JP2003223640A - Object recognizing device - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an object recognizing device capable of improving an S/N value and improving the recognition rate of an object by adjusting an intensity distribution range of feature quantities. <P>SOLUTION: This object recognizing device recognizes the object by selecting the feature quantities from image data of the object and comparing them with feature quantity information of pre-constructed database. This device is provided with an image input device capable of externally changing a set value and incorporating the image data of the object, and a recognizing device selecting the feature quantities based on the image data, adjusting the set value of the image input device, optimizing the intensity distribution range of the feature quantities, collating it with the feature quantity information stored in the database, and recognizing the object. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object recognizing device for recognizing a target object by extracting a feature value from image data of the imaged object and comparing it with the feature value information of a database previously constructed. In particular, the present invention relates to an object recognition device capable of improving the S / N value by adjusting the intensity distribution range of the feature amount and improving the recognition rate of the object.

[0002]

2. Description of the Related Art A conventional object recognition device is an RGB (one of color models that digitizes color information from image data of an object acquired from an image input device such as a camera, a scanner, or a video camera. Red). , Green, Blue
Express color by. ) A feature value such as a value is extracted and compared with feature amount information stored in advance in the database, and when the feature amounts match, the object is identified based on the information in the database. is there.

FIG. 6 is a block diagram showing an example of such a conventional object recognition apparatus. In FIG. 6, 1 is an image input device such as a video camera, 2 is feature amount extraction means for extracting feature amounts from input image data, 3 is recognition processing means such as a computer, 4 is a hard disk or CD-RO.
It is a storage means such as M (Compact Disc Read Only Memory). Further, 2, 3 and 4 constitute a recognition processing device 50.

The output of the image input device 1 is the feature amount extraction means 2
And the output of the feature amount extraction means 2 is connected to the recognition processing means 3
Connected to. The inputs and outputs of the storage means 4 are connected to the recognition processing means 3, respectively.

Here, the operation of the conventional example shown in FIG. 6 is shown in FIG.
This will be described with reference to FIGS. 8 and 9. FIG. 7 shows recognition processing means 3.
FIG. 8 is an explanatory diagram for explaining an example of feature amount extraction, and FIG. 9 is an explanatory diagram for explaining collation between the extracted feature amount and data in the storage unit 4.

In "S001" in FIG. 7, the recognition processing means 3 determines whether the feature amount extraction means 2 has acquired image data from the image input device 1.

If the feature amount extraction means 2 acquires image data in "S001" in FIG. 7, "in FIG. 7"
In step S002 ", the recognition processing unit 3 uses the feature amount extraction unit 2
Control to extract the feature amount from the acquired image data,
When the image data is not acquired, “S00” in FIG.
Return to step 1 ".

For example, when the object shown by "OB11" in FIG. 8 is photographed by the image input device 1 and the image data as shown by "GD11" in FIG. 8 is acquired, the feature quantity extraction means 2 is shown in FIG. A characteristic curve showing the characteristic amount frequency as shown in "CH11" is extracted.

In "S003" in FIG. 7, the recognition processing means 3 collates the feature quantity extracted by the feature quantity extraction means 2 with the feature quantity information of the database previously built in the storage means 4.

For example, the characteristic of the intensity distribution range indicated by "RG11" in FIG. 8 in the characteristic curve indicated by "CH11" in FIG. 8 is compared with the characteristic amount information of the database previously constructed in the storage means 4.

That is, since the intensity distribution range of the characteristic amount of the acquired image is unique and constant by the image input device 1, the object can be recognized by collating such intensity distribution range.

Then, in "S004" in FIG. 7, if the recognition processing means 3 recognizes the object, in other words, within the characteristic curve indicated by "CH21" in FIG.
The characteristics of the intensity distribution range shown in the middle "RG21" and "in Fig. 9"
When the feature amount information of the database previously constructed in the storage means 4 shown in DD21 "matches," S00 "in FIG.
Processing after object recognition shown in 5 "is performed.

In "S004" in FIG. 7, if the recognition processing means 3 does not recognize the object, the recognition processing means 3
Returns to the step indicated by "S001" in FIG.

As a result, it is possible to recognize the object by extracting the feature amount from the image data acquired by the image input device 1 and comparing the feature amount with the feature amount information of the database previously constructed. Become.

[0015]

However, in the conventional example as shown in FIG. 6, the feature amount necessary for recognition of the object is limited to the intensity distribution range peculiar to the image input apparatus 1, and the intensity distribution range is limited. There was a problem that it was difficult to expand appropriately.

Further, since the image data includes a feature amount necessary for recognizing the target object and a feature amount unnecessary for recognizing the target object, recognition of the target object with respect to the feature value necessary for recognizing the target object is performed. Unnecessary feature amount corresponds to the noise component, and S
There was a problem that the / N value deteriorates.

For example, for the intensity distribution range indicated by "RG21" in FIG. 9, "RG22" and "RG23" in FIG.
The intensity distribution range indicated by is a noise component unnecessary for recognition of the object. Therefore, the problem to be solved by the present invention is to adjust the intensity distribution range of the feature quantity to adjust the S / N ratio.
It is to realize an object recognition device that can improve the value and the recognition rate of an object.

[0018]

In order to achieve the above object, the invention according to claim 1 of the present invention is characterized in that a feature amount is extracted from image data of an object and the feature of a database constructed in advance. In an object recognition device that recognizes the target object by comparing it with the amount information, an image input device that can change the set value from the outside and that captures image data of the target object, and a feature value is extracted based on the image data. And a recognition processing device that adjusts the setting value of the image input device to optimize the intensity distribution range of the feature amount and collates with the feature amount information stored in the database to recognize the object. This makes it possible to improve the S / N value by adjusting the intensity distribution range of the feature amount and improve the recognition rate of the target object.

According to a second aspect of the present invention, in the object recognition apparatus according to the first aspect of the present invention, the recognition processing device is:
A feature amount extraction unit that extracts a feature amount based on the image data obtained from the image input device, a storage unit that stores feature amount information of a plurality of objects as the database, and a set value of the image input device. And a setting change unit for changing the setting value of the image input device by controlling the setting change unit to optimize the intensity distribution range of the feature amount output from the feature amount extraction unit and to obtain the obtained feature. The intensity distribution range of the feature amount is adjusted to improve the S / N value by the recognition processing unit that recognizes the object by collating the intensity distribution range of the amount with the feature amount information. It is possible to improve the recognition rate of the object.

According to a third aspect of the present invention, in the object recognition apparatus according to the first or second aspect of the invention, the image input device and the recognition processing means, or the image input device and the feature amount extraction means. Further, since the setting changing means is connected to the network by the network, it is possible to adjust the intensity distribution range of the feature quantity to improve the S / N value and improve the recognition rate of the object.

According to a fourth aspect of the invention, in the object recognition apparatus according to the third aspect of the invention, the network is
Since it is the Internet, it is possible to improve the S / N value by adjusting the intensity distribution range of the feature amount and improve the recognition rate of the object.

According to a fifth aspect of the present invention, in the object recognition apparatus according to the third aspect, the network is
Since it is a LAN, it is possible to adjust the intensity distribution range of the feature quantity, improve the S / N value, and improve the recognition rate of the object.

According to a sixth aspect of the present invention, in the object recognition apparatus according to the third aspect, the network is
By using the wireless LAN, it becomes possible to improve the S / N value by adjusting the intensity distribution range of the feature quantity and improve the recognition rate of the target object.

The invention according to claim 7 is the object recognition apparatus according to claim 3, wherein the network is
The WAN makes it possible to adjust the intensity distribution range of the feature quantity to improve the S / N value and improve the recognition rate of the object.

According to an eighth aspect of the present invention, in the object recognition apparatus according to the first or second aspect of the present invention, the image input device is a video camera, and the intensity distribution range of the feature quantity is adjusted. It is possible to improve the S / N value and improve the recognition rate of the object.

According to a ninth aspect of the present invention, in the object recognition apparatus according to the first or second aspect of the present invention, the image input device is a scanner, and the intensity distribution range of the feature quantity is adjusted. It is possible to improve the S / N value and improve the recognition rate of the object.

According to a tenth aspect of the present invention, in the object recognition device according to the first or second aspect of the present invention, the storage means is a rewritable storage means so that the intensity distribution range of the feature quantity is It becomes possible to adjust the S / N value to improve the recognition rate of the object.

According to an eleventh aspect of the present invention, in the object recognition apparatus according to the first or second aspect of the invention, the storage means is a read-only storage means, so that the intensity distribution range of the feature quantity is It becomes possible to adjust the S / N value to improve the recognition rate of the object.

According to a twelfth aspect of the present invention, in the object recognition apparatus according to the first or second aspect of the present invention, the set value of the image input device is adjusted so that the skin color is most emphasized, and the feature amount is set. By recognizing the human face by the extraction, the recognition rate of the human face can be improved.

[0030]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below with reference to the drawings. FIG. 1 is a configuration block diagram showing an embodiment of an object recognition apparatus according to the present invention. In FIG. 1, reference numerals 2 and 4 are the same as those in FIG. 6, and reference numeral 5 is an image input device such as a video camera whose setting values can be changed by external input or the like.
Is a recognition processing means such as a computer, and 7 is an image input device 5.
It is a setting changing means for changing the set value of.

Also, 2, 4, 6 and 7 are recognition processing devices 5.
1, the optimum feature amount information obtained by adjusting the set values of the image input device 5 is stored in the storage means 4 in advance as a database.

The output of the image input device 5 is the feature amount extraction means 2
And the output of the feature amount extraction means 2 is connected to the recognition processing means 6
Connected to. Further, the input and output of the storage means 4 are mutually connected to the recognition processing means 6.

Further, the control output of the recognition processing means 6 is connected to the control input terminal of the setting changing means 7, and the setting changing means 7 is connected.
The setting change output of is connected to the external input terminal of the image input device 5.

The operation of the embodiment shown in FIG. 1 will now be described with reference to FIG.
This will be described with reference to FIGS. 3 and 4. FIG. 2 shows the recognition processing means 6
4 is a flowchart for explaining the operation of FIG. 4, FIG. 3 is an explanatory view for explaining an example of feature amount extraction and feature amount intensity distribution range adjustment, FIG.
FIG. 4 is an explanatory diagram for explaining the matching between the extracted feature amount and the data of the storage unit 4.

In "S101" in FIG. 2, the recognition processing means 6 determines whether the feature quantity extracting means 2 has acquired image data from the image input device 5.

If the feature amount extraction means 2 acquires image data in "S101" in FIG. 2, "in FIG. 2"
In S102 ″, the recognition processing means 6 causes the feature amount extraction means 2 to operate.
Control to extract the feature amount from the acquired image data,
If the image data is not acquired, "S10" in FIG.
Return to step 1 ".

For example, when the object shown by "OB41" in FIG. 3 is photographed by the image input device 5 and the image data shown by "GD41" in FIG. A characteristic curve showing the characteristic amount frequency as shown in "CH41" is extracted.

In "S103" in FIG. 2, the recognition processing means 6 determines whether or not the intensity distribution range of the feature quantity obtained by the feature quantity extracting means 2 is optimum. In "S104" of 2, the recognition processing means 6 controls the setting changing means 7 to adjust the set value of the image input device 5, and returns to the step shown in "S101" in FIG.

For example, in the characteristic curve showing the frequency of the characteristic quantity as shown by "CH41" in FIG. 3, the characteristic quantity other than the intensity distribution range shown by "RG41" in FIG. 3 becomes a noise component as described above. If the S / N value is not good, the recognition processing means 6
If the determination is made, the setting value of the image input device 5 is changed via the setting changing means 7.

Specifically, the image input device 5 for changing the setting
The setting item corresponds to a dynamic range that is a signal gain of the image input device 5, an aperture amount for adjusting the input light amount so that the dynamic range is optimum, or selection of a filter.

Then, for example, the image input device 5 whose setting has been changed as described above captures an image of the object indicated by "OB41" in FIG. 3 and obtains image data as indicated by "GD42" in FIG. 3 again. In this case, the feature quantity extraction means 2 uses "C" in FIG.
A characteristic curve indicating the frequency of the characteristic amount as shown in H42 ″ is extracted.

For example, in the characteristic curve showing the frequency of the characteristic amount as shown by "CH42" in FIG. 3, there is no characteristic amount other than the intensity distribution range shown by "RG42" in FIG. 3, in other words, the noise component is Since it does not exist, the recognition processing means 6 determines that the intensity distribution range of the feature amount is optimum.

When the intensity distribution range of the feature quantity obtained by the feature quantity extraction means 2 is optimum in "S103" in FIG. 2, the recognition processing means 6 causes the feature quantity extraction means 2 in "S105" in FIG. The feature amount extracted in step 1 is collated with the feature amount information of the database previously built in the storage unit 4.

For example, the characteristic curve indicated by "CH51" in FIG. 4 is changed to "CH5" in FIG. 4 by changing the setting of the image input device 5.
2 ”is the characteristic curve shown in FIG.
Characteristics of intensity distribution range shown in medium "RG51" and storage means 4
The optimal feature amount information of the database previously constructed is collated.

Then, in "S106" in FIG. 2, if the recognition processing means 6 recognizes the object, in other words, within the characteristic curve indicated by "CH52" in FIG.
The characteristics of the intensity distribution range shown in the middle "RG51" and "in Fig. 4"
When the optimum feature amount information of the database previously constructed in the storage means 4 shown in DD51 "matches," S "in FIG.
Processing after object recognition shown in 007 "is performed.

In "S106" in FIG. 2, if the recognition processing means 6 does not recognize the object, the recognition processing means 6
Returns to the step indicated by "S101" in FIG.

As a result, the recognition processing means 6 judges whether or not the intensity distribution range of the feature quantity obtained by the feature quantity extracting means 2 is optimum, and if it is not optimum, the setting changing means 7 is controlled. By adjusting the set value of the image input device 5 by using the image input device 5, the S / N value of the intensity distribution range of the feature amount is improved, and the recognition rate of the target object can be improved.

Although the image input device 5 and the recognition processing device 51 are directly connected in the embodiment shown in FIG. 1, they may be connected from a remote place using the Internet or the like.

FIG. 5 is a block diagram showing the construction of another embodiment of the object recognition apparatus according to the present invention. In FIG. 5, reference numerals 4 and 6 are the same as those in FIG. 1, 8 is an image input device such as a video camera whose setting value can be changed from a remote place by external input, and 9 is input image data. A feature amount extracting means for extracting a feature amount, 10 is a setting changing means for changing a set value of the image input device 8, and 100 is the Internet which is a general-purpose network.

Further, 4, 6, 9 and 10 constitute a recognition processing device 52, and the storage means 4 stores in advance a database of optimum feature amount information obtained by adjusting the set values of the image input device 8. Have been accumulated as.

The output of the image input device 8 is the internet 1
Is connected to the feature amount extraction means 9 via 00, and the output of the feature amount extraction means 9 is connected to the recognition processing means 6. Further, the input and output of the storage means 4 are mutually connected to the recognition processing means 6.

Further, the control output of the recognition processing means 6 is connected to the control input terminal of the setting changing means 10, and the setting change output of the setting changing means 10 is connected to the external input terminal of the image input device 8 via the Internet 100. It Also, FIG.
The operation of the embodiment shown in FIG. 2 is the same as that of the embodiment shown in FIG.

In the embodiment shown in FIG. 5, the Internet is illustrated as a general-purpose network, but LAN is used.
(Local Area Network), wireless LAN, WAN (Wide A
(reaNetwork) etc., of course.

With such a configuration, even in an image input device such as a video camera installed in a remote place, the S / N value can be improved by adjusting the intensity distribution range of the feature quantity, It is possible to improve the recognition rate.

Although a video camera is shown as an example of the image input device in FIG. 1 or FIG. 5, any device capable of outputting the image data of the object may be used, and the object is scanned to obtain the image data. It may be a scanner, a camera (including a digital camera or the like) for obtaining a still image, or the like.

Although a hard disk, a CD-ROM or the like is illustrated as the storage means in FIG. 1 or 5,
Of course, the present invention is not limited to this, and any data can be stored as long as it can store a database regardless of whether or not it can be rewritten.

For example, as a rewritable storage means,
DVD-RAM (Digital Versatile Disc Random Acce
ss Memory), CD-RW (Compact Disc ReWritbl
e), MO (magneto-optic), memory card, etc., and the read-only storage means is a DVD-ROM
(Digital Versatile Disc Read Only Memory), CD
-R (Compact Disc Recordable), etc.

Further, it is possible to improve the recognition rate of a person's face by adjusting the setting value of the image input device so that the "skin color" which is a human feature is most emphasized and extracting the feature amount. Become.

[0059]

As is apparent from the above description,
The present invention has the following effects. Claims 1, 2,
According to the inventions of 3, 4, 5, 6, 7, 8, 9, 10 and claim 11, whether or not the intensity distribution range of the feature quantity obtained by the feature quantity extraction means by the recognition processing means is optimum. If it is not optimal, the setting change means is controlled to adjust the setting value of the image input device, so that the S / N value of the intensity distribution range of the feature amount is improved and the recognition rate of the object is improved. It will be possible to improve.

According to the twelfth aspect of the present invention, the set value of the image input device is adjusted so that the "skin color", which is a human feature, is most emphasized, and the feature amount is extracted to extract the human face. It is possible to improve the recognition rate of.

[Brief description of drawings]

FIG. 1 is a configuration block diagram showing an embodiment of an object recognition device according to the present invention.

FIG. 2 is a flowchart illustrating an operation of a recognition processing unit.

FIG. 3 is an explanatory diagram illustrating an example of feature amount extraction and feature amount intensity distribution range adjustment.

FIG. 4 is an explanatory diagram illustrating matching between the extracted feature amount and data in a storage unit.

FIG. 5 is a configuration block diagram showing another embodiment of the object recognition device according to the present invention.

FIG. 6 is a configuration block diagram showing an example of a conventional object recognition device.

FIG. 7 is a flowchart illustrating the operation of the recognition processing means.

FIG. 8 is an explanatory diagram illustrating an example of feature amount extraction.

FIG. 9 is an explanatory diagram illustrating matching between the extracted feature amount and the data in the storage unit.

[Explanation of symbols]

1, 5, 8 image input device 2.9 Feature extraction means 3,6 Recognition processing means 4 storage means 7, 10 Setting change means 50, 51, 52 Recognition processing device 100 Internet

   ─────────────────────────────────────────────────── ─── Continued front page    (72) Inventor Ryotatsu Iga             2-9-32 Nakamachi, Musashino City, Tokyo Yokogawa             Electric Co., Ltd. (72) Inventor Gentaro Fukano             2-9-32 Nakamachi, Musashino City, Tokyo Yokogawa             Electric Co., Ltd. F-term (reference) 5B047 CA00 CB21                 5C054 CC05 DA06 EA05 EA07 FB03                       FC03 FC07 FC12 FC15 FE13                       GB12 HA05                 5L096 CA02 EA01 JA11

Claims (12)

[Claims] 1. An object recognition apparatus for extracting a feature amount from image data of an object and comparing it with feature amount information of a database constructed in advance to recognize the target object. An image input device capable of capturing image data of the object, and extracting a feature amount based on the image data and adjusting a setting value of the image input device to optimize the intensity distribution range of the feature amount and store it in a database. An object recognition apparatus, comprising: a recognition processing apparatus that recognizes the target object by collating with the generated feature amount information. 2. The recognition processing device stores feature amount extraction means for extracting a feature amount based on the image data obtained from the image input device, and feature amount information of a plurality of objects as the database. Storage means, setting changing means for changing the setting value of the image input device, and intensity of the characteristic amount output from the characteristic amount extracting means by controlling the setting changing means to adjust the setting value of the image input device. 2. The recognition processing means for recognizing the object by optimizing the distribution range and collating the intensity distribution range of the obtained feature amount with the feature amount information. Object recognition device. 3. The image input device and the recognition processing means, or the image input device, the feature amount extraction means and the setting change means are connected by a network. 2. The object recognition device according to 2. 4. The object recognition device according to claim 3, wherein the network is the Internet. 5. The object recognition device according to claim 3, wherein the network is a LAN. 6. The object recognition device according to claim 3, wherein the network is a wireless LAN. 7. The object recognition device according to claim 3, wherein the network is a WAN. 8. The object recognition device according to claim 1, wherein the image input device is a video camera. 9. The object recognition device according to claim 1, wherein the image input device is a scanner. 10. The object recognition device according to claim 1, wherein the storage means is a rewritable storage means. 11. The object recognition device according to claim 1, wherein the storage means is a read-only storage means. 12. A human face is recognized by adjusting a set value of the image input device so that a skin color is most emphasized and extracting a feature amount. The object recognition device described.