JPS61163474A - Control device for optimizing character recognizing - Google Patents

Abstract

PURPOSE:To read always a character by an optimum binary level, to decrease an erroneous recognition and to improve a recognizing ratio by providing a reading object thickness detecting circuit, a binary level correcting circuit and a binary level correcting table. CONSTITUTION:A reading object character 2 of a frame 1 is picked up with an ITV camera 3, the value in which the numerical value of the picked-up picture 4 and the duplex level are analogized by a D/A converter 5, is compared by a comparator 6, and a binary picture 7 is stored in a frame memory 8. Next, a character composite element is detected through a reference point detecting circuit 9 at a reading object thickness detecting circuit 12 by the optimum thickness and when the number of picture elements except it is detected, a binary level correcting circuit 13 is operated. Namely, when the number of five picture elements or above is detected, a binary level is lowered through a binary level correcting table 14, and when the number of five picture elements or below is detected, the binary level is raised. Namely, the character can be always read by the optimum binary level, the erroneous recognition can be decreased, and the recognizing ratio can be improved.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to a device for controlling so that an optimal discriminant value can be obtained when character recognition is performed.

(Prior Art) Conventionally, the following method has been used to read characters written on a label or the like and have a machine recognize them. As shown in FIG.
A comparator 6 compares the photographed image 4 captured by the camera 3 with a value obtained by converting a value of a preset binarization level into an analog value by a D-A converter 5, and the binarized image 7 is a frame. It is stored in memory 8.

Next, the frame detection and reference point detection circuit 9 detects the reference points (the four corners and center of gravity of the frame), and the next segment detection device 10
The segment detection point is determined by Depending on whether the segment thus detected is on or off, the next character determining device 11 determines the content of the 40 characters in the imaged image (see Patent Application No. 59-194343).

In this case, the binarization level during binarization largely depends on the brightness of the base of the label, the brightness and illuminance of the font, or external light. Therefore, in the past, several types of binarization levels were set and entered in the selector 16, and the strength and weakness values of the original image 40 input from the camera 5 and the output value of the selector 16 were set to D-
The values converted into analog by the A converter 5 are compared and binarized.

(Problem to be solved by the invention) However, in order to respond to various conditions, it is necessary to prepare a large number of binarization level setting values, and it is necessary to prepare many setting values for the binarization level, and repeat the process many times until the optimum binarization level is reached. The disadvantage is that it takes a lot of time because binarization processing has to be performed.

Also, if you find it bothersome to change the settings each time you read a label, you can read it with a fixed setting value.
There was a problem in that erroneous recognition occurred because the characters became thicker or thinner. It is an object of the present invention to solve this problem, to always change the optimum image, and to ensure reliable character reading.

(Means for Solving the Problems) The present invention provides a method for solving the above problems by setting in advance the positions of character constituent elements that are necessarily present on an image for which character recognition is to be performed. After imaging this with an imaging device and binarizing it, we created a device that detects the converted thickness of the character component, and then set the binarization level so that the detected thickness becomes a constant value. A device is configured to perform correction and control so that the character component at the specific position always has a constant thickness.

(Function) By installing the above-mentioned control device, the binarization level is always at the optimum value, regardless of the brightness of the label or the presence or absence of external light, and all the compositions of the characters are controlled. The entire element, or character, can always be detected as having a constant thickness. In other words, since the characters are output as images with a certain strength,
This shows that an excellent character recognition device has been realized with fewer errors in character discrimination.

(Example y) An example of the present invention will be explained with reference to the drawings. Fig. 1 is an overall configuration diagram. This is an improved version of the prior art shown in Fig. 6, and the same members are denoted by the same reference numerals. In the figure, the frame 1, the reading object 2, the ITV camera 3, and the photographed image 4 have exactly the same functions as in Figure 6.The device 5 for D-A converting the binary level value is also the same, and the comparator 6
At 9, the binarized image information 7 is compared with the captured image 4 and stored in the frame memory 8, the reference point is detected at 9, the character content is determined by the segment detection device 10 and the character determination device 11. The determination is also the same as in FIG. However, in this embodiment, 7-segment numbers are used as objects to be read, and the representation as shown in FIG. There is always a display segment in f.

Further, in this embodiment, a three-digit number is used, but when the number is 9 or less, there may be only one character as shown in Fig. 2 ←→. Even so, a number always exists in the rightmost digit, and therefore a character component always exists in C or f of the seven segments in the rightmost digit. Further, since the position of the character is correctly detected by the reference point detection circuit 9, the character component of C or f at the rightmost digit can be reliably detected, but this is detected by the reading target thickness detection circuit 12. In this case, the number of pixels in the entire captured image is 256x256 in total, but the number of pixels seen when detecting a character component of C or fK is 0 to 1.
It is normal for the number to reach 0.

In this case, when a number of pixels of about 5 is detected, it can be considered that the character component has been detected with the optimal thickness, and when any other number of pixels is detected, the next binarization level correction circuit 13 works. That is, when the number of pixels of 5 or more is detected, the binarization level is lowered, and when the number of pixels of 5 or less is detected, the binarization level is increased. In this case, since the detected value is also a digital value and the output is also a digital value, digital control is performed. For example, the correction table 14 shown in FIG. 3 is used to register for this control. Character recognition is performed using the binarization level value corrected in this way, and when the optimal binarization level is reached (in this case, when the number of pixels reaches 51C), the label is clipped. Make it. A flowchart showing the process of binarization level correction is shown in FIG.

Even if the character to be read is a character other than a segment numeral, a spot corresponding to the character component of C or f mentioned above is provided at position 15 in the figure as shown in Figure 5, and the thickness of this spot is adjusted. It is possible to detect and correct the binarization level in the same manner as described above.

In this way, whether it is a number or a character, reading can be performed when the highest binarization level is reached.

(Effects of the Invention) As explained above, according to the present invention, characters can always be read at the optimum binarization level without being affected by ambient light during changes in the illuminance of the light source or by the brightness of the background. will be held. This reduces the risk of misrecognition and improves the recognition rate. Maku,
It takes less time to reach the optimal binarization level, and the operating time is also advantageous. Furthermore, restrictions on the installation location will be relaxed, and since a wide variety of colors can be used for labels, it will be possible to indicate by color, and major effects are expected in secondary matters, such as increasing the number of types of labels.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram showing the configuration of the present invention in a block diagram form, and Fig. 2 is an explanatory diagram of the reading target, where (a) is a diagram of the structure of numbers with 7 segments, and (b) is a diagram of the 7-segment numbers. Configuration diagram, (c) is a diagram of the character configuration of the entire label without hundreds and tens digits, and Figure 3 shows the number of segment pixels for thickness detection,
An explanatory diagram showing a correction table for binarization level increments, and FIG. 4 is a flowchart of reading processing and optimum binarization level determination processing. FIG. 5 is a 1-indentation showing a label surface in an embodiment in which characters other than 7-segment numbers are used, and FIG. 6 is a block diagram explanatory diagram of a character reading process according to the prior art. 1 Reading target frame 2 Reading target character 3 IT
V camera 4 Captured image 5D-A converter 6 Comparator 7 Binarized image 8 Frame memory 9 Reference point detection circuit 10 Segment detection circuit 11 Character judgment device 12 Character component thickness detection circuit 13 Binarization level correction circuit 14 Binarization level correction table foil 1 Figure 1... Reading front book 2... - Character 3... ■Ding Vυ Mera 4... ・ Photographed tatami 3 image 7... 200 million image blasphemy report Figure 2 (a) Figure 3 Figure 4

Claims (2)

[Claims] (1) Characters are photographed by an imaging device, the image is binarized, and in a device that recognizes characters, the positions of character constituent elements that will always be present on the image for character recognition are set in advance. A device capable of capturing an image, and the image capturing device,
After imaging and binarizing the character component, a device for detecting the converted thickness of the character component; and a device for binarizing the character component so that the thickness detected by the detection device has a constant value. A character recognition optimization control device comprising a control device for correcting a level. (2) After a character component at a specific position is imaged, binarized, and converted, a device that detects the thickness of the character component detects the number of pixels in the binarized and converted screen segment. The character recognition device according to claim 1, wherein the control device for correcting the binarization level is a device for controlling the number of pixels of the segment to always maintain a constant value. Optimization control device.