SU130711A1 - The method of recognition of printed text and device for its implementation - Google Patents

Description

Methods of machine reading alphabetic, text, and using topological methods are known. However, in these works, a connectivity matrix is used for identification, for the compilation of which a digital computer is used.

The proposed method of recognizing printed text (symbols) with index definition points (where index is the number of arcs exiting a point), besides points with index two, differs from the known ones, that instead of the connectivity matrix there are code combinations that get determining the indices of points, which greatly simplifies the analysis of symbols.

The proposed method allows one to use technically simpler means for text analysis: a reading node connected to a follow-up sweep circuit and a logic circuit defining a sequence of indices.

The operation of the device implementing the proposed method is illustrated in the drawing.

The essence of the described method is as follows. A definite beginning of the bypass of the outer contour is established (for example, the lower left part of the letter in the line) and the points of intersection of the segments are determined and given (the points of the contour break). One arc sang, the point is assigned the index one and the code / corresponds to it. If there are two arcs at the point of convergence (index two), such a point does not take into account. 1-1of the end, the points with an index greater than two correspond to the code “O.

Thus, bypassing the contour, a sequence of binary digits (printed character code) is obtained, which allows to break the studied set of printed characters into classes (the entire alphabet is divided into 17 classes), of which will contain one or several letters. Separation of characters within a class can be done nij

.NO 13071M

a known method of recognition, for example, by the method of minimal description.

A device that implements the proposed method works as follows. (The drawings do not show those nodes that are not fundamental and without which understanding of the operation of the device is not difficult).

The beam from the screen of the cathode ray tube (Fig. 1) is projected onto a letter. With the help of the tracking sweep device, its generator, 2 paraphase voltage generator, keys 3 and 4, storage cells 5 and (J and photocell 7 with amplifier 8, the beam is moved along the black-and-white contour of the mark. In case of violation of smoothness circuit, phase detector 9 with limiter 10 generates a signal that:

a) stops accumulation (opens the key //);

b) increases the amplitude of the generator 2 by the time of one period of its operation, which is achieved by means of a kip-relay 12;

c) directs the pulses appearing at the intersection of the beam making, for the purpose of determining the index of a point, a circular detour of a point with a small radius, with arcs of the image (via the key / 5) to the scheme performing the coding (Fig. 2).

The task of the coding scheme is to give an impulse “/ in the case of a point with an index of one, not to react to a point with an index of two, and to issue an impulse“ O for a point with an index of more than two. (These can be pulses of different polarity at the same output or pmpuls of the same polarity at different outputs) Since the time of complete separation of a point is equal to the generator 2 period (Fig. 1), the duration of the impulse group is always constant and not longer than this period.

The first pulse of the code group starts the Bipy-Relay 14 (Fig. 2) and through the closed keys 15 and 16 goes to the trigger 17, which by this moment is in the "zero state." Under the action of a pulse, the trigger is transferred to the state "/. Trigger 18 is in the "zero state." Thus, by the end of the period, the outputs of the triggers 1 and 18 are characterized by the states: the first - "/, the second -" O.

If the pulse group contains two pulses, the first one will bring the circuit to the state described above, and the second one, having passed through the keys 15 and 16 to the trigger 11, will transfer it to the "zero state." In this case, the key 16 will be closed and the key 19 will be opened. Thus, by the end of the period, the circuit is characterized by the following states of the triggers 17 and 18: the first is “O, the second is“ O. At the end of the period, the KIPP-relay 14 will bring the keys to the initial state, and the triggers to "zero."

If the pulse group has more than two pulses, then the first two will carry out the above, and the third will transfer the trigger l to the state “- /, thus opening the key 15. The pulses following the third will not be able to cause changes in the state x triggers, which in this case will be in the following positions: the first (17) - “O, the second (18) -,“ W. As before, at the end of the period, the initial state of the circuit will be restored.

The encoding results are fed to the scheme (in the drawing is not zhzhazana) for determining classes using the code tree to decode a sequence of values of points.

Before me invented

1. The way of recognizing printed text (characters) with determines cm index of points (where the index is the number of output from the point of the arcs), cr / m