JPH02250187A - Detection of false flection point - Google Patents

Abstract

PURPOSE:To prevent error of false flection point detection by extracting a section where edges are linearly continuous and discriminating the truth or the falsehood of a flection point in the section in accordance with the presence or the absence of a corresponding edge on another orthograph. CONSTITUTION:The section (linear edge continuity section) where an arbitrary number of edges which are probably integrated to one edge roughly are lineary continuous is extracted from each of three orthographs by a linear edge continuity section extracting part 9. A flection point (apex) in the extracted linear edge continuity section is judged to be true or false based on not local information but correspondence relations among orthographs. That is, it is checked whether the edge corresponding to each flection point in the section exists on another orthographs or not by a flection point truth/falsehood discriminating part 13 to discriminate the truth or the falsehood of the flection point. Thus, true flection points and false flection points which are scarecely different in existence condition when locally viewed are discriminated, and false flection points are surely deleted and erroneous deletion of true flection points is prevented, and the performance of the fair copy making processing of a hand- written drawing or the like is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a method for detecting false bending points that should not exist on each side view, in a handwritten three view drawing support system, etc.

[Conventional technology]

A system is being researched that reads a handwritten three-view drawing using an image scanner and outputs the fine-printed drawing using an output device such as a laser beam printer.

In such a three-view drawing support system, each hand-drawn drawing is read into the system by an image scanner, and this drawing data is in the form of image data that is a collection of dots. In addition to the original information on the drawing, this image data includes quantization errors depending on the accuracy of the image scanner.

Vectorization processing is performed on image data that contains such quantization errors, but at this time, due to errors (noise) depending on the vectorization algorithm, what is originally a single edge often turns into multiple edges. is converted into an edge, and a ``false vertex J'' that should not exist is generated.

An example of this is visually shown in FIG. The figure intended by the designer has four edges and four vertices (R) as shown in (a).
Assuming that it is a square consisting of , many false vertices (P) occur in the vectorized figure as shown in (b). Note that the true vertex (R) is almost never lost.

The false vertex (third
In the example shown in the figure, just by eliminating the curved points (the true vertices before and after are connected by straight lines), you can obtain a drawing that visually gives a fairly clean impression.

Conventionally, such false vertices are detected and deleted as follows. An arbitrary number of consecutive edges after vectorization is tracked, and when the sum of the directional fluctuations of the edges exceeds a predetermined threshold, tracking is stopped, and the edges tracked up to that point are integrated into a single edge. That is, a vertex that is being tracked is considered to be a false vertex and is deleted.

An example is shown in FIG. The continuous edges shown in (a) are traced from the left end to the right, and the sum of the "include angles" θ at each vertex (bending point) is calculated. If the sum of the included angles (absolute value) up to the right edge of the diagram is less than a predetermined threshold, and the sum of the included angles (absolute value) of the next edge exceeds the threshold, the consecutive edges from the left edge to the right edge of the diagram are integrated into one edge as shown in (b), and the five bending points in the middle are deleted as false endpoints.

[Problem to be solved by the invention]

However, this method of determining the truth or falsity of a vertex based only on local information may cause errors such as deleting true vertices as false vertices, or conversely leaving false vertices as true vertices. The problem is that it is easy to occur.

This will be explained with reference to FIGS. 5 to 7. All of them are three-view views after being converted into a bag.

Vertex (bending point) of the nearly horizontally continuous edge in the front view (b) of Fig. 5 pl-+ p2r P 3+p4
+ Consider the truth/false judgment of p5. The bending of the edge is emphasized here, but since edge A in plan view (a) exists, plr p 2 + p 3 + p5 is deleted as a false bending point, and only p4 is treated as a true bending point. I would like to leave it as is and hope for a clean copy result like the one shown in Figure 6.

That is, it is expected that a fair copy will be made as a three-dimensional three-dimensional view as shown in the perspective view (d) of FIG.

However, according to the conventional method, all bending points from pl to p5 are erroneously determined to be false bending points and deleted, which may result in clearly incorrect cleanup results as shown in FIG.

Although we have explained here the case of fair copying of handwritten drawings, similar false vertices occur during input processing in systems that recognize printed drawings and machine-generated drawings, so similar false vertices are generated as preprocessing for recognition. A similar problem exists with hand-drawn drawings, which are performed by detecting and deleting vertices.The purpose of the present invention is to provide a method for detecting false bending points that can reliably prevent such errors. It's about doing.

[Means and effects for solving the problem] In a three-view diagram, the projection directions of each projection plane are usually orthogonal to each other, and the edges of an object perpendicular to a certain projection plane (drawing) are one on that plane view. It is projected (drawn) as a point, but it should be projected (drawn) on the remaining two views as an edge parallel to a specific coordinate axis of the view.

For example, the coordinate system of each side view is y-x for the front view and z for the top view.
-x, and the side view is y-z, then the object's z 1lill
An edge parallel to l should be drawn as a vertex (intersection or inflection point) in the front view, and as an edge parallel to the Z axis in the top and side views. In other words, if the vertex of a certain input plan view is a true vertex, it should have a corresponding edge on another plan view.

According to the present invention, first, a section in which an arbitrary number of edges that are likely to be roughly integrated into one edge are linearly continuous (linear edge continuous section) is extracted from each view of the three-view diagram. . Next, the authenticity of the bending points (vertices) in the extracted linear edge continuous section is determined, but instead of using conventional local information as the basis for this determination, the correspondence between the above-mentioned surface views is used. do. That is, the authenticity of each bending point is determined based on whether edges corresponding to each bending point in the section exist on other views.

〔Example〕

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

FIG. 1 is a block diagram showing an example of a functional configuration for three-view fair copy processing to which the method for detecting false bending points of the present invention is applied, and FIG. 2 is a schematic flowchart of the processing. Below, the second
The processing contents will be explained for each processing step in the figure.

Processing Step ■ The image scanner 1 creates each side of the three-view drawing (this can be a handwritten drawing, a printed drawing, or a machine-generated drawing)
is read as binary image data, which is a collection of dots, and stored in the storage unit 2.

Processing step (2) This image data is subjected to vectorization processing by the vectorization section 3, and the image of the input drawing is converted into a line figure with a line width of one dot. The solid data (coordinates of both end points of the vector) obtained by this back-to-back processing is stored in the storage unit 4.

Processing step (2) From this vector data, graph information expressing the connection relationship between edges and vertices is created by the graph information generation section 5 and stored in the storage section 6. This graph information represents a graph corresponding to the input drawing using vectors obtained through vectorization processing and their end points.

Processing Step (2) The intersection recognition unit 7 detects the intersection of the graph from the graph information, and the intersection information is stored in the storage unit 8.

Processing step ① Extract continuous straight edge sections from the graph representing the input drawing (each side view) obtained by back-processing.

This is performed by the linear edge continuous section extraction unit 9,
The data of the extracted linear edge continuous section is stored in the storage unit 10.
is memorized. At the same time, a "clear bending point" is detected and its data is stored in the storage unit 11.

Here, a method for extracting continuous straight edge sections will be explained with reference to FIG.

Assume that a graph as shown in FIG. 8(a) is obtained.

pl to pl2 are bending points, cl and C2 are intersection points (branching points)
It is.

First, focus on a suitable intersection. Assume here that we are focusing on C1.

Next, arbitrarily select edges that constitute this intersection. Here, it is assumed that edge e1 is selected.

Using this edge e1 as a starting edge, successive edges are tracked one after another in the direction opposite to the intersection c1. In this example, tracking is performed in the direction of the arrow in the figure. Tracked edges are given tracked information. One tracking procedure ends when a new intersection is encountered or when the vehicle returns to the starting intersection (cl).

In this example, starting from the intersection c1, el, C2,...C8
The edge is tracked, and one tracking procedure ends when it hits the intersection c2.

In this edge tracking, each time one edge is tracked, the included angle between that edge and the previous edge is calculated, the sum of the angles is calculated, and its absolute value is compared with a predetermined threshold t1. This threshold value t1 is determined by the threshold value setting unit 1 according to the roughness of the input drawing.
2 (FIG. 1).

When the absolute value of the sum of included angles exceeds the threshold t1, the vertex that is the center of the last incorrect included angle is stored in the storage unit 11 as a "clear bending point", and the total sum of included angles is Initialize to O and restart edge extraction and calculation of the sum of included angles in the same manner.

In this way, the edges from the edge where tracking starts to the "clear bending point" are extracted as one linear edge continuous section, and the edges from there to the next "clear bending point" or the end of tracking are linear edges. Extracted as a continuous interval.

For example, if the included angle from cl to p4 is a value as shown in FIG. is detected as a “clear inflection point” and cl
The period from to p3 is extracted as one continuous linear edge section, the sum of included angles is initialized, and tracking continues in the direction of edge e5 with edge e4 as the starting edge.

In this way, in the case of the graph (a) in Figure 8, (b
), seven linear edge continuous sections from ■ to ■ are extracted.

In the extracted continuous straight edge section, there are an arbitrary number of bending points that are neither "clear bending points" nor branching points, and whose truth or falsity is unclear.

Regarding the bending points in each linear edge continuous section on each plan view stored in the wheat star pattern storage unit 10, the bend point authenticity determination unit 13 determines the correspondence with edges on other plan views. By examining it, the authenticity can be determined. The graph information changing unit 14 performs a change process on the graph information in the storage unit 6 to delete the detected false bending point and linearly connect and integrate the edges excluding the edge.

The authenticity determination of the bending point is performed as follows. For example, in the front view (b) of FIG.
Suppose that two edges are extracted as a linear edge continuous section.

If this bending point P is a true bending point, there should be a corresponding edge approximately parallel to the Z axis in the coordinate system (z-x coordinate system) of the plan view (a).

Therefore, let the X coordinate of the bending point A be Ax, and A x −t2
It is checked whether there is an edge or continuous straight edge section that is approximately parallel to the Z axis and has an X coordinate in the range from Ax+t2 (t2 is the threshold set by the threshold value setting unit 15). In this example, since edge A corresponding to the plan view (a) exists, bending point P is determined to be a true bending point. However, if a corresponding edge like this edge A does not exist, the bending point P is determined to be a false bending point and is deleted.

In another example, from pl to p in the plan view (b) of FIG.
Assume that an edge group including 5 is extracted as a linear edge continuous section (threshold value t1 is selected so that such an edge group becomes a linear edge continuous section). In this case, the existence of a similar corresponding edge is checked for each of the inflection points p1 to P5, and only the inflection point p4 is left as a true inflection point because there is a corresponding edge A (or a straight edge continuous section), and the remaining The bending point P 1+ P 2+ p3
゜p5 is deleted as a false bending point, and the drawing represented by the changed graph information (the drawing after the fair copy) is shown in Figure 6 (2)
become that way.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, it is possible to identify true bending points and false bending points, which have almost no difference in their existence status when viewed locally, and to ensure deletion of false bending points. It is possible to prevent erroneous deletion of true bending points, and it is possible to improve the performance of fair copying of handwritten drawings and the like.

[Brief explanation of drawings]

FIGS. 1 and 2 are block diagrams showing an example of the functional configuration for trihedral drawing processing to which the false bending point detection method of the present invention is applied, and a schematic flowchart of the processing, and FIG. Diagram explaining the occurrence of false vertices,
Figure 4 is a diagram explaining the conventional false bending point detection method, Figures 5 to 7 are diagrams explaining the authenticity judgment of the bending point on three sides and the influence of false judgment, and Fig. 8 is a diagram explaining the influence of the false judgment on the three-sided side. It is a figure explaining extraction of a continuous section. DESCRIPTION OF SYMBOLS 1... Image scanner, 2... Image data memory|storage part, 3... Back 1-file conversion part, 4 - Vector information memory|storage part, 5... Graph information generation part, 6... Graph information memory Part, 7. Intersection detection part, 8.
- Intersection information storage unit, 9... Straight edge continuous section extraction unit, 10... Straight edge continuous section storage unit, 11... Clear bending point storage unit, 13... Turning point truth/false determination unit , 14...Graph information changing section. Figure 2 (Shin Figure 5 (α) Thousand-faced funeral map (o: J?i1 Section procedural amendment band (method) 1. Indication of the incident 2. Name of the invention by Tian Wen Yi 1988 Patent Application No. 300051 Detection method of false bending point 3, relationship with the case of the person making the amendment Applicant Address 1 Nakamagome, Ota-ku, Tokyo Name (674) Co., Ltd. Representative 3-6 Hamada-chome Rico - Hiroshi 4, Agent 7. Contents of the supplement (1) Item “2” between the third and fourth lines of the first page of the specification.
, "Claims" shall be inserted. (2) Item 11r2, "Claims" on page 1, line 8 of the specification, is amended to "3. Detailed description of the invention." 5. Date of amendment order: March 29, 1990 (shipment date: 6, specification subject to amendment: April 24, 1990) I

Claims (1)

[Claims] (1) A false method characterized by extracting a section in which edges are linearly continuous in each view of a three-view diagram, and determining the authenticity of a bending point within the section based on the presence or absence of a corresponding edge on another view. How to detect inflection points.