JPS59114681A - Form recognizing system - Google Patents

Abstract

PURPOSE:To improve the reliability with recognition of forms by introducing a recognition discrimination function EX in terms of a vector quantity in order to reduce the recognition errors. CONSTITUTION:The i-th feature parameter value xi (x1: area; x2: circumferential length, x3: centroid, etc.) is measured to a form pattern which is binary coded by an input processing of pictures. Then a recognition discrimination function is obtained with reference to the mean value -Si of the standard feature parameter stored previously, the standard deviation DELTASi and the decision standard value Ui respectively. Then a decision is carried out with said recognition discrimination function, and the incapability of recognition is decided unless the vector is set at zero. Thus the processing is carried out for the recognition error. While it is decided that the desired conditions are satisfied for recognition with the zero vector. Then the processing is done after recognition. Then a resemblance discrimination function E is used as it is for recognition of position of a form pattern when the recognition conditions are satisfied.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Application The present invention relates to a shape recognition method for recognizing the shapes of characters and objects.

Configuration of conventional examples and their problems Conventionally, on production lines, etc., for shape patterns of letters, industrial parts, luggage, etc., classification 2 Positioning, inspection, etc. are performed using characteristic parameters of shape patterns, such as area 2 perimeter 2 center of gravity There is a feature extraction method that identifies a target pattern by measuring the characteristics of the object, etc., and comparing it with pre-stored standard data. However, when the conventional method compares the feature parameters of the target pattern with standard feature parameters stored in advance, the problem is that the misrecognition rate cannot be reduced because the comparison is limited to an integral comparison of scalar quantities. be.

OBJECTS OF THE INVENTION An object of the present invention is to provide a recognition method that achieves a reduction in the false recognition rate in shape pattern recognition in order to solve the above-mentioned problems.

Composition of the Invention The present invention provides a first image of the shape pattern by rotating the attitude angle of the shape pattern detected by a pattern input device within a desired range or by changing observation environment conditions such as illumination conditions within a desired range. The feature parameter value St
(For example, Sl is the area.

S2 is the perimeter, S3 is the center of gravity, etc.), and from the measured feature parameter value Si, its average value SL and standard deviation △
Si is determined, and the measured feature parameter value X of the target pattern to be recognized is determined based on the standard feature parameter in which the determination reference value Ui corresponding to the above feature parameter is set.
For i, find the similarity discriminant function E specific to the shape pattern, and measure the similarity between shapes from the distance on the projected 1-Ei plane on the 1-Ei-F spatial coordinates, This method performs two classifications to identify the shape, and by introducing a recognition discriminant function Ex, which is a vector quantity, to further reduce misrecognition, it has the effect of improving the reliability of shape recognition.

DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to FIGS. 1 to 3. In FIG. 1, a lighting device 3 and a television camera 4 are installed to input the pattern of a target object 1 placed on a conveying means 2 such as a belt conveyor. Here, the television camera 4 is controlled by a camera control circuit 6. The video signal input by the television camera 4 enters an analog-to-digital conversion (hereinafter referred to as A/D conversion) circuit 6, where it is binarized by a binarization control circuit 7 and converted into a C
A microcomputer (CPU-y-) consists of a PU, ROM, RAM, input/output ports, etc.

MN1610. or MN1613).

The pattern recognition device is a judgment control circuit, that is, a CPU 5, to which commands are given from the main controller or operation panel.
, memory control circuit 9. Standard feature parameter memory circuit 1
0. Sample feature parameter circuit 11 calculated based on input cover turns. It is composed of a similarity or recognition discriminant function calculation circuit 12 and an identification circuit 13. The recognition result is sent to the main controller.

By the way, the similarity discriminant function e for the feature parameter i
, , and their frequency distribution f are shown in FIG. 2 as a conceptual diagram expressed as an i-Ei-F space.

Next is a pattern input device configured as described above.

and the operation of the recognition device will be explained.

As shown in FIG. 1, when the recognition object 1 is transported by the transport means 2 into the effective imaging range (effective field of view) of the television camera 4, and the illumination by the illumination device 3 is working effectively,
When a start command is given from the main controller or the operation panel, recognition is started according to the procedure of the shape recognition program using similar shapes shown in the flowchart of FIG.

First, as shown in the flowchart of FIG. 3, in step 1), the feature parameter value Xi (i.e. x
l is the area, X2 is the circumference, X3 is the center of gravity, etc.) in Step 2), referring to the average value Si, standard deviation △St, and judgment reference value Ui of the standard feature parameters stored in advance, or , where, find the similarity discriminant function given by (step 3
), then find the recognition discriminant function given by (Step 4
)O Next, the above recognition discriminant function EH is determined (step 5).
, if it is not a zero vector, it is determined that it is unrecognizable and recognition error processing is performed (step 7), and if it is a -1 zero vector, it is determined that the necessary conditions for recognition are satisfied and post-recognition processing is performed (step 6). Here, if the recognition conditions are satisfied, the above similarity discriminant function E is used as is for position recognition of shape patterns, etc., and on the other hand, for shape pattern classification, the product type with the minimum similarity discriminant function E is adopted. . With the above steps, shape recognition is completed. Note that the results are sent to the main controller.

Effects of the Invention - According to the present invention as described above, it is possible to reduce recognition errors in shape recognition due to similar shapes of shape patterns. Also,
Since it is possible to set a judgment standard value for each feature parameter,
It becomes possible to respond to so-called different consciousness filters for each subject.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an example of a means for shape recognition based on an embodiment of the present invention, FIG. 2 is an explanatory diagram regarding i-Ei-F spatial coordinates, and FIG. 3 is a program for shape recognition. It is a flowchart figure which shows an example. 1... Recognition target, 2... Transfer table, 3... Lighting device, 4... TV camera, 5... Camera control circuit , 6...A
/D conversion circuit, 7...Binarization control circuit, 8...
...Judgment control circuit, e...Memory control circuit, 10...Standard parameter memory circuit, 11.
...Sample parameter memory circuit, 12...
. . . Discriminant function calculation circuit, 13 . . . Discrimination circuit. Name of agent: Patent attorney Toshio Nakao and 1 other person No. 1
Figure 2

Claims (1)

[Claims] A pattern input device detects a shape pattern to be recognized, and then a recognition device detects a first shape pattern for the shape pattern.
The feature parameter value St is measured by changing the input conditions of the shape pattern, and the measured feature parameter value S
From the measured feature parameter value Xi of the shape pattern, for a group of N standard feature parameters for which the average value St and standard deviation ΔSi are calculated from t, and the determination reference value σi corresponding to the feature parameter is set, it is given by Find the similarity discriminant function given by , and find the recognition discriminant function given by , and if a zero vector such as is established as a necessary condition to satisfy the recognition discriminant, then the above similarity discriminant function E
, for i− normalized using the probability density given by F,=f(i,E,)
E, -F Shape recognition is performed on the spatial coordinates by comparing the similarity discriminant function of the standard pattern with the similarity discriminant function of the shape pattern to be recognized, which is determined based on the values of Si, ΔSt, and xi. Shape recognition method.