JP2018146440A - Environment recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an environment recognition device capable of easily recognizing a surrounding environment.SOLUTION: An environment recognition device includes: a first sensor 30 mounted on a base 20 and acquiring positional information of an object existing around the base 20; and a second sensor 40 mounted on the base 20 and having an angle of view narrower than the angle of view of the first sensor 30 and acquiring image information of the object existing around the base 20. The environment recognition device further includes a calculation part for recognizing the surrounding environment from the image information of the object by the second sensor 40 acquired by having the object into the field of view of the second sensor 40 from the positional information of the object acquired by the first sensor 30 and recognizing the surrounding environment from the positional information of the object acquired by the first sensor 30.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an environment recognition device.

  A function of recognizing the surrounding environment is required in a mobile robot or the like. In order to realize this function, an environment recognition device using a distance sensor such as a laser range finder (LRF) has been proposed. In addition, in order to sufficiently perform environment recognition, an environment recognition apparatus in which not only a distance sensor but also a camera is mounted on the environment recognition apparatus has been proposed (for example, Patent Document 1).

Japanese Patent Laid-Open No. 2006-150410

  By the way, in the environment recognition apparatus for recognizing the surrounding environment, improvement is demanded from the viewpoint of information processing load by simply detecting surrounding objects using two types of sensors (distance sensor and camera).

  An object of the present invention is to provide an environment recognition device that can easily recognize the surrounding environment.

  According to the first aspect of the present invention, a first sensor that is mounted on the base and obtains position information of an object existing around the base, and an angle of view of the first sensor that is mounted on the base. A second sensor having a narrower angle of view and acquiring image information of an object existing around the base, and the position information of the object acquired by the first sensor in the field of view of the second sensor The gist is provided with an arithmetic unit that recognizes the surrounding environment from the image information of the object obtained by the second sensor acquired by inserting the object and the position information of the object acquired by the first sensor. .

  According to the first aspect of the present invention, it is possible to acquire position information of an object existing around the base by the first sensor. With the second sensor, image information of an object existing around the base can be acquired. The second sensor has a field angle narrower than that of the first sensor. The calculation unit obtains the image information of the object by the second sensor obtained by putting the object into the field of view of the second sensor from the position information of the object obtained by the first sensor, and the image obtained by the first sensor. The surrounding environment is recognized from the position information of the object. Therefore, the surrounding environment can be easily recognized.

As described in claim 2, in the environment recognition device according to claim 1, the base may further include orientation changing means for directing the second sensor toward the object.
As described in claim 3, in the environment recognition device according to claim 1 or 2, the first sensor may be a laser range finder, and the second sensor may be a linear image sensor.

  According to a fourth aspect of the present invention, in the environment recognition device according to the third aspect, the calculation unit detects the number and positions of objects present around the base by the laser range finder as the first sensor. Then, the shape and color of the object may be detected by the linear image sensor which is the second sensor, and the stored reference information may be collated from the position, shape and color of the object.

  According to the present invention, the surrounding environment can be easily recognized.

(A) is a schematic plan view of the vehicle and its periphery in an embodiment, (b) is a schematic front view of a vehicle and its periphery. The schematic plan view of the vehicle and its periphery for demonstrating an effect | action. The schematic plan view of the vehicle and its periphery for demonstrating an effect | action. The block diagram of an environment recognition apparatus. The perspective view which shows a part of vehicle. The perspective view of the vehicle of another example. The figure which shows the characteristic of various sensors.

DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, an embodiment of the invention will be described with reference to the drawings.
As shown in FIGS. 1A and 1B, an environment recognition device 10 is mounted on the vehicle. The environment recognition device 10 includes one first sensor 30 and one second sensor 40. The first sensor 30 and the second sensor 40 are mounted on the base 20. The first sensor 30 is a laser range finder (LRF). The second sensor 40 is a linear image sensor. The first sensor 30 is for acquiring position information of an object existing around the base 20. The second sensor 40 is for acquiring image information of an object existing around the base 20.

  As shown in FIGS. 1A, 1 </ b> B, and 5, the base 20 includes a rotating member T <b> 1 and an actuator 60. If it demonstrates using FIG. 5, the column-shaped rotation member T1 is supported by the base 20 so that rotation is possible. The upper surface of the rotating member T1 is a horizontal plane. The rotating member T1 is rotated by a motor as the actuator 60.

  A laser range finder (LRF) as the first sensor 30 is fixed on the upper surface of the rotating member T1. The laser range finder (LRF) which is the first sensor 30 has a cylindrical shape. The angle of view α of the laser range finder (LRF) that is the first sensor 30 is, for example, 270 degrees and a wide angle as shown in FIG. The first sensor 30 faces the traveling direction (front) of the vehicle, and the center of the angle of view is the front of the vehicle.

  On the upper surface of the laser range finder (LRF) that is the first sensor 30, a linear image sensor that is the second sensor 40 is fixed. The linear image sensor which is the second sensor 40 has a columnar shape and extends on a vertical line. The linear image sensor, which is the second sensor 40, has a plurality of image sensors 41 arranged in a line on the vertical axis, and obtains uniaxial image information in an orthogonal three-axis coordinate system. This information includes shape and color.

  The angle of view β of the linear image sensor that is the second sensor 40 is, for example, 10 degrees as shown in FIG. That is, the second sensor 40 has an angle of view β that is narrower than the angle of view α of the first sensor 30. More specifically, the visual field in the horizontal direction is such that the first sensor 30 is wide and the second sensor 40 is narrow. Then, the second sensor 40 may be directed to the object so as to acquire image information of the object by the second sensor 40 by rotating the rotating member T1 with the vertical direction as the rotation axis by the actuator 60 as the direction changing means. It can be done.

  As shown in FIG. 4, the environment recognition device 10 includes a calculation unit 50. The computing unit 50 includes the object image information obtained by the second sensor 40 obtained by putting the object into the field of view of the second sensor 40 based on the object position information obtained by the first sensor 30, and the first sensor. The surrounding environment can be recognized from the position information of the object acquired by 30.

  Specifically, the calculation unit 50 includes an environment recognition unit 51, a control unit 52, and an individual identification unit 53. Information acquired from a laser range finder (LRF) as the first sensor 30 is sent to the environment recognition unit 51. Position information is sent from the environment recognition unit 51 to the control unit 52. Based on this position information, the controller 52 drives the actuator 60. As the actuator 60 is driven, information acquired from the linear image sensor that is the second sensor 40 is sent to the individual identification unit 53. The computing unit 50 is connected to the environmental data holding unit 55. The environment data holding unit 55 stores information about the position, shape, and color of the object as environment data. The computing unit 50 can extract environmental data stored in the environmental data holding unit 55 and update the environmental data.

Next, the operation will be described.
As shown in FIGS. 1A and 1B, the calculation unit 50 uses the laser range finder (LRF) as the first sensor 30 to detect the objects (H 1, H 2, W 1) existing around the base 20. Detect number and position. 1A and 1B show a situation in which a person H1, a person H2, and a wall W1 exist as objects. The number of the objects (H1, H2, W1) is “3”, the position of the person H1 as the object is an angle + θ1 with respect to the front of the vehicle, and the distance is L1. The position of the person H2 as an object has an angle of −θ2 with respect to the front of the vehicle, and a distance of L2. The position of the wall W1 as an object is -θ3 with respect to the front of the vehicle, and the distance is L3. In the notation of the angle of the object, the counterclockwise direction is + and the clockwise direction is −.

  Then, as illustrated in FIG. 2, the calculation unit 50 rotates the linear image sensor, which is the second sensor 40, within a range where the person H <b> 1 as an object can be imaged by driving the actuator 60. And the calculating part 50 detects the shape and color of the person H1 as an object with the linear image sensor which is the 2nd sensor 40. FIG.

  Similarly, for the person H2 as an object, the arithmetic unit 50 rotates the linear image sensor, which is the second sensor 40, to the range where the person H2 as an object can be imaged by driving the actuator 60. And the calculating part 50 detects the shape and color of the person H2 as an object with the linear image sensor which is the 2nd sensor 40. FIG.

  As for the wall W1 as the object, as shown in FIG. 3, the calculation unit 50 rotates the linear image sensor, which is the second sensor 40, to the range where the wall W1 as the object can be imaged by driving the actuator 60. And the calculating part 50 detects the shape and color of the wall W1 as an object with the linear image sensor which is the 2nd sensor 40. FIG.

  In this way, the calculation unit 50 detects the shape and color of the person H1, the shape and color of the person H2, and the shape and color of the wall W1, and then the position, shape, and color of the person H1, H2, and the wall W1 as objects. From this, the stored reference information is verified.

  Here, since the calculation unit 50 detects the shape and color of the object by the linear image sensor which is the second sensor 40, the person H1 is wearing blue clothes, for example, and the person H2 is wearing red clothes. For example, the person H1 and the person H2 can be identified. From the position, shape and color of the object, it can be compared with the stored reference information.

This will be described in detail below.
In FIG. 4, the environment recognizing unit 51 grasps the number and positions of objects existing in the vicinity based on data of a laser range finder (LRF) that is the first sensor 30. Specifically, the environment recognition unit 51 receives data of a laser range finder (LRF) that is the first sensor 30, and performs data clustering, that is, region division. By performing clustering, the number and positions of objects such as people H1 and H2 and walls W1 existing around are detected.

  The controller 52 drives the linear image sensor as the second sensor 40 in the yaw direction (horizontal direction) by the actuator 60. That is, the control unit 52 receives the number and position of clusters from the environment recognition unit 51, drives the actuator 60, and sequentially directs the linear image sensor as the second sensor 40 to each cluster.

  The individual identification unit 53 performs individual identification of an object based on information on the shape and color of the object by the linear image sensor that is the second sensor 40 and the position of the object by the laser range finder that is the first sensor 30. Do.

  That is, the individual identification unit 53 receives information about the cluster position from the environment recognition unit 51 and the shape and color of the cluster from the linear image sensor that is the second sensor 40. The individual identification unit 53 refers to the position, shape, and color of the environmental object stored in the environment data holding unit 55 and collates with the received information about the position, shape, and color of the cluster. Then, the position, shape, and color of the environmental object that is most similar to the cluster are updated with the position, shape, and color of the cluster. If there is no environment object similar to the cluster, a new environment object is generated, and the position, shape, and color of the environment object are set as the position, shape, and color of the cluster.

Hereinafter, a description will be given with reference to FIG.
FIG. 7 is a list of characteristics of various sensors for object detection. In FIG. 7, as a sensor type, a laser range finder (LRF) is used, a camera is used, and a laser range finder (LRF) and a camera are used. The characteristics are cost, detection range, and discrimination ability. Environment recognition by two types of sensors has advantages and disadvantages as shown in FIG. 7 in terms of cost, detection range, and discrimination ability. Here, the discrimination ability refers to the ability to identify an object (individual) such as a person or a wall.

In FIG. 7, when the characteristic is excellent, it is indicated by ◯, when it is inferior, it is indicated by ×, and when it is intermediate, it is indicated by Δ.
The characteristics of each sensor will be specifically described.

  First, a laser range finder (LRF) generally has a wider field angle than a camera and has an excellent detection range. However, since it is a sensor that acquires only the position information of an object, when the objects overlap on the detection line, the ability to identify each object is poor.

  Next, since the camera can acquire information about the shape and color of an object, it has an excellent ability to identify individual objects. However, since the angle of view of the camera is narrower than that of a laser range finder (LRF), the detection range is narrow. In addition, since the information obtained from the camera is two-dimensional, the cost for processing information is higher than that of a laser range finder (LRF) that outputs one-dimensional information.

  In order to solve the drawbacks of these sensors, it is conceivable to use a laser range finder (LRF) and a camera in combination. By combining a laser range finder (LRF) and a camera, it is possible to realize both the detection range and the discrimination capability. However, the cost for processing the information output from the sensor increases. In addition, a high-function computing unit is required, and as the processing load increases, the amount of heat generation increases, so that it is necessary to mount a cooling device for cooling the computing unit. For this reason, design restrictions occur when the environment recognition apparatus is incorporated into a robot or the like.

  In the present embodiment, in order to perform environment recognition using a combination of a laser range finder (LRF) and a linear image sensor, first, a wide range of environment recognition is performed by the laser range finder (LRF), and the number and positions of objects present in the vicinity. To figure out. After that, by directing the linear image sensor toward each object, information on the shape and color of the object is acquired. Individual objects are identified by comparing the shape and color of each object. Here, the information output from the linear image sensor is one-dimensional, and the cost required to process the information is less than that of the camera. For this reason, the calculation load is lighter than that of the conventional environment recognition device, so that the power consumption of the cooling device is small, and a compact environment recognition device can be realized by using a linear image sensor compared to the camera. As a result, it is possible to construct a small environment recognition device with a lower calculation cost than the conventional product.

According to the above embodiment, the following effects can be obtained.
(1) As a configuration of the environment recognition device 10, a first sensor 30 that is mounted on the base 20 and acquires position information of an object existing around the base 20 and a base 20 is mounted. A second sensor 40 having an angle of view narrower than the angle of view of the sensor 30 and acquiring image information of an object existing around the base 20 is provided. Furthermore, the environment recognition apparatus 10 includes the object image information obtained by the second sensor 40 obtained by putting the object in the field of view of the second sensor 40 from the position information of the object obtained by the first sensor 30, and the first A calculation unit 50 that recognizes the surrounding environment from the position information of the object acquired by one sensor 30 is provided. Therefore, the surrounding environment can be easily recognized.

  (2) The base 20 further includes an actuator 60 as direction changing means for directing the second sensor 40 toward the object. Therefore, the second sensor 40 can be easily directed to the object.

(3) Since the first sensor 30 is a laser range finder (LRF) and the second sensor 40 is a linear image sensor, it is practical.
(4) The calculation unit 50 detects the number and position of objects existing around the base 20 with the laser range finder that is the first sensor 30, and the shape of the object with the linear image sensor that is the second sensor 40. The color is detected, and the stored reference information is collated from the position, shape, and color of the object. Therefore, the surrounding environment can be recognized in time series.

The embodiment is not limited to the above, and may be embodied as follows, for example.
As shown in FIG. 6, the orientation of the linear image sensor (40) may be changed by turning a vehicle body such as a four-wheel omni-wheel vehicle. In this case, no additional actuator is required. That is, in FIG. 6, wheels 22 and 23 are provided on the left and right of the base 21, and the wheel 22 or the wheel 23 is rotated forward or reverse, or the wheel 22 and the wheel 23 are rotated forward or reverse, thereby The direction of the base 20 is changed by turning in order to make the second sensor 40 face the object.

The first sensor 30 is a laser range finder (LRF), but may be an ultrasonic sensor.
The second sensor 40 is a linear image sensor, but is not limited to this. That is, although information on color (color) can be acquired, monochrome information may be acquired.

○ Although a linear image sensor is used as the second sensor, the present invention is not limited to this. For example, a sensor that can take a two-dimensional image instead of a one-dimensional image may be used.
O Although environment recognition device 10 was carried in vehicles, it is not restricted to this. For example, the environment recognition device 10 may be mounted on a mobile robot.

  DESCRIPTION OF SYMBOLS 10 ... Environment recognition apparatus, 20 ... Base, 30 ... 1st sensor, 40 ... 2nd sensor, 50 ... Calculation part, 60 ... Actuator.

Claims (4)

A first sensor mounted on a base and for acquiring position information of an object existing around the base;
A second sensor mounted on the base and having an angle of view narrower than the angle of view of the first sensor, and acquiring image information of an object existing around the base;
Image information of the object obtained by the second sensor obtained by putting the object in the field of view of the second sensor from the position information of the object obtained by the first sensor, and the object obtained by the first sensor A computing unit that recognizes the surrounding environment from the position information of
An environment recognition apparatus comprising:   The environment recognition apparatus according to claim 1, further comprising orientation changing means for directing the second sensor toward an object on the base.   The environment recognition apparatus according to claim 1, wherein the first sensor is a laser range finder, and the second sensor is a linear image sensor. The computing unit is
The number and position of an object existing around the base are detected by the laser range finder as the first sensor, and the shape and color of the object are detected by the linear image sensor as the second sensor. 4. The environment recognition apparatus according to claim 3, wherein the reference information stored is collated from the position, shape and color.

Images