JPH04141788A - Device and method for thinning line distortion form automatic corrector - Google Patents

Abstract

PURPOSE:To accurately correct a road center line even when roads with different road widths coexist by extracting a pair of line segments which forms a pair with the line segment obtained while connecting the road center line with the crooked points of the contour of the road and correcting it while utilizing the center line of a confidence interval. CONSTITUTION:A road center line thinning line means 2, a whisker elimination means 3 for angular part, a confidence interval extraction means 4, a terminal part correction means 5, an angle correction means 6, an estimation section extraction means 7, a line arc searching means 8, a T-road correction means 9, a Y-road correction means 10, and a bend point reduction means 11 are connected to a CPU 1. In this case, the connection information between arcs is extracted, a pair of line segments which forms a pair with the line segment obtained while connecting the crooked points of the center line of the confidence interval, and obtains the center line which is the parts of the terminal point, angle, branching point, and intersection. Thus, the correction can be accurately performed even when roads with different road widths coexist.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to an apparatus and method for correcting distorted shapes in image processing, and in particular, an apparatus and method for automatically correcting road center lines that can create accurate road center lines. Regarding.

[Conventional technology]

Conventionally, the following methods have been used to thin the raster data of such a line drawing, convert it into vectors, and correct the thinning shape distortion.

This will be explained with reference to FIGS. 52 to 57. Referring first to FIG. 52, dashed lines 110 indicate road areas;
FIG. 52 shows a road center line 111 obtained by thinning the road area and converting it into a vector by polygonal line approximation. Here, a method for obtaining a vector figure from a thin line image by polygonal line approximation will be explained. Referring to FIG. 55, the vector data is shown by thinning the pixel P3, which is represented as a black dot, and the pixels P1 and P2, which are represented by white dots. First, to obtain vector data by polygonal line approximation, find the point P as a branch point,
The pixel is traced from that point P1 to find the next end point P, and a straight line 112 connects the pixel P and pixel P. Point P and end point P
2 is found, and if the distance is 1.5 pixels or more, point P3 is determined as the refraction point.

Next, perform similar processing on each of the two line segments connecting point P and both endpoints P and P2, and repeat the processing until the distances between the approximate vector and each pixel are all 1.5 pixels or less. repeat.

In vectorization, we adopted a method to obtain vectors with small approximation errors from pixels, but since distortion occurs in the figure in the original thinning process, even if the thin line image is faithfully vectorized, the shape may differ from the center line of the original line diagram. It may become. The centerline correction process corrects this so that it becomes the centerline of the original diagram.

Here, the explanation will be given again by returning to FIG. 52. As mentioned above, when a thin line image is faithfully vectorized, distortion occurs in the road centerline. In FIG. 52, areas indicated by circles 112, 113, and 114 are portions with distortion of the road center line.

The following methods are available for automatically correcting these center lines. In other words, a certain part of a line segment with a certain length or less is considered to be distorted and ignored, and the shape of the ignored line segment is restored from the shape of the line segment in its vicinity. .

Another simple method for correcting the thinning shape distortion that appears in the thinning results and finding the centerline of a line figure is a centerline correction method using an interactive processing function on a computer.

[Problem to be solved by the invention]

The first method of the prior art described above will be explained by taking as an example the case where the image of the line figure shown in FIG. 52 is corrected. According to the first method, when the road width is constant, the length of the line segment to be ignored can be corrected fairly well by setting it to an appropriate value less than the road width, but when roads with different widths coexist, If the length of the line segment to be ignored is set to a large value, all line segments below a certain length will be erased, resulting in a figure as shown in FIG. In this case, there is a problem in that all the vectors within the circle mark 114 are also erased, and finally the road center line at the branch point portion 122 is erased.

On the other hand, if the length of the line segment to be ignored is set to a small value, as shown in FIGS. There is a problem in that the center line of the branch point remains uncorrected.

Conventional methods of automatically correcting line segment lengths are estimated based only on the thinning results, so
If the road widths are different, you may end up correcting line segments that should not be corrected.

Further, manual correction using an interactive processing function with the CPU requires a great deal of effort and there is a risk of omission of correction, so automatic correction is desired.

The present invention has been made in view of the above-mentioned problems, and its purpose is to solve the above-mentioned problems by using contour information of original figures without requiring a lot of effort such as manual correction. An object of the present invention is to provide a thinning distortion shape correction device and method.

[Means to solve the problem]

The present invention, which has the above object, is an automatic thinning distortion shape correcting device for correcting the distortion of a thinned line figure of an arbitrary thickness and determining the center line of the line figure, Connection information extraction means for extracting polygonal lines separated by points and intersections and not including other junctions or intersections between the junctions and intersections, and for extracting connection information between the polygonal lines; A line figure bending point extracting means for tracing and extracting bending points of a line figure, and line segments obtained by connecting bending points in the traced order, on both sides of the polygonal line obtained from the thinning result, A confidence interval extracting means for obtaining a confidence interval made of a rectangle inscribed in a pair of line segments, correcting the shape of a polygonal line within the confidence interval using the coordinates of the line segment set, and obtaining a center line of the confidence interval. , a prediction interval correction means for determining a prediction interval consisting of a rectangle circumscribing a set of line segments, and using the prediction interval to modify the shape of a polygonal line at a bending point portion of a line figure, based on connection information between the polygonal lines, It consists of a bifurcation point correction means that corrects the shape of the connecting portion of the polygonal line using the center line of the confidence interval and obtains the center line corresponding to the bifurcation point or intersection portion of the line thinning result.

According to another invention, distortion of a thin line figure is corrected,
A method for finding the center line of a line figure, in which a thinned line figure is divided at its branching points and intersections, a polygonal line that does not include other branching points or intersections is extracted between the branching points and intersections, and the polygonal line is Extract the connection information between each other, trace the outline of the line figure, extract the bending points of the line figure, connect the bending points in the traced order, and among the line segments obtained from the thinning results. Extract a pair of line segments on both sides of the line segment, find a confidence interval consisting of a rectangle inscribed in the set of line segments, and calculate the shape of the line segment inside the confidence interval using the coordinates of the set of line segments. correct it, find the center line of the confidence interval, find a prediction interval consisting of a rectangle circumscribing the set of line segments, use the prediction interval to correct the shape of the polygonal line at the bending point of the line shape, and obtain connection information between the polygonal lines. Based on this, the shape of the connecting part of the polygonal line is corrected using the center line of the confidence interval, and the center line corresponding to the branch point or intersection part of the line thinning result is determined.

[Effect]

According to the present invention, the thinning result of a line figure of arbitrary thickness is divided at its branch points and intersections, and a polygonal line that does not include any branch points or intersections is extracted between the branch points and the intersections. Connection information between the polygonal lines is extracted. At the same time, the outline of the line figure is traced and the bending points of the line figure are extracted. Among the line segments obtained by connecting the bending points in the order of tracing,
A pair of line segments on both sides of the polygonal line obtained from the thinning result is extracted. A confidence interval consisting of a rectangle inscribed in a set of line segments is determined, the shape of a polygonal line within the interval is corrected using the coordinates of the set of line segments, and the center line of the confidence interval is determined. Next, a prediction interval made of rectangles circumscribing the set of line segments is obtained, and the shape of the polygonal line at the bending point of the line figure where the prediction intervals overlap is corrected using the center line of the confidence interval. Then, based on the connection information between the polygon lines, the shape of the connection part of the polygon lines is corrected using the center line of the confidence interval, and the center line corresponding to the branch point or intersection part of the line thinning result is obtained.

In this way, a pair of line segments is extracted using the line segments obtained by connecting the road center line to the bending points of the road outline, and the center line of the confidence interval is used to correct the road center line, so roads with different widths are mixed. To correct a road center line accurately, without requiring much labor, and without omitting any corrections even when the road center line is corrected.

〔Example〕

Hereinafter, the present invention will be explained based on illustrated embodiments. 1st
The figure shows a block diagram of the road center line automatic correction processing device of the present invention. The road centerline automatic correction processing device has a CPUI. A road center line thinning means 2 is connected to the CPUI. The road center line thinning means 2 thins the pixel string representing the road until it becomes one pixel. Here, in order to leave dead-end road sections, the line thinning algorithm produces a single vector at the road corner that is deviated from the road center line, a so-called corner whisker. A corner whisker removal means 3 for removing vectors forming the whiskers is connected. In addition, we extract a rectangular section inscribed in the block parallel area necessary to correct the road center line, a so-called confidence interval,
Confidence interval extracting means 4 for obtaining the road center line within the road is connected thereto. Furthermore, the CPU includes an edge correction means 5 for correcting the road edge, a corner correction means 6 for correcting the road center line at the corner, and a prediction section for extracting a prediction section consisting of a rectangle circumscribing the parallel part of the paired city block. Extraction means 7, intersection
straight arc search means 8 for searching the road center line of a branch road;
A T-junction correction means 9 for correcting T-junctions, a Y-junction correction means 10 for correcting Y-junctions, and a bend point reduction means for processing each bend point are connected.

Next, based on Figure 2, we will discuss the thinning of the road center line in the second section.
This will be explained based on the flowchart shown in the figure.

First, a road raster is obtained by filling in the road area other than the city block. After that, the road is thinned and vectorized, and the vector is determined using the least squares approximation method. The image after the road area is filled in is formed as a screen as shown in FIG. The white line in the center of the road is the result of thinning the road. In FIG. 3, IA represents a road area, and 1B represents a block area. A typical method for thinning the image is the Hildlitch method. Then, find the vector using the least squares approximation method.

A method for determining a vector using the least squares approximation method will be described with reference to FIG. 6, using an example of determining an approximate straight line l for a curve P.

According to this method, first, the area S of a predetermined region formed between the curve P and the straight line l is determined by integration, and the straight line l is determined such that the calculated value is smaller than the maximum allowable error Emax.
seek. In this way, the curved portions of the road center line in the entire screen area are replaced with approximate straight line portions. The screen after replacement is shown in FIG. 4. Then, the bending point of the road center line is the bend point VE, and the branching point or intersection is the node N.
Expressed as These are represented as black dots and added to the screen of FIG. 4, which is shown as the screen of FIG.

With reference to FIG. 7, the removal of whiskers from corners will be explained (step 101). Reference numeral 10L represents a vector between a node and a bend point, which forms a whisker at a road corner. below,
The removal of this beard will be explained in detail. First, a polygonal line or arc having end points is determined. In the example shown in FIG. 7, the arc 1oL is the target. Here, an arc refers to a series of vectors between a branch point and an intersection, or a series of vectors from an end point to a branch point. Next, arc 10L
Delete arcs with endpoints whose arc length is less than or equal to 0.5. Here, 0.5 mm corresponds to the narrow road width.

Therefore, arcs that are smaller than the narrow road width are targeted for deletion. In this case, the arc 10L is 0.5- or less, so it is deleted. After the arc 10L is deleted, the arc 10a above the bend point P and the arc 10b below the bend point P are separated in the data, so the branch point P is set as the bend point, and the lower end point of the arc 10a and the arc 10b Data processing is performed to match the upper end points of the two arcs 10a and 1.
Combine 0b into one line. As a result, arc I in Figure 7
The OL is deleted, resulting in a figure forming one road centerline as shown in FIG.

Next, confidence interval extraction of the thinning result will be explained with reference to FIG. 9. The confidence interval here refers to a rectangular interval inscribed in parallel city blocks. To extract the confidence interval, first, for each inter-bend vector 12 of the road center line, the block bend point vector lla having the shortest distance from each bend vector 12 of the road center line is determined. Next, if the vector lla between the block bend points and the vector 12 between each bend point on the road center line are parallel, the shortest block bend point vector lla
The other block vent point vector 13 (
Find the city block vs. vector).

Here, the block pair vector 13 sets the distance between the block sections at a position symmetrical to the shortest block with the road center line 12 as an axis within the distance between the road center line and the block section distance (d)X2+0.5W.

If the two block pair vectors 11 and 13 are parallel, a rectangular section inscribed in that section is defined as a confidence interval F. To obtain the confidence interval F, a perpendicular line is drawn from the block bend point F1 of the block bend point vector 11 to the block pair vector 13, and an intersection point F2 between the perpendicular line L1 and the block pair vector 13 is determined. Furthermore, draw a perpendicular line L2 from the bend point F of the block pair vector 13 and F1
a, and the rectangles F, F, , F3, F formed by the obtained intersections F and F and the bend points F1 and F are
4 becomes the confidence interval F.

Thereafter, the center line bend points within the confidence interval F are deleted, and a bend point is set at the midpoint of the perpendicular line L and the perpendicular line L2.

To explain with reference to FIG. 12, in the road area in FIG. 12, typically an area F surrounded by a rectangle is
SF F SF, F1112ゝ! 3 1
4 15' FF, FF or confidence interval (S16ゝ 171
8ゝ 19 chip 102).

After that, in the confidence interval within the entire predetermined area, a new bend point is set on the road center line and the road center line is corrected, resulting in the drawing shown in FIG. 13.

The shape of the road center line before and after the confidence interval F is extracted and corrected will be explained with reference to FIGS. 10 and 11.

As shown in FIG. 10, the road center line 12 between the blocks 11a and 13 before correction does not accurately indicate the center of the road between the blocks 11a and 13. After correcting this as described above, the road center line 12 becomes as shown in FIG. 11.

In this way, the intersection points of the center line of the rectangle formed by F SF, , F SF4 and the perpendicular line L1L2 are respectively set as new vent points 12.12N, and the vent points between them are erased, as shown in Fig. 11. The road center line shown in is obtained.

Next, with reference to FIGS. 14, 15, and 16, the correction of the edge of a road having end points will be explained (Step 1
03).

This focuses on arcs with end points and corrects road edges.

First, the corrected road center line within the confidence interval of arc 21,
Alternatively, find the intersection of the road extension and the city block.

To explain using FIG. 14, in FIG.
4 is shown and the arc 2 indicates the road center line 24 at the road edge.
1 end point 25 is shown. First, the confidence interval F of the arc 21 is found, and the road center line corrected using it is extended toward the road edge to find the intersection S with the city block. 1st
In Figure 4, the confidence interval extension E is represented by a dotted line. Next, the distance I3 between the road edge and the previously determined intersection S is determined, and if the distance is less than 0.5W, the position of the end point 25 of the arc 21 is moved to the position of the intersection S.

0.5- corresponds to a narrow road width. Here, set the amount of movement to 0
．． The reason for setting it within 5- is to prevent the following erroneous corrections.

The reason for this will be explained with reference to FIG. 16. For example, suppose there is a road shape as shown in FIG. In such a road shape, a road center line 24 is formed between the block 20 and the block 22.

The road center line 24 branches at a branch point 23a, and the road center line 24 branches at a branch point 23a.
An arc 24 extends from 3a, and end points 25 of the arc 24 are arranged as shown. Inside confidence interval F is arc 24
is corrected linearly.

In this case, unless there is a condition to move only arcs within 0.5-, the arc end point 25 is moved to the point S where the confidence interval is extended and intersects the road block 30. In this case, the bending part (2) of the arc 25 is erased, and the center line inside the bending part V of the road is erased, resulting in erroneous correction.

In order to prevent such erroneous corrections, it was decided to move only the arc within 0.5W.

Next, the prediction interval will be explained using FIG. 17. The prediction interval refers to a rectangular section that circumscribes the four block bend points (■, ■, V 1v4) of the block bend point vector 27 and the block bend point vector 28 that constitute the block pair vector. First, the parallel portions of the vector 27 between city block bend points and the vector 28 between city block bend points are extended. And vector 27 between city block bend points, vector 28 between city block bend points
Perpendicular lines are drawn from the bend point v 1v to the extensions of the vectors 27 and 28 between the opposing confidence interval block bend points, respectively. An intersection point v ′, v ′ is formed at the intersection of the perpendicular line and the extension of the vector 27.28 between the block bend points, and a rectangular section formed by v 1 , v ′, and v 4 including this intersection point is predicted. It is specified as an interval (step 104).

First, the road center line at the corner is corrected using this prediction interval. In the case of a corner where the prediction intervals intersect, the intersection S of the center lines A1B of the two prediction intervals is set as the new bend point V, as shown in FIG. Next, the case of a corner where two prediction intervals do not intersect will be described with reference to FIGS. 20 to 23.

In that case, both ends of the arc of interest are virtually connected with a straight line segment, and the farthest bend point from that line segment is found, and the distance is an error value of ε-0.14 mm (=just over 2 pixels cf).

If the width is 0.5 mm or more, it is divided into two line segments at that point. Within block 27 and block 28, two confidence intervals A
and B. The general method of linear approximation between the end points P1, P of the confidence interval A and the end points P1P4 of the confidence interval B will be explained with reference to FIGS. 21 and 22. Join in a straight line. Find the bend point V1 that is farthest from the straight line, and if the distance is greater than or equal to the above ε, divide it into two line segments at the bend point v1,
Divide into L 1L2. As a result, two straight lines LXL2 shown in FIG. 21(b) are obtained.

Next, as shown in FIG. 21(c), find the farthest bend point V3 from the straight line L1, and if the distance is ε = 0.14 strokes, set that point as the bend point v3, and set that point as the bent point. . Repeat this method until there are no bend points, and if there are no bend points, use the dividing points as approximate points.

As a result, piecewise linear approximations L SL , L , and L4 as shown in FIG. 21(d) are finally obtained. When these linear approximations L , L , L, , L4 are applied between Pl, P , and P SP in FIG. 20 and the same processing as described above is performed, each approximate point P 1P6 in FIG. 22, Pl is obtained. Next, as shown in FIG. 23, the approximate points P 1 and P 2 and P 2 are connected, respectively, and the midpoints of the line segments are set as new bend points P8 and P9 (see FIG. 23) (step 105).

Figures 24 (a) and (b) show the shape before and after correction of the above-mentioned contents in the case of a corner where two prediction intervals intersect.
Let's briefly summarize and explain.

If the prediction intervals A and B intersect, the intersection of the center lines of the prediction intervals is set as the new bend point. The state before such angle correction is shown in Fig. 21(a), and the state after correction is shown in Fig. 21(a).
Shown in b).

In addition, if the prediction intervals do not intersect, correction can be made by performing linear approximation and finding a new bend point (see Figure 25).
The corner correction in the state shown in a) is corrected to the inside of the circle in FIG. 25(b) (step 105).

As a result of performing the angle correction on the entire screen in this manner, an image as shown in FIG. 26 is obtained. In FIG. 26, the corners within the round areas 50, 51, and 52 are corrected.

Next, automatic correction of a T-shaped branch point will be explained with reference to FIGS. 27 to 32. Here, the reason why the correction for the T-shaped branch point and the Y-shaped branch point are separated will be explained.

Note that the correction for T-shaped junctions also includes correction for cross-shaped intersections. Roads are usually built to be as straight as possible. Therefore, junction intersections are often T-shaped or cross-shaped, but in the case of straight roads like this, the center lines need to be connected in a straight line at junctions and intersections. Therefore, branch points and intersections that include straight roads, such as T-junctions and crossroads, are extracted in advance, and the center line of this straight road is corrected so that it becomes straight at this T-junction or crossroad. do. In the case of branch points (Y-junctions) and intersections that do not include straight roads, it is necessary to bend the center line of the road at the branch intersection, so correction is performed using another method. First, a predicted section as a single road at a T-shaped junction is extracted. First, T
Three arcs forming a branch point of a letter 30.31.32
Find the confidence interval for each.

Referring to FIG. 27, a T-junction is formed by the center line arc 30, arc 31, and arc 32 of the block indicated by broken lines. Here, a prediction interval for the arc 31 is determined. First, select four points v, vv, and ■14 of the block-to-vector vector corresponding to arc 31 (
Figure 28). A rectangle R1 circumscribing the bend point v, vvv 1112ゝ 13ゝ 14 is formed (Fig. 30).

The rectangle is defined as a prediction interval regarding the arc 31.

Next, a prediction interval for the arc 32 is determined.

The four points corresponding to arc 32, block bend points ■, v2122ゝ■, and ■, are selected (Fig. 29). The rectangle 2122゜ circumscribing the bend points ■, V ■, ■
2324 Form R2 (Figure 31). The prediction interval R1 includes the prediction interval for the arc 32, and the prediction interval R2 includes the prediction interval for the arc 31. In this way, among the predicted sections in which the predicted sections of other arcs are included, two arcs with high parallelism are defined as a straight road connected in a straight line. Here, arc 31 and arc 32 are assumed to be one road (FIGS. 33 and 34).

Next, the search for a straight line arc will be explained with reference to FIGS. 35 to 38 (step 106).

The state of connection with other arcs is checked for each arc, and arcs that are lined up in a straight line (straight arcs) are determined for each of the starting point side and the ending point side. If the desired arc is not found among the directly connected arcs, arcs without confidence intervals are traced from that location up to two times to find a straight arc.

The above processing will be explained in more detail using FIG. 35. First, it is checked whether there is an arc that is directly connected to the start point or end point of the arc 40 to be investigated and that is lined up in a straight line among the arcs that have a confidence interval. the result,
Arc 41 is selected, but since it is not lined up in a straight line, it is recognized that there are no arcs that are directly connected to arc 40 and lined up in a straight line. Next, it is checked whether there is an arc that is directly connected to the start point or end point of the arc 40 to be investigated but does not have a confidence interval. As a result, arc 44 is selected. If there is no such arc 44, it is assumed that there is no arc linearly aligned with the target arc 40, and the search ends.

Next, the branch point or intersection after tracing the arc 44 is determined. As a result, branch point 45 is selected. It is checked whether there is an arc that is directly connected to this branch point 45 and has a confidence interval and that is lined up in a straight line with respect to the arc 40. In this case, since the arc 42 is lined up in a straight line with respect to the arc 40, the straight arc with respect to the arc 40 is recognized as the arc 42. If there is no matching arc among these, the above-described procedure is repeated one more time to repeat the process of finding arcs that are lined up in a straight line.

Next, a description will be given with reference to FIG. 36. In FIG. 36, F2O is the portion of the modified road centerline that is within the confidence interval of target arc 40, and F42 is the portion of the modified road centerline that is within the confidence interval of arc 42.

There are two conditions for being certified as a straight arc.The first condition is that the angle difference between the confidence intervals of both arcs is 3.
The second condition is that the end point E S of the part of the corrected road center line is within the confidence interval of both sides.
F3 is connected by straight line 46 and the corrected road center line, arc 42, is within the confidence interval F2O of arc 46 from its midpoint M.
A perpendicular line is drawn to each of the corrected road center lines within the confidence interval F42. Next, end points E, F of the modified road center line that are within the confidence interval F40 of both arcs 40.42
Connect the end points E2 of the corrected road center line within F42 with a straight line, and draw from the midpoint M to the corrected road center line within each confidence interval F2O, and the corrected road center line within F42. The length of the perpendicular line must be within 0.10.

If these two conditions are met, two arcs of 40°42
It is determined that the lines are arranged in a straight line.

Therefore, if the arcs 40 and 42 satisfy these two conditions, it is determined that the arcs 40 and 42 are aligned in a straight line (step 106). By performing such a search, even in cases such as the intersection 130 in Fig. 37, which is originally an intersection but has been separated into two branch points due to the distortion of line thinning, it is possible to properly search for a straight arc, and find the next straight arc. With T-junction correction, it is possible to accurately correct the intersection to an intersection like intersection 131 in FIG. 38.

Next, T-junction correction will be explained with reference to FIGS. 39 to 44 (step 107). Based on the results of extracting a predicted section at a T-shaped junction as a single road and searching for straight arcs, the shapes of junctions and intersections containing arc pairs arranged in a straight line are determined based on confidence intervals. to correct.

However, a series of arcs that are consecutively arranged in a straight line is extracted, and a plurality of branch points and intersections included in the arc series are collectively corrected. First, a prediction section is extracted using a series of arcs 50 (main line) that are continuously arranged in a straight line and branch arcs 51 and 52 that are connected to the series of arcs from the side as a T-shaped junction as one road. Extract based on the results of straight arc search. Bent points V that are not within the confidence interval near branch points and intersections are deleted, and the bend points V of the arc string (main line) are
Create an arc by taking out the coordinates of from the end in order.

The erased bend point ■ extends the corrected road center line within the confidence interval, or directly connects the end points of the road center line in the preceding and succeeding confidence interval parts (see FIG. 42).

Branch arcs 51.52 connected to that arc row from the side
After eliminating bend points that are not within the confidence interval near the junction or intersection, the intersection of that arc and the modified road centerline extension within your own confidence interval is set as a new junction, Correct it as an intersection (see Figure 43). If there is a pair of arcs that are lined up in a straight line with each other among the branch arcs, that part is recognized as an intersection, the shapes of both are corrected at the same time, and the intersections are combined into one. Thereafter, a series of arcs (main line) connected in a straight line from the coordinates of the intersection with one arc created from the original series of arcs (main line) is recreated (see Figure 44). The road center lines before and after correction are shown in FIG. 37 and FIG. 38, respectively.

Next, Y-junction correction will be explained using FIGS. 45 to 48 (step 108).

The following processing is performed for all branch points and intersections that were not corrected by the T-junction correction. First, find uncorrected branch points and intersections. Next, check the arcs connected to the branch points and intersections to be corrected, and check whether there are two or more arcs with confidence intervals. Arcs connecting to intersections 60.61.
62 confidence intervals 60F, 61F, and 62F are extracted and intersection points are found for all combinations of corrected road center lines within them. Let K be the intersection of the corrected road center line within the confidence interval 60F and the corrected road center line within the confidence interval 61F, and let K be the intersection of the corrected road center line within the confidence interval 61F and the corrected road center line within the confidence interval 62F. The intersection with the line is K
, the intersection of the corrected road center line within the confidence interval 60F and the corrected road center line within the confidence interval 62F is assumed to be . Then, the center coordinates ○ of the triangle surrounded by this intersection K 1K SK3 are obtained. The obtained center coordinate O is the branch point,
Check the following points for each arc to see if it is a valid intersection. For an arc with a confidence interval, whether the position of the resulting coordinate is in front of the confidence interval for the arc. That is, it is checked whether the end points of the confidence interval are extended and intersect with each other to form a triangle T in FIG. 46. In the case of an arc that does not include a confidence interval, it is checked whether the position of the obtained coordinates is ahead of the bend point closest to the branch point or intersection. If the obtained coordinate position is valid,
For each arc, bend points that are not within the confidence interval near branch points and intersections are deleted, and the center coordinates O are set as new branch points and intersections for correction.

Thereafter, the shape of the arc changes due to T-junction correction and Y-junction correction, so a bend point may exist in the middle of a straight line portion. Such unnecessary bend points are deleted by the bend point reduction means 11 (step 109). FIG. 49 shows the drawing after T-junction correction, FIG. 50 shows the drawing after Y-junction correction, and FIG. 50 shows the drawing after vent point reduction.

In FIGS. 49 and 50, the T-junction and Y-junction marked with O marks 49 and 50 have been corrected, respectively.

〔Effect of the invention〕

As described above, the present invention extracts connection information between arcs, extracts a pair of line segments among the line segments obtained by connecting the bending points of the contour, and uses the center line of the confidence interval. and correct it,
Find center lines corresponding to end points, corners, branch points, and intersections. Therefore, unlike the method of erasing line segments of a certain length or less without finding a confidence interval and filling in those parts, this method allows for accurate correction even when roads with different widths coexist, and it does not require manual correction. To properly correct a road center line without requiring much effort and without frequently omitting corrections.

[Brief explanation of the drawing]

Figure 1 is a block diagram of the thinning distorted shape correcting device of the present invention, Figure 2 is an operation flowchart of the thinning distorted shape correcting device of the present invention, and Figure 3 is an explanatory diagram of center line extraction by thinning. , FIG. 4 is an explanatory diagram showing the least squares vector approximation, FIG. 5 is an explanatory diagram showing the least squares vector approximation with node display, and FIG. 6 is an explanatory diagram explaining the least squares approximation method. , FIG. 7 is a state diagram before removing whiskers from corners, FIG. 8 is a diagram after correction, FIG. 9 is an explanatory diagram of extraction of confidence intervals of thinning results, and FIG. 10 is An explanatory diagram before confidence interval extraction, Figure 11 is an explanatory diagram after the bend point has been corrected after the confidence interval has been extracted, and Figures 12 and 13 are the overall diagram before confidence interval extraction and after the bend point has been corrected. , FIGS. 14 to 16 are illustrations of road end point correction, illustrations of corner correction when predicted sections intersect, and FIGS. 17 to 19 are illustrations of corner correction when predicted sections intersect. , Figures 20 to 2
Figure 3 is an explanatory diagram of corner correction when two prediction intervals do not intersect, Figure 24 is an explanatory diagram when two prediction intervals intersect, and Figure 25 is an illustration of angle correction when two prediction intervals do not intersect. Explanatory drawing, Fig. 26 is an overall explanatory drawing after angle correction, Fig. 27
Figures 34-34 are explanatory diagrams for extracting prediction intervals, Figures 35-36 are explanatory diagrams for searching for straight arcs, and Figure 3 is an explanatory diagram for searching for straight arcs.
7 to 44 are explanatory diagrams of T-junction correction, FIGS. 45 to 48 are explanatory diagrams of Y-junction correction, FIG. 49 is an overall diagram after T-junction correction, and FIG. 50 is the overall view after correcting the Y-junction, Figure 51 is the overall view after removing unnecessary vent points, and Figure 5 is the overall view after removing unnecessary vent points.
2 to 57 are explanatory diagrams of a centerline correction device of the prior art. 2... Road center line thinning means, 3... Corner hair removal means, 4... Confidence interval extraction means, 5... Edge correction means, 6... Corner correction means, 7. . . . Prediction section extraction means, 8 . . . Straight line A, - ri search means, 9 . . . T-junction correction means, 10 .

Claims (1)

[Scope of Claims] 1. A thinning distorted shape automatic correction device for correcting distortion of a thinned line figure of an arbitrary thickness and determining the center line of the line figure, the thinning result being divided into branches. Separated by points and intersections,
Connection information extraction means extracts polygonal lines that do not include other junctions or intersections in the middle between junctions or intersections, and extracts connection information between the polygonal lines; A line figure bending point extracting means extracts bending points, and among the line segments obtained by connecting the bending points in the traced order, line segments that are paired on both sides of the broken line obtained from the thinning result are selected. a confidence interval extracting means for obtaining a confidence interval consisting of a rectangle inscribed in the set, correcting the shape of a polygonal line within the confidence interval using the coordinates of the set of line segments, and obtaining a center line of the confidence interval;
a prediction interval correction means that calculates a prediction interval consisting of a rectangle circumscribing a set of line segments, and uses the prediction interval to modify the shape of a polygonal line at a bending point of a line figure; A thinning distorted shape automatic correction device characterized by comprising a bifurcation point correction means for correcting the shape of the connecting part using the center line of the confidence interval and obtaining the center line corresponding to the bifurcation point or intersection part of the thinning result. . 2. An automatic thinning distortion shape correction method for correcting distortions in thinned line figures and finding the center line of the line figure, which divides the thinned line figures at their branching points and intersections, and It extracts polygonal lines that do not include other branching points or intersections in the middle, extracts connection information between the polygonal lines, traces the outline of the line shape, extracts the bending points of the line shape, and extracts the bending points in the traced order. Among the line segments obtained by connecting the line segments, extract a pair of line segments on both sides of the polygonal line obtained from the thinning result, and calculate the confidence interval made of rectangles inscribed in the set of line segments. Find the center line of the confidence interval, calculate the prediction interval consisting of a rectangle circumscribing the set of line segments, and use the coordinates of the set of line segments to find the center line of the confidence interval. Modify the shape of the polygonal line at the bending point of the line shape, modify the shape of the polygonal line connection using the center line of the confidence interval based on the connection information between the polygonal lines, and correct the shape of the polygonal line at the bending point of the line shape. 1. A method for automatically correcting a thinned distorted shape, the method comprising the step of determining a center line corresponding to .