KR101146128B1 - Navigatioin method and apparatus using sector based image matching - Google Patents

Abstract

Photographing a landmark at a reference point, dividing the photographed image into a first plurality of zones, calculating a reference position value of the landmark for each of the first plurality of zones, and photographing the landmark at the current point Dividing the captured image into a second plurality of zones, predicting an image to be changed in the second plurality of zones by a plurality of arbitrary points when moving from a current point to a plurality of random points, and Calculating an estimated position value of the landmark for each of the plurality of arbitrary points and the second plurality of zones, and comparing the estimated position value with the reference position value according to the image predicted for each arbitrary point. A navigation method and apparatus comprising the step of determining an arbitrary point having an expected position value having the highest matching rate with a reference position value as a moving point, The ball.

Navigation, Image Segmentation, Image Matching, Robots

Description

Navigation method and apparatus using segmentation based image matching {NAVIGATIOIN METHOD AND APPARATUS USING SECTOR BASED IMAGE MATCHING}

The present invention relates to a navigation apparatus and method by image matching based on image segmentation.

Recently, many robot-related technologies such as robot cleaners have been actively studied. In the case of a robot which is movable among these robots, various navigation methods are used to return to a target point.

In nature, insects such as bees and other animals are known to have their own home navigation to return to their homes. Among them, the landmark navigation method, which recognizes and uses natural objects or features around the house as visual information, is known to be used by many insects such as bees and desert ants, and it shows effective performance with a relatively small amount of information. Many studies have demonstrated this.

Various models have been proposed regarding the algorithm of the insect landmark navigation system, and the snapshot model proposed by Cartwright is representative. The model remembers the image taken at home at departure and then compares with the image at the current location at the time of return and takes the highest matching rate. Franz proposed an image matching algorithm based on this model, and showed that it can be returned to the starting point by applying it to a robot. In addition, the image matching method using Anderson's concept of segmentation enables effective home navigation even in low resolution environments.

However, when these methods are applied to a mobile robot, it is not practical and has a problem of inaccuracy because it requires a complicated calculation process for determining the current position of the robot.

Embodiments of the present invention have been made to solve the above-described problems, and provides a navigation method and apparatus using segmentation-based image matching that can be quickly and accurately navigated through a simple calculation process that can be applied to a mobile robot, etc. For the purpose of

According to an aspect of the present invention, the method includes: dividing an image photographed by photographing a landmark at a reference point into a first plurality of zones, and calculating reference position values of the landmarks for each of the first plurality of zones; Dividing the photographed image by taking a landmark at the current point into a second plurality of zones, and a plurality of random points of an image to be changed in the second plurality of zones when moving from the current point to a plurality of random points Predicting each position, calculating an estimated position value of the landmark for each of the plurality of arbitrary points and for each of the second plurality of zones according to the image predicted for each of the plurality of arbitrary points, and the estimated position value. Is compared with the reference position value to determine an arbitrary point having an expected position value having the highest matching rate with the reference position value among the plurality of arbitrary points as the movement point. It provides a navigation method comprising the system.

In one embodiment, the reference position value and the expected position value may be calculated from whether a landmark exists in the corresponding area among the first and second plurality of zones and the average distance to the landmark, respectively.

Meanwhile, in one embodiment, each first plurality of zones may correspond to each second plurality of zones.

The calculating of the predicted position value for the plurality of arbitrary points may include calculating the predicted position value of the landmark in the second plurality of zones for each of the plurality of arbitrary points.

The image photographed in the calculating of the reference position value and the dividing into the second plurality of zones may be photographed by the omnidirectional camera.

According to another aspect of the present invention, there is provided a omnidirectional camera capable of capturing a surrounding image at a time and a controller for processing an image captured by the omnidirectional camera to determine a moving direction, wherein the controller is configured as a omnidirectional camera at a reference point. Taking a landmark and dividing the captured image into a first plurality of zones, calculating a reference position value of the landmark, taking a landmark with the omnidirectional camera at the current point, and taking the captured image into the second plurality of zones. Segmenting, predicting an image to be changed in the second plurality of zones when moving from a current point to a plurality of arbitrary points by a plurality of arbitrary points, and landmarks according to the predicted image for each of the plurality of arbitrary points Calculate the predicted position of for each of the plurality of random points and the second plurality of zones, and apply the predicted image to each of the plurality of random points. Accordingly, the predicted position value of the landmark is calculated for a plurality of arbitrary points, and the plurality of predicted position values are compared with the reference position value to have an expected position value having the highest matching rate with the reference position value among the plurality of arbitrary points. Provided is a navigation device for determining an arbitrary point as a moving point.

According to one embodiment of the present invention, the segmentation-based image matching has the effect of moving quickly and accurately from the current point to the reference point.

A navigation method using segmentation-based image matching according to an embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is a flowchart illustrating a navigation method using segmentation-based image matching according to an embodiment of the present invention. As illustrated in FIG. 1, the navigation method using segmentation-based image matching divides a photographed image by photographing a landmark at a reference point into a first plurality of zones, and landmarks for each of the first plurality of zones. Computing a reference position value of the step (S110), the step of taking a landmark from the current point and dividing the captured image into a second plurality of zones (S120), and when moving to a plurality of random points from the current point Predicting an image to be changed in the second plurality of zones of each of the plurality of arbitrary points (S130), and predicting a position value of the landmark according to the predicted image for each of the plurality of arbitrary points of the plurality of arbitrary points. According to the comparison result in the step (S140) for calculating each of the plurality of points and each of the second plurality of points, the step of comparing the plurality of expected position values with the reference position value (S150), and the comparing step And a step (S160) for determining a moving point. Through this process, a device such as a mobile robot, which is out of the reference point, may return to the reference point.

These processes are described in detail as follows. On the other hand, it is apparent that the embodiments of the present invention can be applied to all of the movable devices in addition to the mobile robot, but for the convenience of description, it will be described with an example applied to the mobile robot.

First, in step S110, the mobile robot photographs a landmark at a reference point using a mounted camera. The camera is preferably an omnidirectional camera. The omnidirectional camera is a device that can capture the environment around the mobile robot all at once. The landmark is a mark installed in an environment in which a mobile robot is located, and a landmark commonly used in the technical field of the present invention may be used. Also instead of this, it may be an object designated as a landmark among various objects existing in the surrounding environment of the mobile robot. When the image is acquired by photographing the landmark, the acquired image is divided into a first plurality of zones, and each image divided into the first plurality of zones is used to map the landmark to each of the first plurality of zones. Calculate the reference position value for. Therefore, in step S110, one set of reference position values having a number of reference position values corresponding to the number of the first plurality of zones is calculated.

The reference position value of the landmark may be calculated by digitizing whether or not the landmark exists in the region and the average distance distance between the mobile robot and the landmark. However, this is only one example, and the present invention is not limited thereto, and any value may be used as long as it can indicate the position of the landmark in each zone.

In this way, the reference position value (reference position value set) at the reference point may be calculated and stored in step S110.

Next, in step S120, the landmark is photographed at the current point where the mobile robot is located. The landmark at the current point is also taken by the omnidirectional camera as in step S110, and preferably by the same camera as the camera used in step S110. When the image is acquired by photographing the landmark, the acquired image is divided into a second plurality of zones as in step S110. The second plurality of zones may respectively correspond to the divided first plurality of zones in step S110, and may correspond to the same area in the captured image. The number of the first plurality of zones and the second plurality of zones divided in steps S110 and S120 may be any number, but the number of the divided zones is preferably small to increase the processing speed.

After the image photographed at the current point is divided into a plurality of zones, a changing image in the second plurality of zones when moving from the current point to a plurality of arbitrary points is predicted in step S130. Assuming a case where the mobile robot moves from the current position to several positions (multiple arbitrary points), the image change in the second plurality of zones for each arbitrary point is predicted. For example, if it is assumed that the current point moves from an arbitrary point, a change in the viewing angle and the distance of each pixel in the image may be predicted for each of the second plurality of zones according to the moving direction and distance. This process is repeated for each arbitrary point, whereby a changing image for each random point can be predicted in step S130.

If the image change for each of the arbitrary points is predicted, in step S140, according to the image predicted for each of the plurality of arbitrary points, the estimated position values of the landmarks are respectively applied to each of the plurality of arbitrary points and to each of the second plurality of points. Calculate Through this process, an estimated position value corresponding to the second plurality of zones is obtained for each arbitrary point. That is, each arbitrary point has one set of estimated position values (having as many as the number of the second plurality of zones), and as a result, an estimated position value of the set number corresponding to the number of arbitrary points is obtained. The expected position value is also calculated in the same manner as the reference position value calculated in step S110. That is, the estimated position value of the landmark may be calculated by digitizing whether or not the landmark exists in the corresponding region of the predicted image and the average distance between the mobile robot and the landmark. In this way, a plurality of estimated position values corresponding to a plurality of arbitrary points may be calculated in step S140.

When the estimated position value is obtained, in step S150, the plurality of estimated position values are compared with the reference position values. In operation S160, an arbitrary point having an optimal estimated position value may be determined among the plurality of arbitrary points according to the comparison result. That is, in steps S150 and S160, the reference position value set is compared with the expected position value set for each arbitrary point, and an arbitrary point having the best predicted position value set is selected. This arbitrary point becomes the optimal path from the current point toward the reference point, and navigation to the reference point can be efficiently performed by moving to the predetermined random point and performing steps S120 to S160 again.

Through this process, the surrounding environment is recognized as a combination of several partitions through segmentation-based image matching, and the mobile robot determines whether a landmark exists for each partition with respect to the image of the current location. By determining the average distance between landmarks, the two factors determine the matching rate with the image of the reference point, and moves to the position with the highest matching rate.

Next, another embodiment of the present invention will be described with reference to FIG. 2 illustrates a navigation apparatus 200 using segmentation based image matching according to an embodiment of the present invention. The navigation device includes an omnidirectional camera 210 capable of capturing a surrounding image at a time, and a controller 220 for processing an image photographed by the omnidirectional camera to determine a moving direction. The controller 220 divides the photographed image by photographing the landmark 300 with the omni-directional camera 210 at a reference point into a first plurality of zones, and landmarks 300 for each of the first plurality of zones. Calculate a reference position value for), split the captured image into a second plurality of zones by shooting the landmark 300 with the omnidirectional camera 210 at the current point, and move to a plurality of random points at the current point. Predict the changed image in the second plurality of zones for each second plurality of zones, and predict the position of the landmark 300 according to the predicted image for the plurality of arbitrary points and the second plurality of zones. A value is calculated for a plurality of arbitrary points, and a plurality of the predicted position values are compared with a reference position value, and any point having an expected position value having the highest matching rate with the reference position value among the plurality of arbitrary points is used as the movement point. texture The.

The controller 220 is configured to perform the above-described navigation method with reference to FIG. 1, and a detailed description thereof is the same as the above-described navigation method, and thus a detailed description thereof will be omitted. The controller 220 may include various components that perform various functions, respectively.

Next, the selected direction vector when the algorithm according to an embodiment of the present invention is applied and the selected direction vector when the algorithm according to the prior art are applied will be described with reference to FIGS. 3A and 3B.

3A illustrates a direction vector selected to return to a reference position in a navigation method using an image matching algorithm according to the prior art, and FIG. 3B illustrates a navigation method according to an embodiment of the present invention, that is, a navigation method using segmentation based image matching. Represents the direction vector selected to return to the reference position in. 3A and 3B show a plane for the environment in which the mobile robot is present, the square in the center represents the reference point, and the circle represents the location of the landmark. When the image matching algorithm according to the prior art is selected as shown in FIG. 3A, since a direction for maximizing the inner product is selected, a tendency to move closer toward the landmark rather than the reference point direction appears when a particular landmark is approached. The performance is poor, and therefore, when the singular mobile robot is outside the area surrounded by the landmark, it is difficult to enter the outside area from the outside area near the boundary of the area surrounded by the landmark.

In contrast, in the case of using image segmentation-based matching according to an embodiment of the present invention, it can be seen that the direction vectors more accurately point to the reference point in all regions. 3B shows an example of the selected direction vector when each image is divided into four sections. Therefore, according to the navigation method using the image segmentation-based matching according to an embodiment of the present invention, there is an effect that more accurate navigation is provided through faster processing than when using the image matching algorithm according to the prior art.

Next, the performance according to the number of compartments is compared using FIGS. 4 and 5. 4 illustrates the division of the surrounding environment according to the number N of compartments when the reference direction of the navigation device according to one embodiment of the present invention is 0 degrees. In FIG. 4, the landmark shows an example in which five cylindrical landmarks are surrounded.

5 is a graph showing an angle error (angle difference from a reference point) according to the number of partitions according to the distance from the reference point. As can be seen in Figure 5, as N increases from 3 to 4, the direction of the reference point tends to be more accurate. However, overall, even if N increases, there is no big difference in the change in performance. Even if N is 3, most of the error rate is 90 degrees or less over the entire region, so even if three partitions are used, it is possible to return to the reference point along the path of the connected vectors.

However, if N is increased by more than a certain number, that is, the resolution is increased, it becomes similar to the pixel-based image matching according to the prior art, and thus has the problems described above. Can be selected accordingly.

Although the present invention has been described above with respect to the embodiments described above, it will be apparent to those skilled in the art that various modifications, changes or variations can be made without departing from the technical scope of the present invention. . However, it is apparent that such modifications, changes or variations are included in the scope of the present invention.

1 is a flowchart illustrating a navigation method using segmentation-based image matching according to an embodiment of the present invention.

2 illustrates a navigation apparatus using segmentation based image matching according to an embodiment of the present invention.

3A illustrates a direction vector selected to return to a reference position in a navigation method using an image matching algorithm according to the prior art, and FIG. 3B illustrates a reference vector in a navigation method using segmentation-based image matching according to an embodiment of the present invention. Represents the direction vector selected for feedback.

4 illustrates the division of the surrounding environment according to the number N of compartments when the reference direction of the navigation device according to one embodiment of the present invention is 0 degrees.

5 is a graph illustrating an angle error according to the number of partitions according to an angle difference from a reference point.

<Explanation of symbols for the main parts of the drawings>

200: navigation device

210: omnidirectional camera

220: control unit

300: landmark

Claims (9)

Photographing a landmark at a reference point, dividing the photographed image into a first plurality of zones, and calculating reference position values of the landmarks for each of the first plurality of zones; Dividing the photographed image into a second plurality of zones by photographing the landmark at a current point; Predicting an image to be changed in the second plurality of zones when moving from the current point to a plurality of arbitrary points by the plurality of arbitrary points; Calculating an estimated position value of the landmark for each of the plurality of arbitrary points and the second plurality of zones, according to the image predicted for each of the plurality of arbitrary points; And Comparing the expected position value with the reference position value to determine an arbitrary point having an expected position value having the highest matching rate with the reference position value among the plurality of arbitrary points as a movement point, Each of the first plurality of zones corresponds to a respective one of the second plurality of zones. The method of claim 1, The reference position value and the expected position value are respectively calculated from the presence or absence of the landmark in the corresponding zone among the first and second plurality of zones and the average distance to the landmark. Navigation method. delete The method of claim 1, The calculating of the predicted position value for the plurality of arbitrary points includes: calculating an estimated position value of the landmark in the second plurality of zones for each of the plurality of arbitrary points, Navigation method. The method of claim 1, The image photographed in the calculating of the reference position value and dividing into the second plurality of zones is an image photographed by an omnidirectional camera. Navigation method. An omnidirectional camera capable of capturing a surrounding image at a time; And A controller configured to process an image photographed by the omnidirectional camera to determine a moving direction; Including; The control unit, Taking a landmark by the omnidirectional camera at a reference point, dividing the photographed image into a first plurality of zones, calculating a reference position value of the landmark, Dividing the image taken by photographing the landmark with the omnidirectional camera at a current point into a second plurality of zones, Predicting the image to be changed in the second plurality of zones when moving from the current point to a plurality of random points by the plurality of random points, Calculate an estimated position value of the landmark for each of the plurality of arbitrary points and the second plurality of zones, according to the image predicted for each of the plurality of arbitrary points, A plurality of expected position values are compared with a reference position value to determine an arbitrary point having an expected position value having the highest matching rate with the reference position value among the plurality of arbitrary points as the movement point, Each first plurality of zones corresponds to each of the second plurality of zones. The method of claim 6, The reference position value and the expected position value are respectively calculated from the presence or absence of the landmark in the corresponding zone among the first and second plurality of zones and the average distance to the landmark. Navigation device. delete The method of claim 6, The control unit calculates, for each of the plurality of arbitrary points, an estimated position value of the landmark in the second plurality of zones. Navigation device.

Images