JP2008234102A - Indicative body recognizing device, indicative body recognition method, and indicative body recognition program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an indicative body recognition method, capable of recognizing efficiently an indicative body in a short processing time, by executing pattern-matching based on an image data and an indicating body pattern, and capable of enhancing the recognition rate. <P>SOLUTION: A matching means of an indicative body recognizer executes the shape pattern-matching with a matching area set in the whole area of the one sheet of image data, sequentially starting from the highest recognition rate of indicative body pattern. An indicating body recognizing means recognizes the indicating body existing within the matching area, based on the result of the shape pattern-matching. A recognition finish area deletion means deletes therein the recognized indicative body area as a recognition finish area from the matching area. A matching area setting means sets a residual area deleted with the recognition finish area from the matching area, as a new matching area, and repeats recognition processing for the indicating body by the shape pattern-matching. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a method for recognizing a sign object by image recognition processing.

  Map data used in a car navigation device or an application that displays a map by operating on a PC includes data of a sign object installed on a road such as a road sign or a signboard of a convenience store. Such data of the sign object is acquired by performing image recognition processing on image data obtained by photographing a road sign or signboard installed on an actual road. An example in which an image of a road sign is taken with an in-vehicle camera and the road sign is recognized by image recognition processing is described in Patent Document 1 or 2.

  For the recognition of the sign object by the image recognition process as described above, a template corresponding to a sign object that actually exists (hereinafter referred to as a “sign object pattern”) is prepared, and the sign object pattern, the captured image data, Is performed by shape pattern matching.

JP 2000-132668 A JP 2000-293670 A

  However, the conventional image recognition processing is non-batch processing, and shape pattern matching with various sign body patterns is performed a plurality of times for one piece of image data. For this reason, there has been a problem that the processing time becomes long and the number of erroneous recognitions increases.

  Examples of the problems to be solved by the present invention include those described above. An object of the present invention is to provide a sign object recognition method capable of efficiently recognizing a sign object in a short processing time and improving the recognition rate by shape pattern matching based on image data and a sign object pattern. And

  According to the first aspect of the present invention, there is provided image data storage means for storing image data to be subjected to shape pattern matching, sign object pattern storage means for storing sign object patterns to be subjected to shape pattern matching, and shape pattern matching. Based on the recognition rate data, recognition rate data storage means for storing the recognition rate of each sign body pattern as recognition rate data, matching area setting means for setting the entire area of one piece of the image data as a matching area, and the recognition rate data In order from the sign object pattern with the highest recognition rate, matching means for performing shape pattern matching with the matching area, and the sign object corresponding to the sign object pattern from the matching area based on the result of the shape pattern matching Sign object recognition means, and a mark recognized by the sign object recognition means A recognized area deletion means for deleting a body area as a recognized area from the matching area, and the matching area setting means has deleted the recognized area from the matching area by the recognized area deletion means The remaining area is set as a new matching area.

  The invention according to claim 5 is an image data storing step for storing image data to be subjected to shape pattern matching, a sign object pattern storing step for storing a sign object pattern to be subjected to shape pattern matching, and a shape pattern matching. Based on the recognition rate data, a recognition rate data storage step for storing the recognition rate of each sign object pattern as recognition rate data, a matching region setting step for setting all regions of one piece of the image data as matching regions, and the recognition rate data In order from the sign object pattern with the highest recognition rate, a matching process for executing shape pattern matching with the matching area, and a sign object corresponding to the sign object pattern from the matching area based on the result of the shape pattern matching. A sign object recognition step and a sign recognized by the sign object recognition means. A recognized area deletion step of deleting a body area as a recognized area from the matching area, and the matching area setting step deletes the recognized area from the matching area by the recognized area deletion step. The remaining area is set as a new matching area.

  The invention described in claim 6 is a sign object recognition program executed on a computer, and is an image data storage means for storing image data to be subjected to shape pattern matching, and a sign object pattern to be subjected to shape pattern matching. The sign object pattern storage means for storing the recognition rate data storage means for storing the recognition rate of each sign object pattern by shape pattern matching as the recognition rate data, and the matching area for setting the entire area of the one piece of the image data as the matching area Setting means, matching means for performing shape pattern matching with the matching area in order from a sign body pattern having a high recognition rate based on the recognition rate data, and the marking from the matching area based on the result of the shape pattern matching Sign recognition that recognizes the sign corresponding to the body pattern Means for causing the computer to function as a recognized area deleting means for deleting the area of the sign object recognized by the sign object recognizing means as a recognized area from the matching area, and the matching area setting means A residual area obtained by deleting the recognized area from the matching area by the area deleting means is set as a new matching area.

  In a preferred embodiment of the present invention, image data storage means for storing image data to be subjected to shape pattern matching, sign object pattern storage means for storing sign object patterns to be subjected to shape pattern matching, and shape pattern matching Based on the recognition rate data, recognition rate data storage means for storing the recognition rate of each sign body pattern as recognition rate data, matching area setting means for setting the entire area of one piece of the image data as a matching area, and the recognition rate data In order from the sign object pattern with the highest recognition rate, matching means for performing shape pattern matching with the matching area, and the sign object corresponding to the sign object pattern from the matching area based on the result of the shape pattern matching A sign object recognizing means that is recognized by the sign object recognizing means. A recognized area deleting unit that deletes the area of the sign body as a recognized area from the matching area, and the matching area setting unit deletes the recognized area from the matching area by the recognized area deleting unit. The remaining area is set as a new matching area.

  In the above sign recognition apparatus, image data to be subjected to shape pattern matching is stored in advance by the image data storage means. The image data is, for example, image data obtained by photographing a sign placed on a road. In the above sign object recognition device, the sign object pattern is stored in advance by the sign object pattern storage means, and the recognition rate data is stored by the recognition rate data storage means. The sign object pattern is a template of a sign object to be subjected to shape pattern matching with image data. The recognition rate data is data verified in advance, and is data of an average accuracy rate when a sign object is recognized by shape pattern matching. The matching area setting means of the sign object recognition apparatus sets the entire area of one piece of image data extracted from previously stored image data as the matching area. The matching area is an area to be subjected to shape pattern matching. Subsequently, the matching unit executes shape pattern matching with the matching region in order from the sign body pattern having the highest recognition rate based on the recognition rate data. Furthermore, the sign object recognition means recognizes the sign object existing in the matching area based on the result of the shape pattern matching. Then, the recognized area deleting means deletes the recognized sign object area as a recognized area from the matching area. The matching area setting means sets the remaining area obtained by deleting the recognized area from the matching area as a new matching area, and repeatedly performs the recognition process of the sign object by shape pattern matching.

  According to this, the sign object recognition apparatus sets the priority order of the sign object pattern based on the recognition rate, and executes shape pattern matching with the matching area in order from the sign object pattern having the highest priority. That is, the sign object is recognized by executing shape pattern matching in order from the sign object pattern with few erroneous recognitions. Further, the sign object recognition apparatus deletes the recognized sign object area as a recognized area from the matching area. In this way, shape pattern matching is performed in order from the sign object pattern with the highest recognition rate, and the area of the recognized sign object is deleted from the matching area, thereby avoiding duplicate recognition, improving the recognition rate, and speeding up the recognition process. Can be realized.

  In one aspect of the sign object recognition apparatus, the sign object pattern storage unit stores a sign object pattern having a recognition target shape and a sign object pattern having a non-recognition shape. According to this, the sign object recognition means is not the recognition object corresponding to the sign object pattern of the non-recognition shape as well as the recognition object sign object corresponding to the sign object pattern of the recognition object shape existing in the matching region. A sign object can also be recognized. Therefore, the recognized area deleting means can delete the area of the sign body from the matching area as the recognized area even when the sign object that is not the recognition target is recognized. For example, when there is a sign object that is not a recognition target but has an approximate shape with a sign object that is a recognition target, a sign object pattern corresponding to the sign object is stored in advance as a sign object pattern of an unrecognized shape. Then, when the sign object recognition apparatus recognizes a sign object having an approximate shape based on the sign object pattern of the non-recognized shape, the sign object area can be deleted from the matching area. It is possible to prevent erroneous recognition as a sign object that is a recognition target. In this way, by using a sign object other than the target sign object at the same time, erroneous recognition can be prevented and the recognition rate can be improved.

  In another aspect of the sign object recognition apparatus, the sign object recognition apparatus further includes a color area extracting unit that extracts a color area having a color that forms a basis of the sign object pattern from the matching area, and the matching area setting unit includes: The color area extracted by the color area extraction means is set as a new matching area. According to this, the matching area setting means sets the color area extracted from the matching area by the color area extracting means as a new matching area. Therefore, the matching unit executes shape pattern matching with a new matching area, that is, a color area in order from the sign body pattern having the highest recognition rate. The sign object recognition means recognizes the sign object existing in the matching area, that is, the color area, based on the result of the shape pattern matching. In this way, the sign object recognition apparatus classifies the sign object patterns, that is, the corresponding sign objects in advance, into groups based on the base color, and narrows down the matching area based on the color. By narrowing down the matching area by color, it is possible to improve the recognition rate and speed up the recognition process.

  In still another aspect of the sign object recognition apparatus, there are a plurality of types of colors that form the sign object pattern, and the color area extracted by the color area extracting unit is deleted from the matching area. The image forming apparatus further includes an area deleting unit, and the matching area setting unit sets a residual area obtained by deleting the color area from the matching area by the color area deleting unit as a new matching area. Since there are multiple types of colors that form the basis pattern of the sign object pattern, the matching area setting means selects the color of the color that forms the basis of the unrecognized pattern from the color area of the color that forms the recognition target pattern. The remaining area from which the area has been deleted can be set as the matching area. As described above, the sign object recognition apparatus can improve the recognition rate and speed up the recognition process by effectively narrowing down the matching area by using the plurality of color areas extracted by the color area extracting unit. Can do.

  In another embodiment of the present invention, a sign object recognition method includes an image data storage step for storing image data to be subjected to shape pattern matching, and a sign object pattern storage for storing a sign object pattern to be subjected to shape pattern matching. A recognition rate data storage step for storing the recognition rate of each sign object pattern by shape pattern matching as recognition rate data, a matching region setting step for setting the entire region of one piece of the image data as a matching region, In order from the sign object pattern having a high recognition rate based on the recognition rate data, a matching step for performing shape pattern matching with the matching area, and from the matching area to the sign object pattern based on the result of the shape pattern matching A sign object recognition step for recognizing a corresponding sign object, and the sign object recognition step And a recognized area deletion step of deleting from the matching area as a recognized area as a recognized area, and the matching area setting step is configured to recognize the recognition area from the matching area by the recognized area deletion step. The remaining area from which the completed area is deleted is set as a new matching area. Even with such a sign object recognition method, shape pattern matching is performed in order from the sign object pattern with the highest recognition rate, and the recognized sign object area is deleted from the matching area, thereby avoiding duplicate recognition and recognition rate. And speeding up of recognition processing can be realized.

  According to still another aspect of the present invention, there is provided a sign recognition program executed on a computer, image data storage means for storing image data to be subjected to shape pattern matching, and a sign object pattern to be subjected to shape pattern matching. The sign object pattern storage means for storing the recognition rate data storage means for storing the recognition rate of each sign object pattern by shape pattern matching as the recognition rate data, and the matching area for setting the entire area of the one piece of the image data as the matching area Setting means, matching means for performing shape pattern matching with the matching area in order from a sign body pattern having a high recognition rate based on the recognition rate data, and the marking from the matching area based on the result of the shape pattern matching A sign that recognizes the sign corresponding to the body pattern Recognizing means, causing the computer to function as a recognized area deleting means for deleting the area of the sign object recognized by the sign object recognizing means as a recognized area from the matching area, and the matching area setting means The remaining area obtained by deleting the recognized area from the matching area by the completed area deleting means is set as a new matching area. By executing such a sign object recognition program on a computer, by executing shape pattern matching in order from a sign object pattern with a high recognition rate, and deleting the recognized sign object area from the matching area, It is possible to avoid duplicate recognition, improve the recognition rate, and speed up recognition processing. This sign recognition program can be suitably handled in a state of being recorded on a recording medium.

  Hereinafter, preferred embodiments of the present invention will be described with reference to the drawings.

[Sign recognition device]
FIG. 1 is a diagram illustrating a schematic configuration of a sign object recognition apparatus 100 to which a sign object recognition method according to an embodiment of the present invention is applied.

  In the present embodiment, the sign object recognition apparatus 100 performs sign pattern matching between image data obtained by photographing a sign object installed on a road and a sign object pattern that is a template corresponding to a sign object that actually exists. It is a device that recognizes the body.

  The sign object recognition apparatus 100 is constituted by, for example, a personal computer, and includes a system bus 11, a CPU (Central Processing Unit) 12, a memory 13, a keyboard 14, a coordinate indicating device 15 such as a mouse, a display 16, and the like. A printer 17 and a database 18 are provided. Here, the keyboard 14 and the coordinate designating device 15 are input devices. The display 16 and the printer 17 are output devices. The CPU 12 controls the entire sign recognition device 100 and controls the input / output device.

  The CPU 12, the memory 13, and the database 18 are connected to the system bus 11. A keyboard 14, a coordinate indicating device 15, a display 16, and a printer 17 are also connected to the system bus 11 via an interface (not shown). The memory 13 is also used as a working memory.

[Sign recognition unit]
FIG. 2 is a functional block diagram of a sign object recognition unit 200 that recognizes a sign object by shape pattern matching between image data and a sign object pattern. The sign object recognition unit 200 is actually constituted by the components of the sign object recognition device 100. As shown in FIG. 2, the sign object recognition unit 200 includes a recognition information storage unit 201, a matching area setting unit 202, a color area extraction unit 203, a matching unit 204, a sign object recognition unit 205, a recognized area deletion unit 206, and A recognition result information storage unit 207 is provided.

  The recognition information storage unit 201 stores image data obtained by photographing a sign placed on a road. The signs installed on the road are signs such as convenience store and gas station signs, regulatory signs and warning signs, etc., and the color does not change.

  Further, the recognition information storage unit 201 stores a sign object pattern that is a template corresponding to a sign object that actually exists, and recognition rate data indicating a recognition rate by shape pattern matching of each sign object pattern.

  First, the sign body pattern will be described in detail with reference to FIGS. 3 and 4. FIG. 3 is an example of a sign object pattern (hereinafter also referred to as “recognition object pattern”) corresponding to a sign object to be recognized. FIG. 4 is an example of a sign object pattern (hereinafter also referred to as “non-recognition pattern”) corresponding to a sign object that is not a recognition target.

  In general, a simple-shaped sign object with few feature points is likely to be erroneously recognized. Therefore, the sign object recognition unit 200 also recognizes a sign object that is a recognition object and a sign object that is not a recognition object including approximate shapes, and excludes them in advance. To do. Therefore, the recognition information storage unit 201 stores a recognition target pattern and a non-recognition pattern as a sign body pattern.

  The recognition target patterns shown in FIGS. 3A to 3C are sign body patterns corresponding to the restriction sign “prohibition of progressing outside the designated direction” whose basic color is “blue”. Also, the recognition target patterns shown in FIGS. 3D to 3G are sign body patterns corresponding to “one-way” restriction signs whose basic color is “blue”. As shown in the figure, the restriction signs such as “prohibition of proceeding outside the designated direction” and “one-way” are basically a figure indicating the direction in which the vehicle can travel by an arrow. In the drawing, a blue region is indicated by a left oblique line.

  Note that the recognition information storage unit 201 considers a three-dimensional depth in the shape pattern matching with the image data, and as shown in FIGS. A plurality of corresponding sign body patterns are prepared. Thereby, even if the sign object in image data inclines by three-dimensional depth, the sign object recognition unit 200 can recognize a sign object exactly by shape pattern matching.

  The non-recognition pattern shown in FIG. 4A is a sign body pattern corresponding to a restriction sign of “warning sound” whose basic color is “blue”. The non-recognition pattern shown in FIG. 4B is a sign body pattern corresponding to the “parking possible” indication sign whose basic color is “blue”.

  In the present embodiment, it is assumed that the restriction signs “prohibition of proceeding outside specified direction” and “one-way” are recognized by shape pattern matching between the image data and the recognition target pattern shown in FIG. Further, in this embodiment, in order to reduce misrecognition of a sign having a similar color pattern, the sign shown in FIG. 4 in which the base color is “blue” and a white mark is drawn. The pattern is a non-recognized pattern.

  However, the present invention is not limited to this, and signs and signs corresponding to the recognition target pattern and the non-recognition target pattern can be arbitrarily set. For example, when recognizing a restriction sign such as “Prohibit traveling out of specified direction” or “One-way”, the base color is “blue” and a “direction guide sign” or “specified direction” with a white arrow drawn on it. The “non-recognized pattern” may be set as “another traffic classification (lane guidance)”.

  Next, recognition rate data indicating the recognition rate of each sign object pattern will be described with reference to FIG. FIG. 5 is a diagram schematically illustrating the data structure of the recognition rate data. As shown in the figure, the recognition rate data is composed of a sign body pattern, a name, and a recognition rate. The sign body pattern indicates a representative shape of each sign body pattern. For example, there are a plurality of types of restriction signs “prohibition of proceeding outside the designated direction”, but in FIG. 5, a sign body pattern corresponding to a sign whose arrow is in the left direction is shown as a representative. The name indicates the name of the sign object corresponding to each sign object pattern. The recognition rate is data verified in advance, and is an average accuracy rate when the sign object is recognized by shape pattern matching based on each sign object pattern. Usually, in the recognition process by shape pattern matching, the recognition rate increases as the sign body pattern has a strong feature.

  The recognition information storage unit 201 functions as image data storage means, sign object pattern storage means, and recognition rate data storage means in the present invention. Moreover, the information storage part 201 for recognition can be comprised in the database 18 shown in FIG.

  The matching area setting unit 202 sets the area for executing the pattern pattern matching with the sign object pattern stored in the recognition information storage unit 201 as the matching area. First, the matching area setting unit 202 sets the entire area of one piece of image data as a matching area. After that, the matching area setting unit 202 sets the color area extracted by the color area extraction unit 203 as a matching area, or deletes the area of the sign object recognized by the sign object recognition part 205, as will be described in detail later. The remaining area is set as a matching area. The matching area setting unit 202 functions as matching area setting means in the present invention.

  The color area extraction unit 203 extracts a color area having a color that is a basis of a specific sign body pattern from the matching area set by the matching area setting unit 202. Here, the specific sign object pattern is a sign object pattern corresponding to the sign object to be recognized, and the base color constituting the sign object pattern is the color constituting the base constituting the corresponding sign object. is there. For example, when recognizing a restriction sign whose base color is “blue”, the color region extraction unit 203 extracts a region containing a lot of blue components from the matching region as a blue region. When recognizing a warning sign whose basic color is “yellow”, the color region extraction unit 203 extracts a region containing a lot of yellow components from the matching region as a yellow region. The color area extraction unit 203 functions as color area extraction means in the present invention.

  The matching area setting unit 202 sets the color area extracted by the color area extracting unit 203 as a new matching area. In this way, by classifying the sign object pattern and the corresponding sign object into a plurality of groups based on the base color, it becomes possible to narrow down the matching area by color, and the processing for recognizing the sign object can be speeded up. .

  Based on the recognition rate data stored in the recognition information storage unit 201, the matching unit 204 performs shape pattern matching with the matching region set by the matching region setting unit 202 in order from the sign body pattern having the highest recognition rate. Execute. Specifically, the matching unit 204 performs shape pattern matching of the recognition target pattern or the non-recognition pattern on the color area extracted from one piece of image data in descending order of recognition rate. In this way, after the sign body pattern and the corresponding sign body are grouped based on the recognition rate verified in advance, the matching unit 204 executes shape pattern matching for each group having a high recognition rate. The matching unit 204 functions as a matching unit in the present invention.

  Based on the result of shape pattern matching by the matching unit 204, the sign object recognition unit 205 determines the sign object existing in the matching area when the matching degree between the matching area and the sign object pattern is equal to or higher than a predetermined level. , It is recognized as a sign corresponding to the sign pattern. That is, the sign object recognition unit 205 recognizes that the sign object in the matching area is a sign object corresponding to the recognition target pattern or a sign object corresponding to the non-recognition pattern. The sign object recognition unit 205 functions as sign object recognition means in the present invention.

  Since specific processing of shape pattern matching between the image data and the sign object pattern is a well-known technique, the description thereof is omitted for the sake of convenience.

  The recognized area deletion unit 206 deletes the area of the sign object recognized by the sign object recognition unit 205 from the matching area as the recognized area. Since the sign object recognition unit 205 recognizes in order from the sign object with the highest recognition rate, the recognized area deletion unit 206 deletes in order from the area of the sign object with the high recognition rate. In other words, the recognized area deletion unit 206 sequentially deletes the areas of the sign object with few erroneous recognitions. The recognized area deletion unit 206 functions as an area deletion unit in the present invention.

  The matching area setting unit 202 sets the remaining area obtained by deleting the recognized area from the matching area by the recognized area deleting unit 206 as a new matching area. Specifically, the matching unit 204 subsequently performs shape pattern matching with a new matching area, that is, a remaining area obtained by deleting a recognized area from a color area, in order from a sign body pattern with a high recognition rate. That is, the recognized area of the matching area corresponding to one piece of image data is deleted every time the sign object recognition unit 205 recognizes the sign object. In addition, the matching unit 204 executes shape pattern matching between a matching region that becomes narrower and a sign body pattern. By performing shape pattern matching for each group with a high recognition rate in this way and deleting the recognized sign object area from the matching area, it avoids duplicate recognition and speeds up and recognizes processing by reducing the matching area. An improvement in rate can be realized.

  The recognition result information storage unit 206 stores, as recognition result information, information related to a sign object that is a recognition target among the sign objects recognized by the sign object recognition unit 205, that is, a sign object corresponding to the recognition object pattern. FIG. 6 is a diagram schematically illustrating the data structure of the recognition result information. The recognition result information includes an image data ID, a shooting location, a direction, a position, and a sign ID. The image data ID is information for identifying image data that is a basis of shape pattern matching by the matching unit 204. The shooting location is a latitude and longitude indicating the shooting location of the corresponding image data. The direction is the traveling direction of the vehicle that has shot at the corresponding shooting location. The position is an X coordinate and a Y coordinate indicating the position of the sign object recognized by the sign object recognition unit 205 on the image data. The sign object ID is information for identifying the sign object recognized by the sign object recognition unit 205. Based on the recognition result information stored in the recognition result storage unit 206, for example, map data stored in the vehicle navigation device can be updated. In this embodiment, the recognition result information storage unit 206 stores recognition result information having a data structure as shown in FIG. 6, but the present invention is not limited to this, and is based on shape pattern matching. Information such as the degree of matching and the inclination of the sign body may be included in the recognition result information.

  In the present embodiment, the sign object recognition unit 200 is configured by the constituent elements of the sign object recognition apparatus 100, but the present invention is not limited to this. It is good also as being comprised by the component of the car navigation apparatus mounted in this vehicle.

  Note that the sign object recognition unit 200 in this embodiment recognizes sign objects corresponding to the sign object patterns shown in FIGS. 3 and 4, and shape pattern matching in this case is a well-known technique. For convenience, the description is omitted. However, the present invention is not limited to this, and it is also conceivable to recognize a direction guide sign indicating the direction, direction, distance, and the like. The matching in this case will be particularly described with reference to FIG.

  The direction guide sign is not a standard sign such as a regulation sign or a warning sign because information unique to the installation location such as a place name and distance is described. Therefore, storing in advance the sign object patterns corresponding to all the direction guide signs in the recognition information storage unit 201 is extremely heavy on the sign object recognizing unit 200. Therefore, when the sign recognition unit 205 recognizes the direction guide sign included in the matching area, the matching part 204 displays a small upward arrow while changing the scale, as shown in FIG. Match with the region. At this time, the matching unit 204 rotates the upward arrow 360 degrees, and executes matching with the area of the direction guide sign for all angles. Then, according to the matching result of the matching unit 204, when there is a sign body including a plurality of arrows, the sign object recognition unit 205 distinguishes it from a restriction sign such as “one-way” or “prohibition of traveling in the specified direction”. Calculate the percentage of the blue area in the whole body. Since the restriction sign usually has a lower proportion of the blue area than the direction guide sign, the sign object recognition unit 205 selects the sign in the matching area when the calculated ratio of the blue area is larger than a preset threshold value. Recognize it as a direction sign.

[Sign recognition method]
Next, a method of recognizing a sign object by shape pattern matching between image data and a sign object pattern by the sign object recognizing unit 200 will be described with reference to FIGS. FIG. 8 is a diagram showing matching regions in the present embodiment. FIG. 9 is a diagram for explaining the sign recognition method in the present embodiment. FIG. 10 is a diagram for explaining a conventional sign recognition method.

  In the sign object recognition unit 200, first, the matching area setting unit 202 sets the entire area of one piece of image data stored in the recognition information storage unit 201 as a matching area. Subsequently, the color area extraction unit 203 extracts a color area having a color as a basis constituting a specific sign body pattern from the matching area set by the matching area setting unit 202. Specifically, the color area extraction unit 203 extracts a blue area containing a lot of blue components as shown in FIG. 8B from the image data shown in FIG. At this time, the color area extraction unit 203 slightly expands the area containing a large amount of blue components (for example, several pixels) and extracts the area including the surrounding area as a blue area. According to this, even if the base color is “blue” and the edge is “white”, the color area extraction unit 203 can extract the sign including the edge as a blue area. .

  The matching area setting unit 202 sets the color area extracted by the color area extraction unit 203 as a new matching area. In other words, portions other than the blue region extracted by the color region extraction unit 203 are not subjected to shape pattern matching. Therefore, in the drawing, a portion that is not a target of shape pattern matching in the image data, that is, a portion other than the matching region is expressed in black for convenience.

  Subsequently, based on the recognition rate data stored in the recognition information storage unit 201, the matching unit 204 starts with the new matching region set by the matching region setting unit 202 in order from the sign body pattern having the highest recognition rate. Perform shape pattern matching. That is, the matching unit 204 first performs shape pattern matching between the sign object pattern having the highest recognition rate and the color area. Specifically, based on the recognition rate data as shown in FIG. 5, the matching unit 204 recognizes a non-recognition pattern corresponding to the restriction mark of “high horn sound” with a recognition rate of 100%, and FIG. The shape pattern matching with the blue region shown in FIG.

  Based on the result of the shape pattern matching by the matching unit 204, the sign object recognition unit 205 selects the sign object existing in the matching area when the matching degree between the matching area and the sign object pattern is equal to or higher than a predetermined level. , It is recognized as a sign corresponding to the sign pattern. Specifically, the sign object recognition unit 205 determines that the degree of matching between the blue region illustrated in FIG. 8B and the non-recognition pattern corresponding to the restriction mark “warning” is equal to or higher than a predetermined level, Recognize the sign in the blue area as a “warning” regulation sign. When the sign object is recognized, the recognized area deletion unit 206 deletes the area of the sign object recognized by the sign object recognition unit 205 from the matching area as the recognized area. Specifically, as shown in FIG. 8C, the recognized area deleting unit 206 deletes the area of the restriction sign of “screaming horn” recognized by the sign object recognition unit 205 from the blue area. The matching area setting unit 202 sets the remaining area obtained by deleting the recognized area from the matching area by the recognized area deleting unit 206, that is, the area shown in FIG. 8C as a new matching area.

  Based on the recognition rate data stored in the recognition information storage unit 201, the matching unit 204 performs shape pattern matching with a new matching region in order from the sign body pattern having the highest recognition rate. Specifically, based on the recognition rate data as shown in FIG. 5, the matching unit 204 has already recognized the recognition target pattern corresponding to the restriction sign “prohibition of traveling in the specified direction” with a recognition rate of 97.10%, and “ The shape pattern matching is performed with the blue region shown in FIG. Based on the result of the shape pattern matching by the matching unit 204, the sign object recognition unit 205 recognizes the blue area shown in FIG. 8C and the recognition mark corresponding to the restriction sign “prohibition of progression outside specified direction (right direction)”. Is determined to be equal to or higher than a predetermined level, and a sign existing in the blue region is recognized as “prohibition of proceeding outside specified direction (right direction)”. When a sign object that is a recognition target is recognized, the recognition result information storage unit 207 stores information on the sign object as recognition result information as shown in FIG. Further, as shown in FIG. 8D, the recognized area deletion unit 206 uses the restriction marker area “prohibition of progression outside specified direction (right direction)” recognized by the sign object recognition unit 205 as a recognized area. Remove from area. The matching area setting unit 202 sets the remaining area obtained by deleting the recognized area from the matching area by the recognized area deleting unit 206, that is, the area shown in FIG. 8D as a new matching area.

  The sign object recognition unit 200 recognizes the sign object by repeating such processing until there is no matching area for one piece of image data. In other words, as shown in FIG. 9A, the sign object recognition unit 200 collectively performs shape pattern matching between one piece of image data and a plurality of sign object patterns prepared in advance, thereby indicating the sign object. The recognition process is executed. At this time, the sign object recognition unit 200 stores the recognition rate of each sign object pattern in advance as recognition rate data, and executes shape pattern matching in order from the sign object pattern having the highest recognition rate based on the recognition rate data. .

  On the other hand, in the conventional recognition process, as shown in FIG. 10, shape pattern matching between one piece of image data and a plurality of prepared object patterns is executed a plurality of times for each type of the object pattern. The processing time was very long. Furthermore, since shape pattern matching for one piece of image data is executed by a plurality of processes for each type of sign body pattern, one sign object included in one piece of image data is erroneously recognized as a plurality of sign objects. Had occurred.

  As in this embodiment, the sign object recognition unit 200 performs batch processing of shape pattern matching between image data and a plurality of sign object patterns prepared in advance based on the recognition rate of the sign object pattern. The speed of time can be increased and the recognition rate can be improved.

  Furthermore, in this embodiment, when the sign object recognition unit 200 recognizes the sign object by shape pattern matching, the sign object recognition unit 200 deletes the recognized sign object area from the matching area as a recognized area. Since the recognition information storage unit 201 of the sign object recognition unit 200 stores the recognition target pattern and the non-recognition pattern as the sign object pattern, according to this, it corresponds to not only the recognition target pattern but also the non-recognition target pattern. Even when a sign object to be recognized is recognized, the area can be deleted from the matching area as a recognized area. Therefore, the matching area can be narrowed down accurately. Further, by deleting the area of the already recognized sign object as the recognized area in this way, one sign object recognized in one piece of image data is redundantly recognized as another sign object. Can be prevented. Therefore, the recognition rate of the sign object can be improved as compared with the conventional recognition process. In particular, the restriction sign “one-way”, which is a simple shape, is likely to be misrecognized, but the “prohibition of traveling in the specified direction” or “direction guidance sign” including approximate shapes in advance is used as a recognition target pattern or non-recognition pattern. By recognizing the corresponding sign and deleting the recognized area from the matching area, the recognition rate of the “one-way” sign can be effectively improved.

  Further, in this embodiment, the sign object recognition unit 200 extracts, as a color area, an area having a color as a basis constituting the recognition target pattern from the matching area, as shown in FIG. The sign recognition process is executed by collectively performing shape pattern matching between the area and a plurality of sign object patterns prepared in advance. By extracting color areas as pre-processing for shape pattern matching in this way, the matching area for executing shape pattern matching is narrowed down in advance, so the processing time can be significantly reduced compared to conventional recognition processing. It becomes.

[Sign recognition processing]
Next, the sign recognition process according to the present embodiment will be described with reference to FIG. FIG. 11 is a flowchart of the sign recognition process according to this embodiment. The sign object recognition unit 200 is configured by the components of the sign object recognition apparatus 100, and the sign object recognition processing is realized by the CPU 12 shown in FIG. 1 executing a program stored in the memory 13. The

  First, the matching area setting unit 202 of the sign object recognition unit 200 extracts one piece of image data from the image data stored in the recognition information storage unit 201, and uses the entire area of the image data as the matching area. Set. Subsequently, the color area extraction unit 203 extracts a color area having a color as a basis constituting a specific sign body pattern from the matching area set by the matching area setting unit 202 (step S1). Here, the specific sign object pattern is a sign object pattern corresponding to the sign object to be recognized from the image data. Therefore, for example, when it is desired to recognize a restriction sign whose base color is “blue”, the color region extraction unit 203 extracts a blue region from the matching region, that is, the entire region of the image data. The matching area setting unit 202 sets the color area extracted by the color area extraction unit 202 as a new matching area.

  Subsequently, based on the recognition rate data stored in the recognition information storage unit 201, the matching unit 204 executes shape pattern matching with the matching region in order from the sign body pattern having the highest recognition rate (step S2). . As a result of the shape pattern matching, when the matching degree between the sign body pattern and the matching area is not equal to or higher than a predetermined level (step S3; No), the matching unit 204 determines that the sign body pattern differs from the matching area based on the recognition rate data. And shape pattern matching is executed (step S2). On the other hand, as a result of the shape pattern matching, when the degree of matching between the sign object pattern and the matching area is equal to or higher than a predetermined level (step S3; Yes), the sign object recognition unit 205 displays the sign object existing in the matching area, It is recognized as a sign corresponding to the sign object pattern (step S4). When the sign object is recognized by the sign object recognition unit 205, the recognized area deletion unit 206 deletes the area of the sign object recognized by the sign object recognition unit 205 from the matching area as a recognized area (step S5). The matching area setting unit 202 sets the remaining area obtained by deleting the recognized area from the matching area by the recognized area deleting unit 206 as a new matching area.

  Also, the sign object recognition unit 205 determines whether or not the recognized sign object is a recognition target (step S6). When the sign object recognized by the sign object recognition unit 205 is a recognition target (step S6; Yes), the recognition result information storage unit 207 stores information on the sign object as recognition result information (step S7). Then, the sign body recognition unit 200 determines whether or not there is an unrecognized area in the matching area (step S8). The unrecognized area is an area where there is a sign body that has not yet been recognized in the matching area. When it is determined that there is an unrecognized area in the matching area (step S8; Yes), the matching unit 204 executes shape pattern matching between the sign body pattern and the matching area based on the recognition rate data (step S2). ). On the other hand, when it is determined that there is no unrecognized area in the matching area (step S8; No), the sign object recognition unit 200 completes the sign object recognition process.

  Until there is no unrecognized area in this way, the sign object recognition unit 200 recognizes a shape pattern matching between the matching area, that is, the remaining remaining area where the recognized area is deleted from the color area, and the sign object pattern with a high recognition rate. Run in one process in order. In other words, each time the sign object corresponding to the recognition target pattern or the non-recognition pattern is recognized, the sign object recognition unit 200 deletes the recognized area from the matching area, and the matching area and the sign object pattern become narrower and narrower. The sign object is recognized by executing shape pattern matching with.

  According to this, when not only the recognition target pattern but also the sign corresponding to the non-recognition target pattern is recognized, the area can be deleted from the matching area in the descending order of recognition rate as a recognized area. Also, by deleting the already recognized area of the sign object as a recognized area, it is possible to prevent one sign object recognized in one piece of image data from being recognized as another sign object. it can. Furthermore, by extracting a color area as a pre-processing for collectively performing shape pattern matching between one image data and a plurality of sign body patterns, the matching area is narrowed down in advance, so that the processing time can be significantly reduced. it can. Therefore, the sign object recognition unit 200 can efficiently recognize the sign object in a short processing time and improve the recognition rate by shape pattern matching based on the image data and the sign object pattern.

[Modification]
In the present embodiment, a color area of a color that forms the basis of a sign object to be recognized is extracted by the color area extracting unit 203 and set as a matching area, and the matching area and the sign object pattern are shape pattern matched. By doing so, the sign body is recognized. However, the present invention is not limited to this, and a color area that is the basis of a sign object that is not a recognition target is extracted by the color area extraction unit 203, and the remaining color area is deleted from the image data. An area may be set as a matching area, and the sign body may be recognized by shape pattern matching between the matching area and the sign body pattern. Also, since there are multiple types of colors that form the basis of the sign object, the color of the base color that constitutes the sign object that is not the recognition target is selected from the color area of the color that forms the sign object that is the recognition target. The remaining area from which the color area has been deleted may be set as a matching area, and the sign body may be recognized by shape pattern matching between the matching area and the sign body pattern. That is, how to use the color area extracted by the color area extraction unit 203 to narrow down the matching area is arbitrary.

  Here, a method for narrowing down the matching area using two types of color areas will be described.

  In this modification, the sign object to be recognized is a restriction sign of “prohibition of traveling in the specified direction” and “one-way”. The color that forms the basis of the restriction signs for “Prohibition of Out-of-Direction” and “One-way” is “Blue”, and is indicated by a left oblique line in the drawing. On the other hand, the sign body which is not the recognition target is assumed to be the restriction signs “Vehicle entry prohibition” and “Parking and stopping prohibited” shown in FIGS. The color that is the basis of the restriction sign of “Vehicle entry prohibition” is “red”, and the color that is the basis of the restriction sign of “parking and parking prohibition” is “red” and “blue”. In the drawing, a red region is indicated by a right oblique line.

  In the sign object recognition unit 200, first, the matching area setting unit 202 sets the entire area of one piece of image data stored in the recognition information storage unit 201 as a matching area. Subsequently, the color area extraction unit 203 extracts a color area having a color as a basis constituting a specific sign body pattern from the matching area set by the matching area setting unit 202. Specifically, this will be described with reference to FIG. FIG. 13 is a diagram illustrating a matching region in the modification. The color area extraction unit 203 extracts a blue area containing a large amount of blue components as shown in FIG. 13B from the image data shown in FIG. The matching area setting unit 202 sets the blue area extracted by the color area extraction unit 203 as a new matching area. As a result, the restriction sign “parking and parking prohibition” whose basic color is “red” exists in the image data but is not extracted as a blue region, and is therefore excluded from the matching region. However, the restriction sign “parking and parking prohibition”, which is the base color of “red” and “blue”, is extracted as a blue region including a blue component, and is included in the matching region.

  Further, the color area extraction unit 203 extracts a color area having a color as a basis constituting a specific sign body pattern from the matching area newly set by the matching area setting unit 202. Specifically, the color area extraction unit 203 extracts a red area containing a lot of red components from the blue area shown in FIG. As the red area, the area of the restriction sign “parking and parking prohibited” which is the base color of “red” and “blue” is extracted. The matching area setting unit 202 sets a remaining area obtained by deleting the red area extracted by the color area extraction unit 203 from the matching area as a new matching area. As a result, the matching area becomes a blue area from which the area of the restriction sign “parking and parking prohibited” is deleted as shown in FIG.

  After this, the sign object recognition process executed by the matching part 204, the sign object recognition part 205, the recognized area deletion part 206, and the recognition result information storage part 207 is substantially the same as that in the above-described embodiment. Omitted.

  Note that the matching area setting unit 202 in the modification functions as a color area deleting unit in the present invention.

  As described above, the sign object recognition unit 200 can effectively narrow down the matching area by using the plurality of color areas extracted by the color area extraction unit 203. As a result, it is possible to increase the speed of the sign object recognition process and improve the recognition rate.

It is a block diagram of a sign recognition apparatus. It is a block diagram which shows the function structure of a sign body recognition unit. It is an example of a recognition object pattern. It is an example of a non-recognition pattern. It is a figure which shows the data structure of recognition rate data. It is a figure which shows the data structure of recognition result information. It is a figure explaining the shape pattern matching of a direction guidance sign. It is a figure which shows the matching area | region in an Example. It is a figure explaining the sign recognition method in an example. It is a figure explaining the conventional sign body recognition method. It is a flowchart of a sign object recognition process. It is an example of the sign object which is not a recognition object in a modification. It is a figure which shows the matching area | region in a modification.

Explanation of symbols

11 System bus 12 CPU
DESCRIPTION OF SYMBOLS 13 Memory 14 Keyboard 15 Coordinate instruction | indication device 16 Display 17 Printer 18 Database 100 Marking object recognition apparatus

Claims (7)

Image data storage means for storing image data to be subjected to shape pattern matching;
A sign object pattern storage means for storing a sign object pattern to be subjected to shape pattern matching;
Recognition rate data storage means for storing the recognition rate of each sign body pattern by shape pattern matching as recognition rate data;
Matching area setting means for setting the entire area of one piece of the image data as a matching area;
A matching unit that performs shape pattern matching with the matching region in order from a sign body pattern with a high recognition rate based on the recognition rate data;
A sign object recognition means for recognizing a sign object corresponding to the sign object pattern from the matching area based on the result of the shape pattern matching;
An area of the sign body recognized by the sign object recognition means as a recognized area, and a recognized area deletion means for deleting from the matching area,
The said matching area | region setting means sets the residual area | region which deleted the said recognized area | region from the said matching area | region by the said recognized area deletion means as a new matching area | region, The sign object recognition apparatus characterized by the above-mentioned.   The sign object recognition device according to claim 1, wherein the sign object pattern storage unit stores a sign object pattern of a recognition target shape and a sign object pattern of a non-recognition shape. A color area extracting means for extracting a color area having a color that forms the basis of the sign body pattern from the matching area;
The sign object recognition apparatus according to claim 1, wherein the matching area setting unit sets the color area extracted by the color area extraction unit as a new matching area. There are multiple types of colors that form the basis of the sign object pattern,
A color area deleting means for deleting the color area extracted by the color area extracting means from the matching area;
4. The sign object recognition apparatus according to claim 3, wherein the matching area setting unit sets a remaining area obtained by deleting the color area from the matching area by the color area deleting unit as a new matching area. An image data storage step for storing image data to be subjected to shape pattern matching;
A sign object pattern storage step for storing a sign object pattern to be subjected to shape pattern matching;
A recognition rate data storage step of storing the recognition rate of each sign object pattern by shape pattern matching as recognition rate data;
A matching area setting step of setting the entire area of one piece of the image data as a matching area;
A matching step for performing shape pattern matching with the matching region in order from a sign body pattern having a high recognition rate based on the recognition rate data,
Based on the result of the shape pattern matching, a sign object recognition step for recognizing a sign object corresponding to the sign object pattern from the matching area;
A recognized area deletion step of deleting the area of the sign body recognized by the sign object recognition means as a recognized area from the matching area, and
The said matching area | region setting process sets the residual area | region which deleted the said recognized area | region from the said matching area | region by the said recognized area deletion process as a new matching area | region, The sign object recognition method characterized by the above-mentioned. A sign recognition program executed on a computer,
Image data storage means for storing image data to be subjected to shape pattern matching;
A sign object pattern storage means for storing a sign object pattern to be subjected to shape pattern matching;
Recognition rate data storage means for storing the recognition rate of each sign object pattern by shape pattern matching as recognition rate data;
Matching area setting means for setting the entire area of one image data as a matching area;
Matching means for executing shape pattern matching with the matching region in order from a sign body pattern having a high recognition rate based on the recognition rate data,
A sign object recognition means for recognizing a sign object corresponding to the sign object pattern from the matching area based on the result of the shape pattern matching;
The area of the sign object recognized by the sign object recognition means is the recognized area, and the computer functions as a recognized area deletion means for deleting from the matching area,
The said matching area setting means sets the residual area | region which deleted the said recognized area from the said matching area by the said recognized area deletion means to a new matching area, The sign object recognition program characterized by the above-mentioned.   A recording medium on which the sign recognition program according to claim 6 is recorded.

Images