JPS62127984A - Rotation-deviated character recognizing device - Google Patents

Abstract

PURPOSE:To read a character at a real time by placing a mark on a character to obtain the direction of rotation and to apply rotation correction processing to a character area only. CONSTITUTION:A mark 4 recognizing the direction of a character is provided, an angle is obtained from the mark 4, the correction of rotation is applied to recognize the character. When the mark is close to a horizontal or a vertical position, four points A-D are not obtained accurately. This is due to noises, then both ends of the mark is forecast from the point A (leftmost point) and the point D (rightmost point) being points obtained accurately, four points A'-D' are obtained by picture data compression and both ends P1, P2 of the mark are obtained from the said four points. The angle is obtained from the ends P1, P2, and at which side of the area around the mark the character exists is obtained to decide the quadrant. In converting tilted information into a regular position, a vector for information reading is decided by the quadrant and angle obtained is this way.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to reading characters with rotational deviation, and particularly to a character rotation correction device suitable for reading in real time.

[Background of the invention]

Conventional devices, as described in Japanese Patent Application Laid-open No. 60-23390, read only when one character is placed in the correct position. There was no mention of character reading when the direction is undefined.

[Purpose of the invention]

The purpose of the present invention is to read characters shifted by one rotation in real time,
The purpose is to provide information. This makes it possible to automate the sorting process, which was conventionally done by one person, and to automate the entire system.

[Summary of the invention]

Conventionally, it has been common practice to process character readings on the assumption that there are no variations in the inclination, rotation, or size of a single character. In particular, there is no precedent for reading characters when one character is rotated out of alignment, and it has been difficult to implement screen rotation correction processing in real time to correct characters to a normal state.

The present invention determines the direction of rotation by placing a mark on one character, and performs rotation correction processing on only the area of one character.

Achieved the reading @shi of characters.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to FIG.

This embodiment is a drum conveying section and a drum discriminating section.

It consists of a drum sorting section. A truck loaded with a mixture of drums with different contents stops at a predetermined position below the television camera 1, and the image processing device of the drum identification section reads the 42-character mark on the top of the drum, which describes the contents and drum type, and identifies the drum. Fully automatic management. The drums that have been read are sorted by the drum sorting section according to their identification numbers. In the figure, 2 is a character reading device, and 3 is a management computer for sorting.

With conventional equipment, the drum had to be electrically connected to the TV camera in the correct position as shown in Figure 2. Second
Figure 4 shows an image of the drum can in the position shown in the figure, captured by a television camera. In the case of the image shown in FIG. 4, it was possible to read it with a conventional device.

However, in this embodiment, in which a large number of drums are carried in, the drums are installed relative to the television camera as shown in FIG. FIG. 5 shows an image of the drum can in the position shown in FIG. 3, captured by the television camera 1.

In the case of the image of FIG. 5, it is impossible to read it with conventional equipment.

In this embodiment, as shown in FIG. 5, a cake 4 from which the direction of a character can be determined is provided, and the angle is determined from this mark 4 and rotation correction is applied to perform single character recognition.

Next, mark detection according to this embodiment will be explained. The input t-video contains noise and other drums, etc., and the projection distribution of image information in the X direction, X direction, and \ shown in FIG. 6 is performed, and the main parts are cut out as shown in FIG. 7. 7th
The image information contour tracking process shown in FIG. 8 is performed on the video in question. From the contour ruins, four points can be found: point A (leftmost point), point B (lowest point), point 0 (topmost point), and point D (leftmost point) shown in FIG. From these four points, the midpoint of AB is P1
．． Find the midpoint P2 of the CD, and find the angle θ as both ends of the mark.

At this time, as shown in FIG. 10, if the mark is close to horizontal or close to vertical, A, B, C.

Is it possible to accurately find the four points of D? this is.

Although it depends on the noise, for this reason, in Fig. 10, both ends of the mark are predicted from the points that can be accurately determined, point A (leftmost point) and point D (leftmost point), and by image data compression shown in Fig. 11. Four points A', B', C', and D' are found, and from these four points both ends of the mark PI and P2 are found.

The angles of PI and P2 are determined, and as shown in FIG. 12, the quadrant is determined by determining on which side of the t area the character is located with the mark centered.

To convert the tilted information to the correct position, the information reading vector is determined based on the obtained quadrant and angle.

According to the read vector, reading is performed n times depending on the size of the area as shown in FIG. 13, and the data is written in a fixed vector to another memory shown in FIG. 44.

Through the above processing, Figure 5 is replaced with the image in Figure 15,
Performs character recognition processing.

According to this embodiment, the center position of the drum, and.

The objective distance between the television camera and the drum can be sent to the drum sorting section, which has the effect of creating crane operation data.

The time required for one-turn correction is about 0.5 seconds, and compared to conventional character reading devices, the processing speed is almost on par with that of conventional character reading devices. Furthermore, because it uses a software-based character recognition algorithm and can determine character areas using marks, it achieves higher recognition accuracy than conventional character reading devices.

In this way, it is possible to read fixed-rotation characters at high speed and with high precision, which is unachievable with conventional character reading devices, and the sorting work that was previously done by humans and logistics management can be fully automated. , which made it possible to reduce the cost of the entire system.

〔Effect of the invention〕

According to the present invention, since it is possible to read characters that may be tilted in all directions, there is a great effect on sorting and management in actual distribution lines.

[Brief explanation of drawings]

Fig. 81 is a system configuration diagram of an embodiment of the present invention; Fig. 2;
3 is a front view of the television camera, FIGS. 4 and 5 are input video diagrams, FIG. 6 is a projection information diagram of FIG. 5, and FIG. 7 is a cutout video diagram of FIG. 5. Figure 8 is a diagram of the contour tracking process, Figure 9 is a diagram showing the points obtained from the results of Figure 8, Figure 10 is a diagram showing the results of Figure 8, and points that are incorrect due to noise, and Figure 11 is a diagram showing the points found as a result of Figure 8. A diagram showing the points obtained by image data compression, Figure 12 is a diagram showing the quadrant determination method, Figures 13 and 14 are illustrations of the principle of rotation correction, and Figure 45 is an image diagram of Figure 5 after rotation correction. It is. 1...TV camera, 2...Character reading device, 3...
- For sorting, use the management computer, mark 4.

Claims (1)

[Claims] 1. Inputting symbolic information into a storage device using an optical device,
The rotational shift character recognition device is characterized in that the symbol information is provided with a mark indicating a direction in the symbol information recognition device. 2. In claim 1, the symbol information is input into a two-dimensional image memory, the inclination angle of the symbol is determined by recognizing a mark indicating the direction of the symbol information, and the symbol information is input into the two-dimensional image memory. A rotational misalignment character recognition device characterized by correcting the character to the correct position when transferring the character to a computer.