WO2013154062A1

WO2013154062A1 - Image recognition system, image recognition method, and program

Info

Publication number: WO2013154062A1
Application number: PCT/JP2013/060564
Authority: WO
Inventors: 雅文矢野; 雄馬松田
Original assignee: 日本電気株式会社
Priority date: 2012-04-11
Filing date: 2013-04-01
Publication date: 2013-10-17
Also published as: JPWO2013154062A1

Abstract

The present invention performs generic object recognition on the basis of silhouette information containing multiple objects. The invention implements an image recognition method for performing recognition processing on an object or a shape, said method characterized in comprising: a contour extraction step of extracting a contour of an object or a shape on the basis of object image data as an object of detection; a feature point detection step of detecting a feature point group on the basis of the extracted contour; a partial contour generation step of using the detected feature point information to partition the contours between the feature points into one or more regions in order to generate the partitioned regions as partial contours; and a contour matching step of selecting one or more of the generated partial contours and matching same with contour information.

Description

Image identification system, image identification method, and program

The present invention relates to a system for recognizing or searching a general object or shape included in image information, and more particularly to a method for identifying a target using silhouette information of the object or shape.

In recent years, with the rapid spread of digital video equipment such as digital cameras, there is an increasing expectation for general object recognition that identifies what objects are included in captured images and videos. Yes.
In general object recognition, appropriate classification of image data stored without being classified in a database, retrieval of necessary image data, and the like are performed. Furthermore, general object recognition has the possibility of being applicable to various uses such as extraction of a desired scene from a moving image and re-editing by cutting out only the desired scene.
Various recognition technologies such as face recognition and fingerprint recognition have been developed so far as technologies related to object recognition, but these are often directed to specific applications. Such a recognition technology specialized for a specific application does not operate efficiently when it is used for another application. This problem includes problems such as a decrease in recognition rate, misidentification, and an increase in the amount of information processing. For this reason, development of a technique for efficiently recognizing a general object is expected.
As a method for recognizing a general object by an information processing apparatus, a method using a silhouette of an object shape has been proposed and widely used (Patent Documents 1 and 2). In these documents, a method for recognizing from silhouette information using curvature information of each point on the contour is proposed.
The curvature is defined as the reciprocal of the radius of the local arc (curvature radius). Therefore, the curvature itself is not an invariant with respect to the size of the object. However, by performing relativity, the curvature can be extracted invariably with respect to the size of the object.
In addition, since it is known that the curvature information of contours of similar objects are similar, the above technique using the curvature information can calculate the similarity even for contours with slightly different contour shapes. Is possible.
Further, from the silhouette information of the object shape, it is possible to extract not only the curvature information but also various feature quantities related to the object shape such as inflection points that are the characteristic points of curvature and the position coordinates of the contour (non- Patent Document 1).
While silhouette information can be used to extract various features of object outlines, the object identification process from silhouettes can be easily separated when multiple objects are adjacent and silhouettes are combined. There is a problem that it cannot be performed (Non-Patent Document 2). Non-Patent Document 2 points out a problem that when handwritten characters are adjacently combined, it is not possible to determine where the combined points are.
In order to deal with such problems, a coping method using a learned coupling pattern has been proposed when analyzing silhouette coupling. For example, after restricting the number of characters to be combined to two characters, learning is performed in advance for a combination pattern of two characters, and processing such as cutting the silhouette at the point where the learned combination pattern appears is performed. Thereafter, the target is identified using the divided silhouette. However, there are a number of problems when using such learned coupling patterns. For example, there is a problem that cannot be dealt with when the number of connected characters (the number of objects) is unknown. As another example, there is a point that cannot be dealt with in the case of a connection pattern that is not a previously learned connection pattern. Another problem is that the coupling pattern is substantially different depending on the degree of coupling.
As described above, various problems remain in the method using the learned connection pattern. For this reason, it is necessary to cut out the shape freely and efficiently from silhouette information without limiting the conditions used for recognition such as the number of characters (number of objects) and without learning the connection pattern in advance.
As a technique for extracting a shape from silhouette information without learning a connection pattern in advance, a characteristic point existing on the outline of the silhouette is used as a extraction candidate point, and the outline information between the extraction candidate points is complemented. A method of cutting out a shape has been proposed (Patent Document 3). By using this method, it is possible to cut out many shapes from silhouette information without learning the connection pattern in advance. However, according to this method, when there is no characteristic point on the outline of the silhouette, the shape cannot be cut out.

JP 10-055447 A PCT / JP2011 / 075599 Japanese Patent Application No. 2011-268907

When using a general object recognition method using silhouette information, it is necessary to extract silhouette information corresponding to a single object from silhouette information in which a plurality of objects are combined. In order to extract silhouette information for each single object from silhouette information of a plurality of objects, conventionally, this is performed by extracting feature points to be cut out.
However, if there is no characteristic point between the silhouettes of two different objects, the characteristic point cannot be extracted, and it has been difficult to cut out with the conventional method.
That is, in the silhouette information of one or more object shapes, when there is no characteristic point between two different silhouettes, an appropriate method for solving the extraction of a single closed contour has been devised so far. It wasn't.
An object of the present invention is to provide an image identification system that performs general object recognition of an object or a shape included in silhouette information based on silhouette information, a method thereof, and a program.
Another object of the present invention is to provide a processing method for performing general object recognition from silhouette information and efficiently trying to extract a single closed contour.

A contour extraction step for extracting a contour of an object or shape from target image data to be detected, a feature point detection step for detecting a feature point group from the extracted contour, and a feature point using the detected feature point information A contour generation step for generating a partial contour as a partial contour, and selecting one or more of the generated partial contours and performing contour information matching processing And an image identifying method for identifying and processing an object or shape.

ADVANTAGE OF THE INVENTION According to this invention, the image identification system, its method, and program which perform general object recognition of the object or shape contained in this silhouette information based on silhouette information can be provided.
In addition, according to the present invention, it is possible to provide a processing method that performs general object recognition from silhouette information and efficiently attempts to extract a single closed contour.

FIG. 1 is a configuration diagram of an image identification system according to the first embodiment.
FIG. 2 is an explanatory diagram illustrating an example of an object shape to be subjected to pattern recognition.
FIG. 3 is an explanatory diagram showing an example of contour information in silhouette information.
FIG. 4 is an explanatory diagram showing an example of feature points on the contour.
FIG. 5 is an explanatory diagram showing an example of the selected partial contour.
FIG. 6 is an explanatory diagram illustrating an example of an object shape stored in the storage unit.
FIG. 7 is a flowchart showing an example of the operation of the image identification system in the first embodiment.
FIG. 8 is a configuration diagram of an image identification system according to the second embodiment.
FIG. 9 is an explanatory diagram illustrating an example of a method for reconstructing a contour.
FIG. 10 is a flowchart illustrating an example of the operation of the image identification system according to the second embodiment.
FIG. 11 is a configuration diagram of an image identification system according to the third embodiment.
FIG. 12 is an explanatory diagram illustrating an example of a method for correcting a contour between feature points.
FIG. 13 is an explanatory diagram illustrating an example of an object shape to be subjected to pattern recognition.
FIG. 14 is an explanatory diagram illustrating an example of the complemented contour.
FIG. 15 is a flowchart illustrating an example of the operation of the image identification system according to the third embodiment.

Conventionally, when using a general object recognition method using silhouette information, it is necessary to extract silhouette information corresponding to a single object from silhouette information in which a plurality of objects are combined. In order to extract silhouette information for each single object from silhouette information of a plurality of objects, conventionally, this is performed by extracting feature points to be cut out.
However, if there is no characteristic point between the silhouettes of two different objects, the characteristic point cannot be extracted, and it has been difficult to cut out with the conventional method.
That is, in the silhouette information of one or more object shapes, when there is no characteristic point between two different silhouettes, an appropriate method for solving the extraction of a single closed contour has been devised so far. It wasn't.
In response to these problems, the inventors propose a system for performing general object recognition as follows.
The processing operation of the system can be shown in four steps.
In the first step, the entire outline that can be acquired from the silhouette information of an object including one or more objects is extracted.
In the second step, characteristic points on the extracted contour are detected.
In the third step, the contour is divided into one or a plurality of parts based on the characteristic points extracted in the second step, and a partial contour is generated.
In the fourth step, an object is specified by collating the generated partial contour information and database information.
Hereinafter, an image identification system that performs each step will be described using an embodiment.
An embodiment of the present invention will be described with reference to FIGS.
[First Embodiment]
By adopting the present embodiment, it is possible to provide an information processing system that performs image identification processing that enables recognition of a single object included therein from silhouette information including a plurality of object shapes.
[Description of configuration]
FIG. 1 is a block diagram illustrating a configuration of a recognition system according to an embodiment. In addition, about the structure with little connection with this invention, description is simplified or abbreviate | omitted.
The image identification system includes a control unit 10 that manages the entire information processing, an image information acquisition unit 20 that acquires image data to be detected, an image information storage unit 30 that stores acquired image data, and the like. It includes a contour information storage unit 40 in which features of objects and shapes are stored and collected in a database, and a matching result output unit 50 that outputs a recognition result. In addition, the image recognition system includes a contour extraction unit 101, a feature point detection unit 102, a partial contour generation unit 103, and a partial contour matching unit 104 as the image identification unit 100 according to the present invention.
The control unit 10 manages the overall operation of information processing related to image identification.
In the image information acquisition unit 20, image data specified by the user such as a moving image or a photograph is taken into the system and stored in the image information storage unit 30. As the capturing method, the image information specified by the user may be acquired as it is, or may be acquired by performing conversion that facilitates subsequent processing, such as black and white conversion using luminance information or the like. An example of the image information acquired by the image information acquisition unit 20 is illustrated in FIG. It is also possible to automatically collect image information from a moving image or the like at an arbitrary interval or the like.
The image information storage unit 30 stores acquired image data, results obtained by collation, intermediate data (closed contour shape, complementary contour, feature point coordinates, etc.) and the like as necessary. The image information storage unit 30 may be configured by a memory, an HDD, or a storage unit.
In the contour information storage unit 40, data used for inquiry is extracted and stored. Note that an external database may be referred to as the contour information storage unit 40.
The collation result output unit 50 outputs the result obtained by the image identification unit 100. For example, together with the collation result acquired from the image information storage unit 30, the target object or shape of the collation result recorded in the contour information storage unit 40 is output to a monitor or the like. Note that any information may be output as the output form.
The contour extraction unit 101 extracts, as contour information (silhouette information), all or part of the contour that can be acquired from the image data to be detected that is acquired by the image information acquisition unit 20 and recorded in the image information storage unit 30. To do. Here, “all” means that the individual contours are all targeted for detection, and “only part” means that only one contour in the vicinity specified by the user or an arbitrary number of contours is targeted for detection. Point to.
The contour extraction unit 101 extracts, from the target image data, for example, points where the hue, saturation, brightness, and the like change rapidly using a Laplacian / Gaussian filter. The point of sudden change here may be determined by whether differential values such as hue, saturation and brightness exceed a predetermined threshold. Note that the method of extracting the contour information is not limited to the exemplified method.
The contour of the extracted object may be expressed as a collection of contour points, for example, and each point may be represented by (x, y) or the like using a Cartesian coordinate system.
An example of the contour information extracted by the contour extraction unit 101 is shown in FIG. As described above, when there is one silhouette information composed of the number patterns shown in FIG. 2 (the silhouettes of two objects joined by “2” and “3”), the outer contour as shown in FIG. 3 is extracted. The
A plurality of contours may be obtained from the target image data. What is necessary is just to perform the process demonstrated below with respect to this acquired each outline separately. In the following description, it is assumed that only one contour has been acquired.
The feature point detection unit 102 detects a characteristic point (a feature point group) on the contour to be detected. As this feature point, for example, a point at which the value of the curvature k (t) defined by the following formula (1) becomes zero, that is, an inflection point is used.
The curvature k (t) is the contour coordinate t taken so as to go around the contour starting from an arbitrary point on the contour, and x, x when the contour is expressed in an orthogonal coordinate system (x, y coordinate system) It is defined by a primary differential value and a secondary differential value with respect to t of y (see formula (1)).

The result of detecting the feature points defined by this method from the contour shown in FIG. 3 is indicated by a circle in FIG. Note that the method of detecting the feature points is not limited to the above. For example, a method may be considered in which the point at which the value of the curvature k (t) described above changes abruptly is defined as a “connection point” and the connection point is adopted as a feature point. The point of sudden change here is determined by, for example, whether or not the differential value of the curvature exceeds a predetermined threshold value. The definition of “connection point” is not limited to the above definition, and any definition may be used as long as it is a characteristic point at which an object connects.
The partial contour generation unit 103 divides the contour into one or a plurality of regions using the feature point information detected by the feature point detection unit 102, and generates each divided region as a partial contour. For the partial contour, any means using the feature point information may be used, but when two or more feature points are detected, it is desirable to generate a region having the feature points as both ends as the partial contour. When the contour extracted by the contour extraction unit 101 is an open contour, a region having both the feature points on the contour and the end points of the contour as both ends may be used as the partial contour.
The partial contour matching unit 104 selects one or a plurality of partial contours from the partial contours generated by the partial contour generation unit 103, and the selected one or a plurality of partial contours are recorded in the contour information storage unit 40. The object is identified by referring to the inquiry data. When selecting partial contours, it is desirable to select continuous partial contours. Many existing methods can be used for the matching process. For example, collation by comparison of feature amounts for each partial contour, collation using object positions using all selected partial contours, and the like can be mentioned. The partial contour selected in the example of FIG. 2 is indicated by a thick line in FIG. 5, and an example of the inquiry data recorded in the contour information storage unit 40 is shown in FIG. Similar processing is performed on the inquiry data shown in FIG. 6 to generate and select a partial contour, and to refer to the selected partial contour in FIG. The collation process may be any method as long as the outline information collation process is performed by using the selected complementary outline and the partial outline corresponding to the part obtained by dividing the previously extracted outline.
[Description of operation]
Next, an operation example of the embodiment will be described. FIG. 7 is a flowchart showing an example of the operation of the present embodiment.
First, the image information acquisition unit 20 acquires target image data designated by the user and records it in the image information storage unit 30 (S1001). The acquisition of image information is not limited to the one specified by the user, and the system may acquire it automatically or semi-automatically.
Next, the contour extraction unit 101 extracts the contour information (silhouette information) of the object or shape from the target image data to be detected recorded in the image information storage unit 30 (S1002). For example, the contour information can be extracted only when the user designates a reference in advance, such as a pixel having a luminance equal to or higher than a threshold, and satisfies the reference.
Next, the feature point detection unit 102 detects a characteristic point on the contour and detects it as a feature point group (S1003).
Next, the partial contour generation unit 103 divides the contour into one or a plurality of regions using the feature point information detected by the feature point detection unit 102, and generates each divided region as a partial contour ( S1004).
Next, the partial contour matching unit 104 selects one or more contours from the generated partial contours using the contour information (silhouette information) extracted from the target image data and the corrected complementary contour. (S1005) Then, by comparing the selected partial contour with the contour information stored in the contour information storage unit 40, an object matching process is performed (S1006). When a desired object cannot be obtained when compared with the contour information stored in the contour information storage unit 40, the partial contour is selected again (S1005), and the same processing is performed (S1007).
At this time, the control unit 10 may perform processing for changing the above processing and parameters as necessary. If processing is performed under different conditions, the accuracy can be improved. Thereafter, the control unit 10 outputs a collation result from the collation result output unit 50.
By operating the image identification system in this way, it can be expected that recognition of a single object included in the silhouette can be efficiently acquired from silhouette information in which the number of included objects is not clear.
Note that each unit of the image identification system may be realized by using a combination of hardware and software. In a form in which hardware and software are combined, an image identification program is developed in the RAM, and hardware such as a control unit (CPU) is operated based on the program, thereby realizing each unit as various means. Further, the program may be recorded on a storage medium and distributed. The program recorded on the recording medium is read into a memory via a wired, wireless, or recording medium itself, and operates a control unit or the like. Examples of the recording medium include an optical disk, a magnetic disk, a semiconductor memory device, and a hard disk.
To describe the above-described embodiment in another expression, an information processing apparatus that operates as an image identification system is based on an image identification program developed in a RAM, contour extraction means, feature point detection means, feature point pair generation means, It can be realized by operating the control unit as a complementary contour selection unit and a contour collation unit.
As described above, according to the present invention, it is possible to provide an image identification system, a method, and a program for performing general object recognition of an object or a shape included in silhouette information based on silhouette information.
In addition, according to the present invention, it is possible to provide a processing method that performs general object recognition from silhouette information and efficiently attempts to extract a single closed contour.
In addition, the specific configuration of the present invention is not limited to the above-described embodiment, and changes within a range not departing from the gist of the present invention are included in the present invention.
[Second Embodiment]
A second embodiment for carrying out the present invention will be described in detail with reference to the drawings. The present embodiment assumes a case where information included in the partial contour information in the first embodiment is not sufficient. In such a case, the second embodiment employs a technique for reconstructing the partial contour with reference to the contour information stored in the contour information storage unit. For this reason, the second embodiment is different from the first embodiment in that a contour information reconstruction unit 111 is newly provided. In addition, some elements have functions different from those of the first embodiment.
[Description of configuration]
FIG. 8 is a block diagram illustrating a configuration of a recognition system according to an embodiment. In addition, about the structure with little connection with this invention, description is simplified or abbreviate | omitted.
The contour information reconstruction unit 111 reconstructs a partial contour by a method of cutting or combining necessary partial contours with respect to the partial contour generated by the partial contour generation unit 103. Any method may be used as the reconstruction method at this time, but it is preferable to refer to the already stored contour information of the object recorded in the contour information storage unit 40. Hereinafter, an example of a technique adopted by the contour information reconstruction unit 111 using the silhouette information of FIG. 2 will be described using FIG. 9.
First, regarding the silhouette information shown in the lower part of (a) in FIG. 9, a partial outline generated and selected by the feature point detection unit 102, the partial outline generation unit 103, and the partial outline matching unit 104 is shown in (b). . At the same time, (b) the partial outline corresponding to the lower partial outline is selected as shown in (b) upper part of the partial outline of the shape shown in (a) upper part included in the outline information storage unit 40. At this time, for the shape in the memory shown in (b) above, the partial contour not selected is determined as shown in (c). The partial contour in the lower shape corresponding to this may be composed of two partial contours in the lower portion as the partial contour adjacent to the partial contour selected in (b). This reconstruction can be performed by using geometric information such as the length and curvature of the partial contour shown in (c). Here, (d) shows the result of reconstruction using the length information of the partial contour shown in (c). The contour information is reconstructed by the series of processes described above.
[Description of operation]
Next, an operation example of the embodiment will be described. FIG. 10 is a flowchart showing an example of the operation of the present embodiment. The description of operations that are not related to the present invention will be omitted.
After S1004, the contour information reconstruction unit 111 cuts or joins the necessary partial contours to the partial contours generated by the partial contour generation unit 103 (S1101).
The partial contour matching unit 104 selects one or a plurality of contours from the generated partial contours using the contour information (silhouette information) extracted from the target image data and the corrected complementary contour (S1005). Then, the object matching process is performed by comparing the selected partial contour with the contour information stored in the contour information storage unit 40 (S1006). When the desired information cannot be obtained when comparing the contour information storage unit 40, the contour information is reconstructed (S1101) and the partial contour is selected (S1005), and the same processing is performed (S1007).
At this time, the control unit 10 may perform processing for changing the above processing and parameters as necessary. If processing is performed under different conditions, the accuracy can be improved. Thereafter, the control unit 10 outputs a collation result from the collation result output unit 50.
As described above, by implementing this embodiment, it is possible to collate an object even when the information included in the partial contour information is not sufficient.
In addition, the specific configuration of the present invention is not limited to the above-described embodiment, and changes within a range not departing from the gist of the present invention are included in the present invention.
[Third Embodiment]
A third embodiment for carrying out the present invention will be described in detail with reference to the drawings. In the present embodiment, it is assumed that the information included in the contour information in the images in the first and second embodiments is not sufficient. In such a case, the third embodiment employs a technique of referring to the contour information stored in the contour information storage unit and complementing the contour. For this reason, the third embodiment differs from the second embodiment in that a contour complementing part 121 is newly provided. In addition, some elements have functions different from those of the second embodiment.
[Description of configuration]
FIG. 11 is a block diagram illustrating a configuration of a recognition system according to an embodiment. In addition, about the structure with little connection with this invention, description is simplified or abbreviate | omitted.
The contour complementing unit 121 supplements necessary contour information with respect to the contour extracted by the contour extracting unit 101. Any method may be used as the complementing method at this time, but it is desirable to perform complementation using the feature points detected by the feature point detecting unit 102 as described below. Hereinafter, a complementing method using the feature points detected by the feature point detection unit 102 will be described.
Among the feature points detected by the feature point detection unit 102, two points in an appropriate combination are complemented. The two points may be two arbitrary points, but it is desirable to select an appropriate one in consideration of a geometric condition or the like. For example, the condition that the angle between the two tangents on the two points is less than the threshold, or the curvature value becomes extremely discontinuous after complementation (for example, it becomes discontinuous by switching between positive and negative). It is desirable to select on the condition that there is not. In addition, any correction method may be adopted for the correction here, and an approximation by a linear curve (straight line) adopted in the first small step may be adopted. A method of using an angle formed by a tangent line will be described with reference to FIG. First, the angle formed by the tangent line between the point 10001 and the point 10002 is measured as 10003. Next, a line segment length 10004 having both ends of the point 10001 and the point 10002 is derived as a distance between two points of the point 10001 and the point 10002. Further, assuming that a small area having both ends of the

points

10001 and 10002 is an arc, the representative point of the circular arc is 10005, the distance 10006 between the point 10001 and the point 10005 is a radius when the small area is an arc. It matches (curvature radius). Also, by assuming a circular arc, the values of

angles

10003 and 10007 are equal. Using the geometric relationship of these variables, the radius of curvature 10006 can be derived from equation (2).

Thereby, it is possible to extract a contour as shown by a bold line in FIG. 14 from the image information as shown in FIG. Contour information can be collated by operating the partial contour collating unit 104 for the contour thus reconstructed.
[Description of operation]
Next, an operation example of the embodiment will be described. FIG. 15 is a flowchart showing an example of the operation of the present embodiment. The description of operations that are not related to the present invention will be omitted.
The contour complementing unit 121 supplements the contour extracted by the contour extracting unit 101 after S1003 (S1201).
Next, the partial contour generation unit 103 divides the contour into one or a plurality of regions using the feature point information detected by the feature point detection unit 102, and generates each divided region as a partial contour ( S1004).
After S1004, the contour information reconstruction unit 111 reconstructs the partial contour by cutting or combining the necessary partial contours with the partial contour generated by the partial contour generation unit 103 (S1101). .
The partial contour matching unit 104 selects one or a plurality of contours from the generated partial contours using the contour information (silhouette information) extracted from the target image data and the corrected complementary contour (S1005). Then, the object matching process is performed by comparing the selected partial contour with the contour information stored in the contour information storage unit 40 (S1006). If the desired object is not obtained when the contour information storage unit 40 is compared, the steps after S1201 are performed again, and the same processing is performed (S1007).
As described above, by implementing this embodiment, it is possible to collate an object even when the information included in the contour information is not sufficient.
In each of the above-described embodiments, the processing of the embodiment may be executed by a computer-readable storage medium encoded with a program, software, or an instruction that can be executed by a computer. . The storage medium includes not only a portable recording medium such as an optical disk, a floppy (registered trademark) disk, and a hard disk, but also a transmission medium that temporarily records and holds data such as a network.
In addition, the specific configuration of the present invention is not limited to the above-described embodiment, and changes within a range not departing from the gist of the present invention are included in the present invention.

According to the present invention, it is possible to recognize a general object by extracting one piece of information from silhouette information of one or a plurality of object shapes. Applicable for use.
This application claims its benefit on the basis of priority from Japanese Patent Application No. 2012-90225 filed on April 11, 2012, the disclosure of which is hereby incorporated by reference in its entirety. take in.

10 Control unit (control means)
20 Image information acquisition unit (image information acquisition means)
30 Image information storage unit (target image, result)
40 Contour information storage (inquiry data)
50 Verification result output unit (Verification result output means)
100 Image identification unit (image identification means)
101 Contour extraction unit (contour extraction means)
102 feature point detection unit (feature point detection means)
103 Partial contour generation unit (partial contour generation means)
104 Partial contour matching unit (partial contour matching means)
111 Outline information reconstruction unit (contour information reconstruction means)
121 Contour complement part (contour complement means)
10001: A point of a feature point pair 10002: A point of a feature point pair 10003: An angle formed by a tangent line of the feature point pair 10004: A line segment length between the feature point pairs 10005: An arc center point 10006 when an arc is formed between the feature point pairs : Line segment with both ends of arc center point and each point of feature point pair 10007: Angle formed by both ends of feature point pair with arc center point as center

Claims

A contour extraction step for extracting a contour of an object or shape from target image data to be detected;
A feature point detecting step for detecting a feature point group from the extracted contour;
Using the detected feature point information, the contour between the feature points is divided into one or more regions, and a partial contour generation step for generating the divided regions as partial contours;
An image identification method for identifying an object or shape, comprising: a contour matching step that selects one or more of the generated partial contours and performs a contour information matching process.
The image identification method includes:
2. The contour information reconstructing step for newly reconstructing a partial contour from other than the selected partial contour is further included, and the contour matching step is performed using the reconstructed partial contour. An image identification method for identifying and processing the object or shape described in 1.
The image identification method includes:
3. The method further comprises a contour complementing step of complementing a contour between any two points among the detected feature points, and performing the partial contour generating step using the complemented contour. An image identification method for identifying and processing the object or shape described in 1.
4. The feature point according to claim 1, wherein the feature point is a point at which the azimuth or curvature of the tangent line on the contour changes rapidly or is included in a predetermined value range. 5. An image identification method for performing identification processing on an object or shape according to claim 1.
The reconstructing of the partial contour is performed by referring to the contour to be collated and using any or all of the length, orientation, and curvature of the contour. An image identification method for identifying an object or shape according to any one of the above.
When the contour is complemented, correction is performed only when one or a plurality of pieces of information on the positions, tangent directions, and curvatures of the plurality of feature points are included within a predetermined value range. An image identification method for identifying an object or shape according to any one of claims 1 to 5.
Contour extraction means for extracting a contour of an object or shape from target image data to be detected;
Feature point detection means for detecting a feature point group from the extracted contour;
Using the detected feature point information, the contour between the feature points is divided into one or more regions, and a partial contour generating means for generating the divided regions as partial contours;
An image identification system for identifying an object or a shape, comprising: an outline collating unit that selects one or a plurality of generated partial outlines and performs an outline information matching process.
The image identification system includes:
The contour information reconstructing means for newly reconstructing a partial contour from other than the selected partial contour is further provided, and the contour matching means is implemented using the reconstructed partial contour. An image identification system for identifying and processing the object or shape described in 1.
The image identification system includes:
9. The apparatus according to claim 7, further comprising contour complementing means for complementing a contour between arbitrary two points among the detected feature points, wherein the partial contour generating means is implemented using the complemented contour. An image identification system for identifying and processing the object or shape described in 1.
10. The feature point according to claim 7, wherein the feature point is a point at which the azimuth or curvature of the tangent line on the contour changes rapidly or a point included in a predetermined value range. An image identification system for identifying and processing the object or shape described in the above.
11. The reconstructing of the partial contour is performed by referring to the contour to be collated and using any or all of the length, direction, and curvature of the contour. An image identification system for performing identification processing on any object or shape.
When the contour is complemented, correction is performed only when one or a plurality of pieces of information on the positions, tangent directions, and curvatures of the plurality of feature points are included within a predetermined value range. The image identification system for identifying an object or a shape according to any one of claims 7 to 11.
Contour extraction processing for extracting the contour of an object or shape from target image data to be detected; and
A feature point detection process for detecting a feature point group from the extracted contour;
Using the detected feature point information, the contour between the feature points is divided into one or more regions, and a partial contour generation process for generating the divided regions as partial contours;
An image identification program that causes a computer to execute one or more of the generated partial contours and perform contour matching processing for matching contour information.
The image identification program includes:
The contour information reconstructing process for newly reconstructing a partial contour from other than the selected partial contour is further included, and the computer performs the contour matching process using the reconstructed partial contour. 13. The image identification program according to 13.
The image identification program includes:
14. The method according to claim 13, further comprising a contour complementing process for complementing a contour between any two points among the detected feature points, and causing the computer to execute the partial contour generating process using the complemented contour. 14. The image identification program according to 14.
16. The feature point according to any one of claims 13 to 15, wherein the feature point is a point at which the azimuth or curvature of the tangent line on the contour changes abruptly or falls within a predetermined value range. The image identification program according to Crab.
The reconstruction of the partial contour is performed by referring to the contour to be collated and using any or all of the length, direction, and curvature of the contour. The image identification program in any one.
When the contour is complemented, correction is performed only when one or a plurality of pieces of information on the positions, tangent directions, and curvatures of the plurality of feature points are included within a predetermined value range. The image identification program according to claim 13, wherein:
A computer-readable storage medium in which the image identification program according to any one of claims 13 to 18 is recorded.