the partial image into a predetermined number of partial image pieces corresponding to each template image piece; and the correlation values of the template image piece and the partial image piece Calculating means for calculating a total value by counting the number of correlation values that are equal to or greater than a predetermined threshold, and a preset value among the total values obtained by the calculation means as a result of scanning the entire searched image One aggregate value that is the maximum value is extracted from the aggregate values that are equal to or greater than the first predetermined value, and the position of the partial image corresponding to the aggregate value is extracted. A recognizing means for recognizing a matching position, and a second value in which the total value when the matching position is recognized by the recognizing means is set to a value smaller than the predetermined number and larger than the first predetermined value. And updating means for replacing the template image with a partial image corresponding to the matching position recognized by the recognition means when the value is equal to or less than a predetermined value.

  In the object recognition apparatus that scans a template image in a searched image and performs pattern matching, the present invention provides a template image dividing unit that divides the template image into a predetermined number of template image pieces, and the searched image. A partial image having the same size as the template image is extracted and divided into a predetermined number of partial image pieces corresponding to each template image piece, and each of the template image piece and the partial image piece As a result of scanning the whole of the image to be searched as a result of scanning the whole of the searched image, calculating means for calculating a correlation value, totaling the number of correlation values that are equal to or greater than a predetermined threshold, One aggregate value that is the maximum value is extracted from the aggregate values that are equal to or greater than the first predetermined value that is set in advance, and the partial image corresponding to the aggregate value is extracted. A recognition means for recognizing a position as a matching position, and a total value when the matching position is recognized by the recognition means in the current scan is a value smaller than the predetermined number and larger than the first predetermined value And a total value when the matching position is recognized by the recognition means in the past scan and when the matching position is recognized by the recognition means in the current scan. Updating means for replacing the template image with a partial image corresponding to the matching position recognized by the recognition means in the current scan when the difference value from the total value is equal to or less than a preset third predetermined value; , Provided.

  According to the present invention, the partial image corresponding to the total value that is the maximum value is recognized as the matching position among the total values that are the total value of the correlation values that are equal to or greater than the predetermined threshold value that are equal to or greater than the first predetermined value. As a result, luminance fluctuations partially occur in the partial image, and the object can be recognized even if a part of the correlation value is lowered, and the object recognition accuracy is improved against short-term and long-term luminance fluctuations. be able to.

  Furthermore, by replacing the template image based on the total value of the recognized partial images, even if the luminance gradually changes and long-term luminance fluctuation occurs in a part of the image, it is close to the illumination environment of the searched image Updated to template image. Therefore, the recognition accuracy of the object can be improved over a long period of time.

It is explanatory drawing which shows schematic structure of the object recognition apparatus which concerns on 1st Embodiment of this invention. It is a figure which shows a to-be-searched image and a template image. It is a flowchart which shows the recognition operation of an object recognition apparatus. It is a flowchart which shows the processing operation | movement of the connector recognition shown in FIG. It is a figure for demonstrating the operation | movement which scans the inside of a to-be-searched image using the divided template image, and performs pattern matching. It is a figure for demonstrating the operation | movement which calculates | requires a correlation value. It is a flowchart which shows the recognition operation | movement of the object recognition apparatus which concerns on 2nd Embodiment of this invention. It is a flowchart which shows the recognition operation | movement of the object recognition apparatus which concerns on 3rd Embodiment of this invention. It is a figure for demonstrating the operation | movement which updates a template image.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention will be described in detail with reference to the drawings.

[First Embodiment]
FIG. 1 is an explanatory diagram showing a schematic configuration of an object recognition apparatus according to the first embodiment of the present invention. An object recognition device 30 shown in FIG. 1A includes a surface plate 32 on which a printed circuit board 31 is placed, a camera 33 as an imaging unit that images the printed circuit board 31, and an illumination device that irradiates the printed circuit board 31 with light. 34. Further, the object recognition device 30 includes an image processing device 35 that captures a search target image captured by the camera 33, scans a template image in the search target image, and performs pattern matching processing. The camera 33 includes a connector 36 that forms an optical image with a lens (not shown), converts the optical image into an imaging signal with an imaging device (not shown), and provides the imaging signal to the image processing device 35 as an image to be searched. It is an imaging means for obtaining an image. The illumination device 34 is ring illumination attached to the camera 33.

  The image processing device 35 is a computer having a CPU 35A that controls the entire object recognition device, a ROM 35B that stores a program P for operating the CPU 35A, and a RAM 35C that temporarily stores data as a work area. In addition, the image processing device 35 includes a storage unit 35D configured by a rewritable nonvolatile memory (for example, EEPROM) or an HDD as a storage unit for storing data such as a template image and a search target image. As shown in FIG. 1B, the printed circuit board 31 is provided with a connector 36. In the first embodiment, the connector 36 is a detection object (object recognition target).

2 shows, the search image I and the template image I _T is illustrated. In FIG. 2A, the searched image I including the printed circuit board 31 captured by the camera 33 is stored in the storage unit 35D. The storage unit 35D, in advance was stored the image cut out a peripheral portion of the image from the connector 36 including a printed circuit board serving as a reference as the template image I _T. The scope of the template image I _T, it is possible for the user to arbitrarily set.

  Next, the operation flow of the CPU 35A based on the program P shown in FIG. 3 will be described with reference to FIGS. First, as shown in FIG. 3, the CPU 35A gives an instruction to the camera 33 to start imaging, acquires the searched image I, and stores the searched image I in the storage unit 35D (S1).

Then, CPU 35a reads out the template image _{I T} which has been stored in advance in the storage unit 35D (S2), divided into the template image piece _{i t} in FIG. 5 a predetermined number as shown in (a) N4 (e.g. 12) ( S3: Template image dividing step). The process of S3 by the CPU 35A functions as a template image dividing unit.

Further, CPU 35a, as shown in FIG. 5 (b), extracting a partial image _{I S} of the template image _{I T} and the same size from the search image I, the predetermined number corresponding to each template image pieces _{i t} N4 ( = 12) is divided into partial image section _{i s} of (S4: partial image dividing step). The process of S4 by the CPU 35A functions as a partial image dividing unit.

Then, CPU 35a performs pattern matching by using the divided template image piece _{i t} and the partial image section _{i s.} Specifically, CPU 35a is extracted partial image _I with allocating numbers 1-12 in each partial image section _{i s} of _S, allocate numbers 1-12 in each template image pieces _{i t,} each image piece _i t, correspondence between the i _s with each other to identify the (same part number). Then, CPU 35a sets the variable k indicating the number 1 (S5), the image pieces _i t corresponding to the variable k, obtained by calculating the correlation values of _{i s} between (S6). For the calculation of the correlation value, a normalized correlation method is used. The following formula 1 is used as a calculation formula for obtaining the correlation value.

The position of the image pieces (i, j) and, as shown in FIG. 6, the partial image piece i of the partial image I _S obtained by extracting the image of a size corresponding to the template image I _T from the search image I a position or positions in the template image piece i _t of the template image I _T in _s. That is, the position (i, j) are the coordinates expressed in pixels, determine the _{f ij, g ij} for each pixel in each image piece _i t, _{i s,} obtains a correlation value R. For the correlation value R, 1 represents the highest similarity, and -1 represents the lowest similarity.

Next, the CPU 35A determines whether or not k = N4 (12) (S7). If it is determined that k is smaller than N4 (S7: NO), k is incremented (S8). The process proceeds to S6 in which a correlation value between the image pieces i _t and i _s corresponding to the number of is obtained by calculation.

CPU35A is, k = N4, i.e. all the image pieces _i t, if it is determined that the correlation value of the _{i s} between (S7: YES), or greater than a predetermined threshold value to a certain similarity (e.g. 0.95 or higher) The total number of correlation values is calculated to obtain a total value (S9). That is, S6 to S9 correspond to calculation steps, and the processing operations of S35 to S9 of the CPU 35A function as calculation means. The predetermined threshold is set to a value that is recognized as a match between the image pieces.

Next, the CPU 35A determines whether or not the total value obtained in S9 is equal to or greater than a preset first predetermined value N1 (for example, 5) (S10). The first predetermined value N1 is set to a value smaller than the predetermined number N4. CPU35A, if the aggregate value determined in S9 equal to or greater than the first predetermined value N1 (S10: YES), recognizes the position of the partial image _{I S} corresponding to the aggregate value as match candidate position (S11). The first predetermined value N1, it and the template image I _T and the partial image I _S is set to a value that is determined to potentially meets. That is, if the total value is greater than or equal to the first predetermined value (five or more), it is recognized as a candidate for a specific detected object (connector 36). At this time, the position and the aggregate value of the partial image I _S in the search image in the I as match candidate position of the connector 36, and stores a correlation value between each template image piece i _t and the partial image section i _s in the storage unit 35D .

Incidentally, CPU 35a, when the aggregate value determined in S9 below a first predetermined value N1 (S10: NO), that is, if four or less, does not match the partial image _{I S} and the template image _{I T,} i.e. It is recognized that it is not a specific connector 36 (S12).

Next, the CPU 35A determines whether or not the entire searched image I has been scanned (S13). CPU35A, if not to scan the entire search subject image I (S13: NO), shifting the region in pixels, then specifying a partial image _{I S} for pattern matching (stored in the storage unit 35D) and (S14) The process proceeds to step S4.

When the entire scanning of the searched image I is completed (S13: YES), the CPU 35A proceeds to connector recognition processing (S15: recognition step). In S15, the CPU 35A repeats the process of S14, and among the one or more aggregate values obtained in S9, the aggregate value that is the maximum value among the aggregate values that are equal to or greater than the first predetermined value N1 is 1 One extracted, recognizes the position of the partial image I _S corresponding to the total value and the matching position. That, CPU 35a recognizes the matching position of the partial image _{I S} that matches the template image _{I T.} The processing operation of S15 by the CPU 35A functions as a recognition unit. Next, the CPU 35A outputs the recognized position of the connector 36 as a recognition result in two-dimensional coordinates (x1, y1). The coordinates (x1, y1) are positions in the searched image I.

  Next, the CPU 35A sets a second predetermined value N2 (for example, 7) in which the total value when the matching position is recognized in S15 is set to a value smaller than the predetermined number N4 and larger than the first predetermined value N1. It is determined whether or not the following is true (S16). If the CPU 35A determines in S16 that the value exceeds the second predetermined value N2 (S16: NO), it ends the pattern matching. The second predetermined value N2 is set to a boundary value at which the brightness may affect the accuracy of object recognition between the recognized partial image and the set template image. That is, when the total value is equal to or smaller than the second predetermined value, the brightness of the lighting device 34 is lower than when the template image is set.

Therefore, the CPU 35A updates the template in S16 if it is equal to or smaller than the second predetermined value N2 (= 7) (S17: update step). That, CPU 35a replaces the template image _{I T,} the partial image _{I S} corresponding to the recognized matching position S15. Specifically, CPU 35a stores the partial image _{I S} position recognized as a connector 36 in the storage unit 35D as a new template image _{I T.}

  The object recognition operation on the first printed circuit board 31 is completed by the above operations S1 to S17. However, the recognized printed circuit board 31 is replaced with the next printed circuit board 31, and the object recognition on the next printed circuit board 31 is performed. When the operation is performed, the process returns to S1 to execute the process.

By the operation of S15, occurs partially brightness variation in the partial image I _S, a part of it is that the correlation value to recognize an object be reduced, the recognition of the object with respect to short-term and long-term brightness variation Accuracy can be improved. Further, for example, even when the illumination device 34 deteriorates due to long-term use of the illumination device 34 and the luminance gradually decreases and a partial luminance change occurs in the captured image I to be searched, the partial image I recognized in S15. _When the total value of _S is equal to or smaller than the second predetermined value N2, the template image _IT is replaced. Therefore, the template image _IT can be updated to a state close to the current lighting environment when the image to be searched I is imaged, and the object can be prevented from becoming unrecognizable. Accuracy can be improved. Further, instead of updating every time the template image I _T every time the pattern matching, since the provided criteria in S16, will contain a large error in error is accumulated in the template image I _T as the recognition criteria Can be prevented in advance. Thereby, it can recognize correctly over a long period of time.

  Here, the connector recognition process in S15 will be described in detail. As shown in FIG. 4, as a result of scanning the entire searched image I, the CPU 35A determines whether or not the number of match candidate positions recognized in S11 is plural (S21). That is, it is determined whether or not there are a plurality of aggregate values that are equal to or greater than the first predetermined value N1 (= 5).

When the CPU 35A determines that there are a plurality of match candidate positions in the determination process of S21 (S21: YES), the partial image corresponding to the maximum value among the total values corresponding to the match candidate positions. extracting a matching candidate positions I _S (S22).

  Next, the CPU 35A determines whether or not the number of match candidate positions extracted in S22 is plural (S23). That is, it is determined whether or not there are a plurality of aggregate values having the maximum value. Next, the CPU 35A calculates the total value of the correlation values used when totaling the total values corresponding to these match candidate positions (S24). Next, the CPU 35A recognizes a match candidate position corresponding to the total value that is the maximum value among the total values as a match position (S25), and proceeds to the determination process of S16 shown in FIG.

  Further, when the CPU 35A determines that the number of match candidate positions is 1 in S21 (S21: NO), or determines that the number of match candidate positions is 1 in S23 (S23: NO), the match candidate position Is recognized as a match position (S26). Note that the CPU 35A causes the storage unit 35D to store the total value when it is recognized as the matching position in the processing of S25 and S26.

Thus, S3, S4 divides the template image _{I T} and the partial image _{I S} in the template image piece _{i t} and the partial image sections _{i s} of a predetermined number N4 in each image piece _i t at S6, _{i s} together The correlation value is obtained. Then, it occurs term brightness variations short-term brightness variations and shading partially shadow occurs in the partial image I _S is generated, even if some of the correlation values is reduced, the aggregate value is determined in S10 first If it is greater than or equal to the predetermined value N1, it is recognized as a match position in S25 and S26. Therefore, the recognition accuracy of an object can be improved with respect to a part of luminance fluctuations. Furthermore, if it is determined in the determination process of S21 that there are a plurality of match candidate positions (total values that are equal to or greater than the first predetermined value N1), match candidate positions corresponding to the maximum total value among the total values Is the matching position. Therefore, even when there are a plurality of match candidate positions, the match positions can be accurately determined.

[Second Embodiment]
Next, an object recognition apparatus according to the second embodiment will be described. The configuration of the object recognition apparatus of the second embodiment is the same as the configuration shown in FIG. FIG. 7 shows an operation flow of the object recognition apparatus of the second embodiment. In addition, the same code | symbol as FIG. 3 is attached | subjected to the step which performs the process same as the process demonstrated in the said 1st Embodiment.

First, the CPU 35A sequentially executes the processes of S1 to S5 as in the first embodiment. Then, CPU 35a obtains the image piece _i t corresponding to the number k, by calculating the correlation values of _{i s} between (S6-1). For the calculation of the correlation value, a normalized correlation method is used. The above-described formula 1 is used as a calculation formula for obtaining the correlation value.

  Next, the CPU 35A counts the number m of correlation values smaller than a predetermined threshold (= 0.95) among the k correlation values obtained in S6-1 (S6-2). Next, the CPU 35A determines whether or not the number m of correlation values that are smaller than a predetermined threshold exceeds a difference value obtained by subtracting the first predetermined value N1 from the predetermined number N4, that is, whether m> N4−N1. It is determined whether or not (S6-3). For example, in the case of N4 = 12, N1 = 5, it is determined whether the number m of correlation values exceeds 8 and becomes 8, that is, whether m> 7.

CPU35A, when obtaining a correlation value sequentially, when the number m of the correlation value has exceeded a difference value (N4-N1) (S6-3: YES), the remainder of the template image piece _{i t} and the partial image section i The calculation of the correlation value with _s is terminated, that is, the process proceeds to S12. That is, in the image to be searched I, the number of partial images that are match candidates is extremely small compared to the total number of partial images that scan the image to be searched. Therefore, for the partial image that cannot be a match candidate position in S6-3, the calculation of the correlation value is terminated halfway. If the CPU 35A determines that the number m of correlation values does not exceed the difference value (N4-N1) (S6-3: NO), the CPU 35A proceeds to S7. In S7, as described above, it is determined whether or not k = N4. If k = N4 is not satisfied, k is incremented (S8), and the process proceeds to S6-1. If k = N4, the process proceeds to S9.

As described above, according to the second embodiment, the effects similar to those of the first embodiment, since the calculation of the correlation values with the rest of the template image piece i _t and the partial image section i _s aborted, the first Compared to the first embodiment, the amount of calculation is reduced, and the object recognition process is speeded up.

[Third Embodiment]
Next, an object recognition apparatus according to the third embodiment will be described. The configuration of the object recognition apparatus of the third embodiment is the same as the configuration shown in FIG. FIG. 8 shows an operation flow of the object recognition apparatus of the third embodiment. In addition, the same code | symbol as FIG. 3 is attached | subjected to the step which performs the process same as the process demonstrated in the said 1st Embodiment. In the third embodiment, the determination whether or not to update the template image is different from that in the first embodiment.

The searched image is picked up by the camera 33 every time the printed circuit board 31 shown in FIG. First, the flow shown in FIG. 3 is executed until the scanning of the _first image to be searched I1 is completed. At this time, as shown in FIG. 9, the aggregate value first partial image I _S that has been recognized as the matching position in the search images I ₁ is stored in the storage unit 35D (FIG. 1).

Next, when the printed circuit board 31 is replaced and the second and subsequent searched images I ₂ are captured by the camera 33, the CPU 35A executes each process shown in FIG. Specifically, the CPU 35A executes the processes of S1 to S15, and then the total value 2 when the matching position is recognized in S25 or S26 (FIG. 4) in the current scan is the second predetermined value N2 ( = 7) It is determined whether or not (S16-1). Here, the storage unit 35D, past aggregate value 1 at the time of the matched position is recognized by S25 or S26 in the search image I ₁ of the scanning of the (last) is stored.

  Therefore, when the total value 2 is equal to or smaller than the second predetermined value N2 (S16-1: YES), the CPU 35A determines the total value 1 stored in the storage unit 35D when the matching position is recognized in the previous scan and the current time. In this scanning, a difference value from the total value 2 when the matching position is recognized in S25 and S26 is obtained. Then, the CPU 35A determines whether or not the obtained difference value is equal to or less than a preset third predetermined value N3 (S16-2). The third predetermined value N3 is a value (for example, 3) equal to or less than (N4−N1).

That is, the third predetermined value N3 becomes the partial images recognized by the previous scan of the search image I _1, and significantly different luminance change in the current partial image recognized by the scanning of the search image I ₂ Set to the boundary value. That is, in S16-2, the magnitude of short-term luminance fluctuation of the lighting device 34 is determined. When the luminance variation of the lighting device 34 is small, the obtained difference value is equal to or smaller than the third predetermined value, and when the luminance variation of the lighting device 34 is large, the difference value exceeds the third predetermined value.

CPU35A, if the difference value is less than the third predetermined value N3 (S16-2: YES), the template image _{I T,} replaced by current scan in S25, the recognized partial image and the matching position at S26, the template update The process is executed (S17). When the total value 2 exceeds the second predetermined value N2 (S16-1: NO), or when the difference value exceeds the third predetermined value N3 (S16-2: NO), the CPU 35A does not update the template. The pattern matching is finished.

  In the third embodiment, it is possible to determine the magnitude of the luminance variation between the connector region (partial image) recognized by the previous pattern matching and the connector region (partial image) recognized by the current pattern matching. Thus, if NO in S16-2, it is a short-term luminance change, and if YES in S16-2, it can be distinguished that it is a long-term luminance change, not a short-term luminance change. Therefore, it is possible to prevent the template image from being updated when a shadow is produced in the short term due to the operation of the apparatus or human beings, and it is possible to prevent an increase in the error of the template image. Therefore, the recognition accuracy of the object can be improved over a long period of time.

  In addition, although this invention was demonstrated based on the said embodiment, this invention is not limited to this. The processing of the third embodiment can be applied to the processing of the second embodiment. Further, since the functions of the present invention are implemented by computer, the program code installed in the computer also implements the present invention, and the present invention includes the computer program itself. The program can be supplied by connecting to a homepage on the Internet using a browser of a client computer and downloading the program of the present invention from the homepage to a recording medium such as a hard disk. It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer is also included in the present invention. In addition, the case where the computer performs the function of the above-described embodiment by performing part or all of the actual processing based on the instruction of the program and by the OS running on the computer being included.

30 Object recognition device 31 Printed circuit board 32 Surface plate 33 Camera 34 Illumination device 35 Image processing device 35A CPU (template image division means, partial image division means, calculation means, recognition means, update means)
I the search image _{I S} partial image _{i s} partial image fragment _{I T} template image _{i t} template image piece

Claims (9)

In an object recognition device that scans a template image in a searched image and performs pattern matching,
Template image dividing means for dividing the template image into a predetermined number of template image pieces;
A partial image dividing means for extracting a partial image having the same size as the template image from the searched image and dividing it into a predetermined number of partial image pieces corresponding to the template image pieces;
Calculating means for obtaining a correlation value between each of the template image piece and the partial image piece, and calculating a total value by counting the number of the correlation values equal to or greater than a predetermined threshold;
As a result of scanning the entire image to be searched, one total value that is the maximum value is extracted from the total values that are equal to or greater than the first predetermined value that is set in advance, among the total values obtained by the calculation means. Recognizing means for recognizing the position of the partial image corresponding to the total value as a matching position;
When the total value when the matching position is recognized by the recognition means is equal to or smaller than the second predetermined value set to a value smaller than the predetermined number and larger than the first predetermined value, the template An object recognizing apparatus comprising: an updating unit that replaces an image with a partial image corresponding to the matching position recognized by the recognizing unit. In an object recognition device that scans a template image in a searched image and performs pattern matching,
Template image dividing means for dividing the template image into a predetermined number of template image pieces;
A partial image dividing means for extracting a partial image having the same size as the template image from the searched image and dividing it into a predetermined number of partial image pieces corresponding to the template image pieces;
Calculating means for obtaining a correlation value between each of the template image piece and the partial image piece, and calculating a total value by counting the number of the correlation values equal to or greater than a predetermined threshold;
As a result of scanning the entire image to be searched, one total value that is the maximum value is extracted from the total values that are equal to or greater than the first predetermined value that is set in advance, among the total values obtained by the calculation means. Recognizing means for recognizing the position of the partial image corresponding to the total value as a matching position;
The total value when the matching position is recognized by the recognition means in the current scan is less than or equal to a second predetermined value set to a value smaller than the predetermined number and larger than the first predetermined value. Thus, a difference value between the total value when the matching position is recognized by the recognition unit in the previous scan and the total value when the matching position is recognized by the recognition unit in the current scan is set in advance. And an update unit that replaces the template image with a partial image corresponding to the matching position recognized by the recognition unit in a current scan when the value is equal to or less than a third predetermined value. apparatus.   When there are a plurality of aggregate values that are maximum values among the aggregate values that are equal to or greater than the first predetermined value, the recognizing unit calculates a total value of correlation values that are used when the aggregate values that are the maximum values are aggregated. The object recognition apparatus according to claim 1, wherein the object recognition device calculates each and recognizes the position of the partial image corresponding to the maximum value among the total values as a matching position.   When calculating the correlation values of the template image piece and the partial image piece in order, the calculating means is configured such that the number of correlation values that are smaller than the predetermined threshold is from the predetermined number to the first predetermined value. 4. The object recognition according to claim 1, wherein when the difference value obtained by subtracting is exceeded, the calculation of the correlation value between the remaining template image piece and the partial image piece is terminated. apparatus. In an object recognition method that scans a template image in a searched image and performs pattern matching,
A template image dividing step for dividing the template image into a predetermined number of template image pieces;
A partial image dividing step of extracting a partial image of the same size as the template image from the searched image and dividing it into a predetermined number of partial image pieces corresponding to the template image pieces;
Calculating a correlation value between each of the template image piece and the partial image piece, calculating a total value by counting the number of the correlation values that are equal to or greater than a predetermined threshold; and
As a result of scanning the entire searched image, one total value that is the maximum value is extracted from the total values that are equal to or greater than the first predetermined value that is set in advance, among the total values obtained in the calculation step. A recognition step for recognizing the position of the partial image corresponding to the total value as a matching position;
When the total value when the matching position is recognized in the recognition step is equal to or less than a second predetermined value set to a value smaller than the predetermined number and larger than the first predetermined value, An object recognition method comprising: an updating step of replacing a template image with a partial image corresponding to the matching position recognized in the recognition step. In an object recognition method that scans a template image in a searched image and performs pattern matching,
A template image dividing step for dividing the template image into a predetermined number of template image pieces;
A partial image dividing step of extracting a partial image of the same size as the template image from the searched image and dividing it into a predetermined number of partial image pieces corresponding to the template image pieces;
Calculating a correlation value between each of the template image piece and the partial image piece, calculating a total value by counting the number of the correlation values that are equal to or greater than a predetermined threshold; and
As a result of scanning the entire searched image, one total value that is the maximum value is extracted from the total values that are equal to or greater than the first predetermined value that is set in advance, among the total values obtained in the calculation step. A recognition step for recognizing the position of the partial image corresponding to the total value as a matching position;
The total value when the matching position is recognized in the recognition step in the current scan is less than or equal to a second predetermined value set to a value smaller than the predetermined number and larger than the first predetermined value. The difference value between the total value when the matching position is recognized in the recognition step in the previous scan and the total value when the matching position is recognized in the recognition step in the current scan is, An update step that replaces the template image with a partial image corresponding to the matching position recognized in the recognition step in the current scan when the third predetermined value or less is set in advance. An object recognition method.   In the recognition step, when there are a plurality of aggregate values that are maximum values among the aggregate values that are equal to or greater than the first predetermined value, a total value of correlation values that are used when the aggregate values that are the maximum values are aggregated is calculated. 7. The object recognition method according to claim 5 or 6, wherein each object is calculated, and the position of the partial image corresponding to the maximum value among the total values is recognized as a matching position.   In the calculation step, when the correlation values of the template image piece and the partial image piece are sequentially obtained, the number of correlation values that are smaller than the predetermined threshold is changed from the predetermined number to the first predetermined value. 8. The object recognition according to claim 5, wherein when the difference value obtained by subtracting is exceeded, the calculation of the correlation value between the remaining template image piece and the partial image piece is terminated. Method.   A program for causing a computer to execute the object recognition method according to any one of claims 5 to 8.

Images