JPS5971581A - Picture recognizing method - Google Patents

Abstract

PURPOSE:To recognize an object with high accuracy in a short time by judging the degree of difficulty of the recognition of the object depending on the level noise existing at a recognizing area to process in response to the degree. CONSTITUTION:After the picture information in the recognizing area obtained by a TV camera 11 is processed by a picture processor 12, the result is stored in a picture memory 13. The picture element information is read out from the memory 13 according to the direction of scanning, a prescribed threshold is provided by a picture binary-coding circuit 14 in a recognition control section 20 for attaining the binary-coding, and the degree (1-3) of difficulty is discriminated depending on the length of scanning direction of noise by a difficulty judging circuit 15. One of processing circuits 161, 162, 163 is selected based on the discriminated difficulty, allowing to attain the processing according to each algorithm. As a result, noise is discriminated and eliminated and only X, Y coordinates of the object are stored in an X-Y table 17. Thus, the objective is recognized with high accuracy at high speed.

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Field of the Invention The present invention is directed to the recognition area obtained by a video camera, etc.
The present invention relates to an image recognition method for gRing one position of the shape of an object from +2.

(2) Prior Art and Problems Conventionally, there have been various methods for recognizing the shape and position of an object based on the image power obtained by a video camera or the like.

For example, by scanning in all directions of the object that exists with 6 noises within the recognition field 1, we find the change point of the pixel R scene "1" and 1g signal relative to "0", and use this point as the reference point. The contour line is traced as a point, and each point is determined to determine whether it is noise or an object. This has the disadvantage that it is subject to noise of various sizes and is very time consuming compared to when a single algorithm is applied.

The present inventor set several degrees of recognition depending on the size of noise for four known distances, and applied the following algorithm to each of them. This product focuses on the ability to significantly shorten the overall processing time for rice.

(3) Purpose of the Invention The purpose of the present invention is to set the difficulty level of object recognition according to the size of noise existing in the recognition area, and to recognize a single image in a short time and with high precision. The purpose is to provide a method.

(4) Structure of the Invention - Previous In order to differentiate the purpose, the image recognition method of the present invention recognizes the shape and position of the object from the painter within the image recognition area. The size of noise existing in the recognition area is determined based on multiple difficulty levels set in relation to the shape of the object, and the optimal processing method is selected according to the determined difficulty level, 1.It is characterized by recognizing household objects.

(5) Embodiments of the Invention To explain the present invention in detail, first, if the recognition area is "1"
1) If the noise is present and not overwhelming, then the R view consisting of the clear report of the heavy object and the clear information of 17
Algorithm A, which is 11 points in length, can easily recognize objects on the ball. However, if there is noise that has the same information as the target object, Algorithm A, which does not take noise into account, will likely misrecognize it. However, if these are combined, the overall processing time can be shortened, and the setting of actual browning according to the following example is and g abbreviation of the processing means corresponding to each will be explained.

FIG. 1 (α) shows an example of the rules applicable to the embodiments of the present invention.

In the figure, the lower trapezoidal portion of the bonding pad 2 is shown as an object consisting of "1" pixel within the recognition area 1.

Then, the center P (z+1/) of the base l in the row direction is determined, and the noise 4 of length m, composed of "1" pixels, resides in the R landscape area 6 composed of "0" picture system. think about it.

The degree of difficulty in recognizing the trapezoidal part of this object is divided into the following 6a categories based on the relationship between t and m.

Difficulty level ■; If z7L = 0 Difficulty level ■; 0 (m (If l Difficulty level ■; m ≥ l (D) From this, if difficulty level ■, the same length as l without considering noise Set the line of ``S'', use it as the y point of point tp, check the leftmost pixel of ``1'' of the line you found, and add 1/2 to that point. demand.

In the case of difficulty level ■, the V coordinate may be the same as in the case of browning 1, but if the X coordinate is detected in the same way, there is a possibility of misrecognition due to isolated noise. Point P is determined by matching with the corner detection template 5 of the trapezoidal part.

-When browning ■ occurs, noise with the same length as l exists, so when determining the y coordinate, determine it by examining the rate of change in the number of pixels in each row. In the case of a bonding pad, this is a 45 degree angle (A-A) as shown in Figure Cb).
'), so the rate of change is accordingly. In the case of noise, it is irregular and can be easily distinguished.

Also, z! The collar is determined by matching with a template 6 for a bonding pad that matches the trapezoidal part shown in FIG. 2(c).

FIG. 2 is an explanatory diagram of the configuration of an embodiment of the present invention based on the above-mentioned principle.

In the same figure, image information within the recognition area obtained by the TV left camera 1? After being processed by the image processing device 12, each image is stored in the image memory 16. The recognition control unit 2 reads out pixel information from this image memory 16 in accordance with the scanning direction, and
A binarization circuit 14 sets a predetermined threshold value and binarizes the image, and a difficulty level determination circuit 15 determines the difficulty level ■, ■, ■ according to the length of the noise in the scanning direction. Processing circuit (1) 16x - (II) 1 according to this determined difficulty level
62. (l [D) 16g and performs all the processing according to each of the algorithms described above. As a result, noise is identified and removed, and only the x and Y coordinates of the object are stored in the XY table 17. You can see the object at high speed and in the garden.

FIG. 6 is a flow chart showing the operation of the embodiment of FIG.

In the figure, the recognition control unit 20 of the present invention is based on the input theory of one song.
As described above, step 9 is to binarize the pixel information and determine the difficulty level.

In the case of difficulty level ■, a line equal to l is found because there is no noise, positions where the value changes to "1" are extracted from both ends of the line, and the object is recognized.

In the case of difficulty level ■, a line equal to l is found, matching is performed from both ends of the line using corner templates, and the object is recognized.

In the case of Usui brown discoloration (■), l is determined by checking the pixel coverage ratio in the row direction to find l. Then, matching is performed using the ten-rate for the pad corners to determine both ends, and the entire object is recognized.

(6) Effects of the invention As explained above, according to the present invention, the difficulty level is determined based on the size of noise existing in the recognition area
11f The process is performed by selecting a processing means using a corresponding algorithm according to the difficulty level. To this ↓shi,
In contrast to conventional processing methods that use a single algorithm to process uniformly, this method simplifies the processing as a whole, and
It can be made faster. Furthermore, since it is possible to apply the processing means that is most suitable for the level of difficulty, it is possible to recognize the accuracy.

[Brief explanation of drawings]

1C(L) to (c) are schematic diagrams of an embodiment of the present invention and explanatory diagrams of essential parts; FIG. 2 is a diagram illustrating the configuration of an embodiment of the present invention;
The figure is a flowchart showing the operation of the embodiment of FIG.
1 is recognition area, 2 is object, 6 is RM14 is noise, 5
, 6 is Tenblay), 11f-1: TV camera, 12 is image processing device, 16 is image memory, 14 is image 1#! A binarization circuit, 15 is a difficulty level -■ constant circuit, 16 to 163 are processing circuits, and 17 is an XY table. Patent Applicant: Fujitsu Limited Temporary Agent Patent Attorney 1) Yoshi Saka ]j Figure 1 (a) Figure 2

Claims (2)

[Claims] (1) In an image recognition method that recognizes two positions of the shape of an object from an image within a recognition region, the magnitude of noise existing within the recognition region is determined by comparing the magnitude of noise existing within the recognition region with the shape of the object. Determine the difficulty level, select the optimal processing method according to the determined difficulty level, and recognize the target object.
Featured image recognition method. (2) The setting is based on the presence or absence of noise within the self-recognition area and the relationship between the noise and the length of the object in the row direction. The image recognition method according to scope 1.