JPH02268375A - Line segment determining method for machine drawing - Google Patents

Abstract

PURPOSE:To efficiently determine the coupling of line segments by modeling an area according to the condition of connection and the component of linkage with the other area of the line segment for the divided area, charging a limit condition, determining the coupling of the line segments according to the limit condition and propagating the coupling to the other area. CONSTITUTION:The machine drawing of a recognition subject is read by a scanner 11 and a binary picture is divided by the areas. Then, a result to classify the respective divided areas by feature quantity is inputted and held to the abstract slot of an area object in a node corresponding to the respective areas of tree structure information in a memory 15. A dimension line (d) is determined by using the information of an arrow in the area object. Next, by using the limit condition to the coupling of the line segments to be allowed in the area, the other line segment element of the area, for which the dimension line (d) is determined, is determined. The name of the line segment is propagated from the area, for which the dimension line (d) is determined, to the area connected to the above mentioned area. When the line segment element is determined, the other line segment element in the area is determined by using the limit condition and further propagated to the other area.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for determining line segments included in divided regions in a system that recognizes a mechanical drawing by dividing it into a plurality of regions.

[Conventional technology and its problems]

When a mechanical drawing recognition system recognizes a complex mechanical drawing, it is impossible to recognize the entire image of a given drawing at once. Therefore, the entire drawing is divided into a plurality of regions, each of which has a relatively simple structure, and recognition of the entire drawing is achieved by recognizing each of the divided regions. In this case, the entire drawing is generally first divided into two to four parts, as described in, for example, Japanese Unexamined Patent Publication No. 61-220077.

A so-called hierarchical division method is used in which the process of further dividing each area is repeated.

On the other hand, when recognizing mechanical drawings, it is important to identify the meaning of the line segments represented in one drawing (outline lines, dimension lines, extension lines, leader lines, etc.) in order to recognize the entire drawing. This is a clue. However, when determining the line segments included in each divided area, consider all possible combinations of line segments.

The combinations are vast and almost impossible.

An object of the present invention is to provide a method for efficiently determining combinations of line segments included in divided regions in a mechanical drawing recognition system that employs a region division method.

[Means to solve the problem]

In order to achieve the above object, the present invention models the divided regions based on the connections between the line segments and other regions and the connected components, imposes constraints on each model, and The method is characterized in that the combinations of line segments in a region are determined using the method, and the combinations are propagated to other regions.

[For production]

First, using the information of the arrow in the feature amount of a certain area,
Find a pair of arrows in other areas and determine the dimension line. Next, the name of the line segment is propagated from the area where the dimension line has been determined to the area connected to it, and the line segments are determined one by one using constraints. This allows most of the line segments to be determined. . In addition, for areas with undetermined line segment elements, leader lines are determined using constraint conditions and feature quantities such as bending points and end points.
What is necessary is to determine the dimension extension line, outline line, etc.

〔Example〕

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

FIG. 1 is a block diagram of an embodiment of a mechanical drawing recognition system targeted by the present invention. The scanner 11 reads a mechanical drawing to be recognized and stores the binary image in the image memory 13. The processor 12 has a similar configuration to a microprocessor or other data processing device, and uses the working memory 14 to divide the input drawing into regions, extract features of each divided region, model using the features,
Then, processing such as line segment determination, which will be described later, related to the present invention is performed.Here, it is assumed that area division is performed hierarchically. Each of the divided areas is managed by tree structure information in the tree structure table memory 15. Model dictionary memory 1
Reference numeral 6 stores information (referred to as a model) that abstractly expresses the shape of the divided region and its constraint conditions in advance.

FIG. 2 shows the correspondence between each area of the entire divided drawing and the tree structure information in the tree structure table memory 15. This is an example in which the tree structure information has an 8-ary tree structure, and one divided area is expressed by one node in the main structure. One division is, for example, one of 2 to 4, and only 2 to 4 of the 8 nodes at the same level are valid, and the other nodes are invalid (marked with an x). One node has the coordinates and feature amounts of, for example, the upper left end and lower right end of the corresponding region, and the regions created by dividing this region are expressed as its child nodes. The propagation order when determining division and line segments is realized by searching this tree structure information.

Here, it is assumed that the feature amount of the divided region has the following structure. This is called a region object. For each divided region, one region object instance is generated and recorded in the corresponding node.

Connection status of area object with other areas Value: ((Xu Yu) (Xr Yr) (X
d Yd) (Xi Yl)) Connection status of line segments Value: (nl n2 n3 n4) Line segment on which the arrow feather is placed and direction of arrow Value: (U/R/D/L +/-) Line segment End point value: Value of bending point of line segment; Abstraction value: Connection status with other areas (Xu Yu), etc. are the absolute coordinates of the representative point connected to other areas.

The connection status of line segments nl n2 n3 n4 corresponds to the top, right, bottom, and left of each area, and represents the number of connected components.

For example, (nl n2 n3 n4) is (1120), indicating that the upper and right line segments of the region are connected, and the lower component is not. For example, the line segment on which the arrow is riding and the direction of the arrow have a value of (R-), which is,
An arrow feather is placed on a line segment that connects to the right, indicating that the arrow points to the left, and a line segment having an end point is recorded at the end point of the line segment. As for the bending point of a line segment, a line segment having a bending point is recorded.

The above information is recorded at the time when the area is divided and the feature amount is detected. In the abstraction slot, the connection status of this area with other areas obtained as a result of feature extraction and the connection status of line segments,
Determine the corresponding model and enter its instance name.

The model is an abstract representation of the area, with only information about the connection status with other areas and the connection status of line segments in the 5-divided area. In this recognition system, dividing regions are integrated and a line name is determined by determining the name of a line segment in a model instance corresponding to the region.

FIG. 3 shows a concrete example of the model shape, where, for example, m
odel-1100 is connected to two orthogonal adjacent areas, and their line segments (in this example, the upper and right line segments) are connected within the area. In model-1200, they are not connected. Although omitted in Figure 3, there is more.

Rotate these clockwise, 0, 90, 180°270, respectively.
For example, a model rotated by 0.180 degrees is model-1010e and model-1020.

The models are hierarchically classified, first by the connection status with other areas, and then by the connection status of line segments, and are stored in the model dictionary memory 16 in advance together with their constraints. FIG. 4 shows an example of the relationship between a model and constraints (permitted combinations of line segments). In addition,
In Figure 4, the outline line is V, the dimension line is d, the dimension extension line is e,
The leader line is abbreviated as Ω.

FIG. 5 is a flowchart of an embodiment of the line segment determination method of the present invention, and the following description will be based on this.

Input The mechanical drawing for crab recognition is read with the scanner 11, and the second
The value image is divided into regions as shown in FIG. 2, and the results of classifying each divided region into one of the models shown in FIG. 3 based on the feature amount are input. This is already held in the abstract slot of the area object in the node corresponding to each area of the tree structure information in the tree structure table memory 15.

Filter l11 missing decision-: Determine the dimension line d using the information on the fletching of the area object. That is, in the region corresponding to the model, it is checked whether there is a corresponding arrow at the connection destination of the line segment on which the arrow is located. If there is, the line segment corresponding to the line segment on which the arrow feather is placed in the area is determined as the dimension line d. At this time, the direction in which the line segment on which the arrow feathers are connected to other areas.

The dimension line is determined only when a pair of fletching feathers can be found in another region depending on the direction of the fletching feathers.

Figure 6 Ca) shows this.

Next, other line segment elements in the area where the dimension line d has been determined are determined using constraints on the set of line segments allowed within the area.

The name of the line segment is propagated from the area where the dimension line d has been determined to the area connected to it. Once the values of line segment elements in the region are determined by propagation from other regions, use constraints to determine what other line segment elements in the region can be determined, and then propagate to other regions. . Propagation is dimension, @d, dimension assistance I
It is propagated in the order of JIAe and outline IIAv. This determines most of d, e, and v. FIG. 6(b) shows this.

of and of
:Finally, for areas with undetermined line segment elements.

Determination of leader line Q using constraints, bending points, and end points,
Determine the dimension line auxiliary line e and the outline line V.

LU: The line segments included in each divided area are converted into outline lines (including hidden lines).
The results of classification into V, dimension IIAd, dimension extension line (including center line) e, and leader line a are displayed.

〔Effect of the invention〕

As described above, according to the present invention, in a mechanical drawing recognition system that uses an area division method, determining line segments within and between divided areas is given more weight according to constraint conditions.

[Brief explanation of drawings]

Figure 1 is a block diagram of an embodiment of the mechanical drawing recognition system targeted by the present invention, Figure 2 is a diagram showing the correspondence between region division and tree structure information, and Figure 3 is an example of a region shape model. Figure 4 is a diagram showing an example of the relationship between the model and constraints,
FIG. 5 is a flowchart of an embodiment of the line segment determining method of the present invention, and FIG. 6 is a diagram showing a specific example of the process. 11...Scanner, 12...Processor, 13.
...Image memory, 14...Working memory. 15... Tree structure table memory, 16... Model dictionary memory. (b) Figure 3 (e)

Claims (1)

[Claims] (1) In a mechanical drawing recognition system that recognizes the entire drawing by dividing a drawing into multiple regions and classifying each divided region into one of the pre-prepared models based on its feature values, the information contained in each divided region is A method for determining line segments (outline lines, dimension lines, extension lines, leader lines, etc.), which applies constraints determined by the connections of line segments to other regions and connected components within the region to each model. First, determine the dimension line using the arrow information in the feature amount of the area,
Next, the line segment determination for mechanical drawings is characterized in that the line segments are propagated from the area where the dimension line is determined to the area connected to the area, and the line segments in each area are sequentially determined using constraint conditions. Method.