KR100881228B1 - Object detection method using ultrasonic sensor - Google Patents

Abstract

A object detecting method using the ultrasonic sensor is provided to improve the response speed by exactly detecting the object using simple method and omitting data in which the reliability is low. A object detecting method using the ultrasonic sensor comprises a step for initializing the lattices of the lattice map(S101); a step for collecting data about scatter it uses sensor(S102); step for confirming the location in which the section paper data is stored in phase of lattice(S103); a step for giving index it is shape according to the degree of accordance in which data recognize of shape(S104); a step for giving the weighted value according to index it is shape in data(S105); and a step for renewing the occupancy probability about lattice it uses data in which the weighted value is given(S106).

Description

Object detection method using ultrasonic sensor {OBJECT DETECTION METHOD USING ULTRASONIC SENSOR}

The present invention relates to a method for detecting an object using an ultrasonic sensor, and more particularly, when data measured by ultrasonic sensors are continuously input, weights are measured in proportion to a shape recognition index. To combine these data to detect an object.

Ambient environment information recognized by a sensor mounted on the robot can have a large impact on the autonomous driving of the mobile robot as a whole. To provide a wealth of information about the environment, expensive laser or vision sensors outperform lower cost ultrasonic sensors. However, using the ultrasonic sensor to realistically represent the surrounding environment is worth it economically.

As a method of fusing measured data to detect an object by using data input from an ultrasonic sensor attached to a mobile robot, the probability grid map method proposed by Morabe and Elfes is widely used. have. Unlike the shape-based map, the probability grid map can update the map immediately by the data measured by the sensor, and can express the shape efficiently and realistically even for the complex object.

The representation of the environment in the form of a probability grid can be usefully applied when the robot is trying to form a path or avoid obstacles by making the shape of the environment intuitive and realistic.

The principle by which the ultrasonic sensor measures the distance to an object in front is the time of flight until the sound wave is transmitted from the transmitter and then reflected by the nearest object to receive the reflected wave at the receiver of the sensor. It calculates and converts it into a distance value.

However, the actual ultrasonic sensor often fails to detect the closest object in the first half because of mirror reflection. Mirror reflection refers to a phenomenon in which the transmitted sound wave does not arrive at the receiver after reflection if the incident angle of the sound wave is greater than half the effective angle of the sound wave.

This phenomenon causes errors by providing incorrect information about the grid when constructing the map using the Bayesian grid probability update model. In order to make up for this point, studies have been proposed to construct a filter that predicts the mirror reflection phenomenon and to remove false information. However, this method is difficult to accurately predict the mirror reflection phenomenon and the update algorithm is complicated, resulting in a slow response time. have.

The present invention compares the data for the same grating measured by the ultrasonic sensor and weights it according to the conformity of the shape to filter the data incorrectly measured due to mirror reflection, etc., and the highly reliable data lead the completion of information of the grating. The present invention provides an object detection method using an ultrasonic sensor with improved accuracy.

An object detection method using an ultrasonic sensor according to an embodiment of the present invention comprises the steps of initializing the grids of the grid map, collecting data on the surrounding objects using the sensor, the grid to store the data on the grid map Determining a position of the symbol, assigning a shape recognition index according to the degree of matching of the shapes recognized by the data, weighting the data according to the shape recognition index, and weighted data to the grid. Updating the occupancy probability for the method.

The sensor may be an ultrasonic sensor, and the ultrasonic sensor may be mounted on a mobile robot.

로 정의될 수 있다.When the number of data measured in an interval is N and the number of data representing the same shape among the measured data is n, the shape recognition index NRF is

It can be defined as.

The shape may be any one selected from the group consisting of circles, lines, and points.

The object detecting method using the ultrasonic sensor according to an embodiment of the present invention may include classifying the collected data into occupied data and non-occupied data, and may further include obtaining a measurement probability for the data. .

As described above, according to the present invention, an object can be detected more accurately by weighting highly reliable data and participating in a result update.

In addition, low-reliability data can be dropped in a simple way, resulting in improved response speed.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art may easily implement the present invention. As those skilled in the art would realize, the described embodiments may be modified in various different ways, all without departing from the spirit or scope of the present invention. In the drawings, parts irrelevant to the description are omitted in order to clearly describe the present invention, and like reference numerals designate like elements throughout the specification.

1 is a perspective view showing a mobile robot equipped with an ultrasonic sensor according to an embodiment of the present invention.

The mobile robot 10 includes a moving part 1 capable of moving by magnetic force and an ultrasonic sensor system 5 mounted on the moving part 1.

The ultrasonic sensor system 5 comprises ultrasonic sensors mounted along the outer circumference of the supporting plate 4 at intervals of π / 4 with respect to the center of the supporting plate 4 and the supporting plate 4 fixed on the upper part of the moving part 1 ( 3) consists of. The radius R of the support plate 4 is 16 cm, the height H on the ground is 63 cm, and the effective beam width W of the ultrasonic sensor 3 is π / 4.

2 is a flowchart illustrating a process of detecting an object using an ultrasonic sensor according to an embodiment of the present invention.

Referring to FIG. 2, the object detection method according to the present embodiment includes initializing grids of a grid map (S101), collecting data on surrounding objects using a sensor (S102), and a grid map. Confirming the position of the grid to store the data on the image (S103), giving a shape recognition index according to the degree of conformity of the shape recognized by the data (S104) and weighting the data according to the shape recognition index (S105), and updating the occupation probability for the grid using the weighted data (S106).

In the present embodiment, the object is detected by the probability grid method. The probability grid method is a method of identifying the occupied state of the grid by dividing a space into a grid and combining measured data. Before collecting data about an object, first initialize each grid to get ready to accept the data.

Next, while moving in the set space by using the mobile robot 10 to recognize the object in front of the plurality of ultrasonic sensors 3 to collect data. According to the probability grid method, the data measured by the ultrasonic sensor 3 is stored as information about each preset grid. Based on these data, a grid map of the environment is created in a probability grid method, and the mobile robot generates a plurality of data for one grid during the movement.

The data generated at different positions determines whether the data measured through the data occlusal filter is morphologically derived from an object of a line, a point or a circle shape.

Hereinafter, a method of identifying a shape using ultrasonic data will be described.

3 shows the morphological relationship of two single ultrasound data to a circular shape, FIG. 4 shows the morphological relationship of two single ultrasound data to a line shape, and FIG. 5 is the morphological relationship of two single ultrasound data to a point shape. Indicates a relationship.

First, referring to FIG. 3, FIG. 2 shows that ultrasonic sensors having measurement origins of O ₁ and O ₂ measure a circle having a center A and a radius R. FIG. The ultrasonic sensor emits a fan-shaped sound wave and uses the reflection of the fan-shaped sound wave to diagnose the shape of the object. The data measured at the measurement origin O ₁ is defined as the data whose measurement origin is O ₁ , the radius is Z ₁ , and the direction angle from the measurement origin (O ₁ ) to the circumscribed circle is Φ ₁ , and the data measured at the measurement origin O ₂ Is defined as the data whose measurement origin is O ₂ , the radius Z _2, and the direction angle Φ ₂ . At this time, the distance between the measurement origin (O _{1 ,} O ₂ ) is d.

The measured data is defined as a measurement origin, a radius, and a direction angle. Based on this, a grid to store measured data on a grid map can be designated.

The hatched effective direction angle area S refers to an area that simultaneously satisfies the effective direction angle range of the ultrasonic data generated at the respective midpoints O ₁ and O ₂ . When two ultrasound data originate from one circle shape, the direction angles Φ ₁ , Φ ₂ up to the circumscribed circle midpoint are included in the effective direction angle region S, so the direction angles Φ ₁ , Φ ₂ are circular. In the triangle formed by the center (A) of and the measurement points (O ₁ , O ₂ ) can be defined by Equation 1 by the second law of cosine.

[Equation 1]

for ˚,

for ˚,

과

는 i번째 초음파 데이터의 각각 최소 및 최대 유효 지향각을 나타내며 [수학식 2]와 같이 나타낼 수 있다.In Equation 1

and

Denotes the minimum and maximum effective directivity angles of the i-th ultrasound data, respectively, and may be expressed as shown in Equation 2.

[Equation 2]

는 i번째 초음파 데이터의 유효 지향각에 대한 중심각을 나타내며 ω는 초음파 데이터의 유효 빔 폭이다. 물리적으로 초음파 데이터의 유효 빔 폭은 탐지 거리와 물체의 종류에 따라 정의될 수 있지만, 실제 적용 시에는 초음파 센서에 의한 물체 탐지 거리에 따른 편차와 물체 표면 특성에 따른 영향이 크지 않기 때문에 탐지 거리와 물체의 종류에 상관 없이 사용되는 초음파 센서의 특성에 따라 상수 값으로 설정한다.In [Equation 2]

Is the center angle with respect to the effective directivity angle of the i-th ultrasound data, and ω is the effective beam width of the ultrasound data. Physically, the effective beam width of the ultrasonic data can be defined according to the detection distance and the type of the object. However, in actual application, the detection distance and the detection distance and the influence of the object surface characteristics are not significant because of the variation of the object detection distance by the ultrasonic sensor. Regardless of the type of object, set the constant value according to the characteristics of the ultrasonic sensor used.

According to Equation 1 and Equation 2, the i-th data on the circular shape may be represented by a radius Z ₁ and Z ₂ and a direction angle φ ₁ and φ ₂ , respectively.

Referring to Figure 4 looks at the morphological relationship of the two ultrasound data to the line shape. In the case of the linear phase, the same result as in the case where R is infinite in the circle shape.

Therefore, if R is set to infinity in Equation 1 and the equation is rearranged, the same result as in Equation 3 below can be obtained.

[Equation 3]

for |,

for |,

for |,

for |,

Referring to Figure 5 looks at the morphological relationship of the two ultrasound data to the point shape. In the case of the point shape, the same result as in the case where R is 0 in [Equation 1] for the circle shape described above.

Therefore, if R is set to 0 in Equation 1 and the equation is rearranged, a result as shown in Equation 4 below can be obtained.

[Equation 4]

for point,

for point,

for point,

for point,

Due to the difference as described above, it is possible to determine what type of object the ultrasonically measured data originates from. If the radius of the virtual shape is close to zero, the virtual shape may be regarded as a point shape, and if the radius is close to infinity, it may be regarded as a line shape. In addition, if the radius of the virtual shape exists in the area corresponding to the effective beam width of the two ultrasound data, it may be regarded as a circular shape satisfying the condition.

Hereinafter, a process of assigning a shape recognition index to each data based on the detected shape will be described.

6 shows a process in which the mobile robot 10 detects the same object while moving on a path. In FIG. 5, k is the number of steps of the mobile robot 10, and n _i represents the number of data for detecting the same shape as the i-th data. The mobile robot 10 searches for surrounding information by using an ultrasonic sensor while moving a path. At this time, it is determined whether the data stored in the data pool detect the same shape.

That is, the shape recognition index for the i-th data is given by identifying the number of data that recognizes the same shape as the i-th data in the combination of the data measured in the predetermined section. Such shape recognition index can be expressed as Equation 5 below.

[Equation 5]

Here, NRF _i means a shape recognition factor for the i-th data. N denotes the total number of data in the corresponding section, and n _i denotes the number of data that recognizes the same shape as the i-th data.

Therefore, if the total number of data is 100 and the number of data that recognizes the same shape as the i-th data is 90, the shape recognition index is 0.9 while the total number of data is 100 and the same shape as the i-th data is obtained. If the number of recognized data is 10, the shape recognition index is 0.1.

The shape recognition index is added as a weight to the data when the measured data participates in the data update. Accordingly, the data having a low shape recognition index plays little role in data update, whereas the data having a high shape recognition index is used. Can fully participate in data updates.

In this way, if the shape recognition index is assigned to each data, data that is incorrectly measured due to a mirror reflection phenomenon may play a role in the update process, and thus may be naturally filtered.

Hereinafter, a process of converting measured data into probability data will be described.

7 shows a conceptual diagram of an ultrasonic sensor measurement model.

Referring to FIG. 7, a single ultrasonic wave typically has an effective directivity angle ω, a minimum detection distance R _min , and a detection distance R. FIG. In addition, the distance detected by the ultrasonic sensor has a constant error (2ε), this error (2ε) is due to the density of the air and the moving speed of the robot.

By setting the distance r _i and the direction θ _i from the measurement point to any grating i in a single ultrasound data region, the probability of the distance can be expressed as P (r _i ) and the direction probability is P ( θ _i ).

Occupancy probability according to the distance and direction of a single ultrasound data can be considered divided into non-occupied area and occupied area. That is, when the ultrasonic wave is generated and reaches the occupied area, the probability of the occupied area can be obtained, and at the same time, the probability of the unoccupied area of the space between the occupied areas can be obtained.

When the direction probability of the grid for the non-occupied area is called P _emp (θ) and the distance probability is called P _emp (r), the measurement probability of the grid for the unoccupied area is expressed as shown in [Equation 6] below. Can be.

[Equation 6]

When the direction probability of the grid for the occupied area is P _occ (θ) and the distance probability is P _occ (r), the measurement probability of the grid for the occupied area can be expressed by Equation 7 below. have.

[Equation 7]

As shown in [Equation 7], since the ultrasonic wave reaches the area where the object is located, the occupied area is located adjacent to the ultrasonic detection distance (R), so the grid distance (r) of the occupied area is between R-ε and R + ε. Will be located.

Thus, according to Equations 6 and 7, data about the measured position can be converted into data about the direction probability and the distance probability.

A weight may be added to the probability data by using the shape recognition index, and the measurement probability of adding the shape recognition index to the probability data may be expressed by Equation 8 below.

 [Equation 8]

P _emp , _i = P _emp (r _i ), P _emp (θ _i ), NSF _i

P _occ , i = P _occ (r _i ) · P _occ (θ _i ) · NSF _i

In Equation 8, P _emp , _i represents a non-occupancy probability in an arbitrary lattice, and P _occ , i represents an occupancy probability in an arbitrary lattice.

When a single ultrasound data z for an arbitrary grating i is measured, the occupancy probability and the non-occupancy probability can be expressed quantitatively as shown in Equation 9 below.

[Equation 9]

, 격자 i가 비점유 영역에 속하는 경우1) sensor model (m _i , z) =

, If grid i belongs to an unoccupied region

, 격자 i가 점유 영역에 속하는 경우2) sensor model (mi, z) =

, If grid i belongs to occupied area

As shown in [Equation 9], the data is set to have a positive value when the data belongs to the occupied area and to have a negative value when the data belongs to the non-occupied area.

Thus, in the process of updating the data, the occupied data adds positive values to increase the probability value, while the non-occupied data adds negative values to reduce the probability value, so that the newly calculated confidence value and the previous stored value in each grid The confidence value of each is updated to update the confidence value of each grid.

Hereinafter, a method of updating information about a cell by combining measured data with conventional data will be described.

In this embodiment, a grid probability update model based on Bayesian theory is applied to determine and update the confidence value of each grid. The grid probability update model using Bayesian theory has great significance in that it uses the grid probability of a previously created map in updating the occupancy probability of the grid using newly measured ultrasound data. The updated lattice probability for any lattice i updated by Bayesian theory can be expressed as Equation 10 below.

[Equation 10]

In Equation 10, p (m _i ) represents the occupancy probability of the i-th grating, z _{1 : t} represents the ultrasound data measured from 1 to t time. In Bayesian theory, P (z _t | z _{1 : t-1} ) of the denominator is a normalizing constant.

The occupancy ratio of the probability grid may be defined as a ratio obtained by dividing the occupancy probability of the grid i by the non-occupancy probability. The occupancy ratio of the probability grid may be expressed by Equation 11 below.

[Equation 11]

In Equation 11, when the occupancy probability of the grid approaches 0 or 1, a problem may occur in the calculation, so that it can be converted into a log form and expressed as Equation 12.

[Equation 12]

In Equation 12, sensor model (m _i , z _t ) represents the probability of occupancy at t seconds for any lattice i, and l _{t -1} (m _i ) represents any probability at t-1 seconds. The ratio of grating occupancy to grating i is shown, and p (m _i ) represents the initial occupancy probability of grating i.

Equation 12 can be expressed as Equation 13 by rearranging the updated occupancy probability p (m _i | z _{1 : t} ) of an arbitrary grid i.

[Equation 13]

As shown in Equation 12, the occupancy probability sensor model (m _i , z _t ) at t seconds for any lattice i and the lattice occupancy ratio l _{t -1} (m _i ), and the initial occupancy probability, p (m _i ), to obtain the grid occupancy rate l _t (m _i ) up to t seconds, and the updated occupancy probability for the grid (i) through the grid occupancy rate. have.

In this way, by applying the Bayesian theory to add the shape recognition index to the data to update the occupancy probability of the grid (i), it is possible to more accurately and easily identify the surrounding objects by actively participating in the update process.

In addition, the response speed is improved because the incorrectly measured data can be filtered only by comparing shapes with a shape recognition index.

Although the preferred embodiments of the present invention have been described above, the present invention is not limited thereto, and various modifications and changes can be made within the scope of the claims and the detailed description of the invention and the accompanying drawings. Naturally, it belongs to the scope of the invention.

1 is a perspective view showing a mobile robot in an embodiment of the present invention.

2 is in accordance with an embodiment of the present invention

3 is a diagram for describing a morphological relationship between two single ultrasound data with respect to a circular shape.

4 is a diagram for describing a morphological relationship between two single ultrasound data with respect to a line shape.

5 is a diagram for describing a morphological relationship between two single ultrasound data with respect to a point shape.

6 is a schematic diagram illustrating a process in which a moving ultrasonic sensor measures the same object.

7 is a view for explaining a distance probability and a direction probability of data measured using an ultrasonic sensor.

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

10: mobile robot 1: moving unit

3: ultrasonic sensor 4: support plate

5: ultrasonic sensor system

Claims (7)

Initializing the grids of the grid map; Collecting data on surrounding objects using a sensor; Identifying a location of a grid on which the data is to be stored on the grid map; Assigning a shape recognition index according to the degree of agreement of shapes recognized by the data; Weighting the data according to the shape recognition index; Updating the occupancy probabilities for the grid using the weighted data; Object detection method comprising a. According to claim 1, And the sensor is an ultrasonic sensor. The method of claim 2, The ultrasonic sensor is an object detection method mounted on the mobile robot. According to claim 1, When the number of measured data is N, and the number of data representing the same shape among the measured data is n, the shape recognition index NRF satisfies the following equation.

The method of claim 4, wherein The shape is any one selected from the group consisting of circles, lines, points. According to claim 1, Classifying the data collected by the sensor into occupied data and non-occupied data. The method of claim 6, Obtaining a measurement probability for the data.

Images