SU1543431A1 - Device for identification of text images - Google Patents

Abstract

The invention makes it possible to identify text images with high confidence and can be used to automate the quality control of marking on electronic products. An increase in the reliability of identification is achieved by multiplying the values of the video signals coming from different parts of the image by the coefficients corresponding to the semantic significance of the image zones. The values of the semantic significance coefficients are stored in a weight memory block, the sample of which is synchronized with the scan of the text image. The decision on the quality of the textual image is made depending on the value of the sum of the signals from different parts of the studied and reference images multiplied by the coefficients of semantic significance. 4 il.

Description

The invention relates to automation and can be used in the automation of quality control of the application of marking on electronic products (IET).

The purpose of the invention is to improve the accuracy of the device.

FIG. 1 shows an example of identifying text images with various defects; in fig. 2 is a block diagram of the device; in fig. 3 is a control block diagram; in fig. 4 - a group of images with the same code distance.

FIG. 1a shows a reference defect image of the letter A, on

FIG. 1b and 1c - the image of the letter A with defects in the form of line breaks,

The text image identification device (Fig. 2) contains a television sensor 1, a comparison unit 2, a control unit 3, a measurement memory unit 4, a weight memory unit 5, a multiplier 6, an accumulator adder 7, a decision unit 8, an information input 9 devices, output 10 of the initial installation of the device and output II of the device.

9

The control unit (Fig. 3) contains a generator of 12 pulses and a counter 13. The width of the counter is 13

Ј

Ј

is made by the control inputs of blocks 4 and 5.

The device works as follows.

A pulse is applied to the input 10, which sets the blocks 6 and 7 to the initial state, the counter 13 of the block | 3 is set to the zero state. After that, the television sensor 1 converts the optical image of text characters, read from input 8, into an electrical signal, for example, a text label printed on the IET. At the output of block 1, digital signals are formed corresponding to the values of luminance at the points of the text image and background being studied. The signal from the output of block 1 is fed to the input of block 2 of the comparison. The second input of block 2 synchronously receives the signals of the reference image from block 4 of the memory of the standards. The information from block 4 is read at the address specified by the counter 13 of control block 3, and the reading and comparison is controlled by the sync pulse coming from generator 12 of block 3. Thus, the image information sequentially input to the input of comparator 2 corresponds to the reference image information sequentially coming from block 4 | in accordance with the change in the content of the counter 13 of block 3. When defects occur in the read image, a digital code appears at the output of the comparator 2, corresponding to the area of the defect. This signal enters multiplier 6 and is multiplied by a factor from the weight memory block 5. The reading from block 5 is conducted at the address specified by the counter 13 so that at the moment the result on the output of block 6 corresponds to the zone coefficients read from block 5. The multiplied results accumulate in block 7 for as long as this image is identified. The result of the accumulation of defects in block 7 enters the decision block 8. If the result of the summation of defects exceeds the threshold, the image under study is not identified, and vice versa.

The weights recorded in block 5 are chosen so that the signals from the defect

five

0

five

0

five

0

five

0

five

sites with high semantic significance, made large increments in the calculation of image defects. Conversely, the areas of defects located in the zones with the usual romantic significance contributed minimal increments.

The numerical values of the coefficients are selected as follows.

The reference image is interpreted as a numeric binary sequence. For the image in question (a numerical sequence), there are sequences (images) that have a minimum code distance, where n is the number of elementary partitions of the zone determined by the technical characteristics of the optical system. Code distance is defined as the number of mismatched positions in binary sequences. FIG. 4a shows a group of images having a minimum code distance. The areas in which these images do not match are assigned the values of the coefficients K 2N / l-n N a threshold value, which is chosen arbitrarily. Thus, these areas are assigned maximum weights. Next, the images that are separated from the considered one by the distance n are determined. Those zones that do not match in the group of selected images are assigned the values of the coefficients K 2N / 2 "n N / n, i.e. the coefficients in these zones are smaller. FIG. 46 depicts a group of characters having a code distance. Thus, according to the Formula, the values of the weighting coefficients of all zones of the image under consideration are determined after comparing the code distance with all other images. By a similar algorithm, all coefficients for all the studied images are calculated.

The introduction of new blocks and links, which allowed to take into account the semantic significance of various parts of the image, improves the accuracy of the device.

Claims (1)

Invention Formula A device for identifying text images containing a block the comparison unit, the memory of the standards, the decision unit and the television sensor whose output is connected to the information input of the comparator whose other input is connected to the output of the memory of the standards, the first output of the control unit is connected to the control input of the television sensor whose information input is the input of the device, the output of the decision block is the output of the device, characterized in that, in order to improve the accuracy of the device, it contains a weight memory block, a multiplier and a drip ivayuschy adder whose output is connected to the input deciding unit, an output the comparison unit is connected to the first injectivating input of the multiplier, the output of which is connected to the information input of the accumulating adder, the first output of the control unit is connected to the synchronization inputs of the comparison unit, the multiplier and accumulating adder These weights, the output of which is connected to the second information input of the multiplier, is the zero input of the control unit connected to the zero inputs of the multiplier and akaplivayuschego adder is input and the initial installation of the device. 3 W 2 figure 1 & A k ji tiff fa E 5 n Riga