KR20130118116A - Apparatus and method avoiding collision with moving obstacles in automatic parking assistance system - Google Patents

Abstract

PURPOSE: An automatic parking assistant system is provided to improve the detection performance by detecting the movement of an obstacle using an obstacle image characteristic value and a main component analysis method. CONSTITUTION: An apparatus for avoiding collision with moving obstacles in an automatic parking assistant system which detects the obstacles and avoids the collision when parking a vehicle comprises the following: an automatic parking control unit (160) searching a parking space through an ultrasonic sensor, creating a park route to the park target position based on the searched parking space, and controlling the parking of a vehicle; an obstacle detection unit (140) determining the obstacles based on detection information of the parking space, movement information, and the location of the vehicle; a movement detection unit (150) detecting the movement of the obstacles using machine studying algorithm; and an obstacle collision avoiding control unit (170) recreating the park route to avoid the moving obstacles. [Reference numerals] (110) Camera image unit; (120) Image signal processing unit; (130) Sensor unit; (140) Obstacle detection unit; (150) Movement detection unit; (160) Automatic parking control unit; (170) Obstacle collision avoiding control unit; (AA) Distance from obstacle; (BB) Obstacle movement signal

Description

Apparatus and Method Avoiding Collision with Moving Obstacles in Automatic Parking Assistance System}

The present invention relates to an obstacle collision avoidance apparatus and method, and more particularly to an apparatus and a method for detecting the movement of the obstacle to control to avoid collision with the obstacle.

The conventional obstacle detection system has a problem that the error rate is high when detecting the obstacle, the algorithm for detecting the obstacle is inaccurate, the recognition of the obstacle's movement is impossible, and the collision with the obstacle is not accurately avoided.

In other words, the conventional obstacle detection system has a high error rate for obstacle recognition when detecting obstacles such as failure to detect blind spots such as blind distance detection because only a simple sensor is installed, so when parking a vehicle in a narrow place, The vehicle may be damaged due to a collision with a wall or other vehicle located in the vehicle, and the lower part of the vehicle may be damaged during parking due to a collision with a low object such as a stone even though there is a free space in the parking place.

In addition, the conventional obstacle detection system uses an incorrect detection algorithm such as the MLP method, there is a problem that does not accurately detect the obstacle.

For example, the MLP method minimizes errors due to given sample data, so that the accuracy is high in the training data, but the accuracy is low in the non-learning data, and in an environment where various changes in brightness, position, shape, etc., such as obstacle image detection, occur. There is a problem that it is difficult to secure reliability when the number of sample data is not large.

Conventional obstacle detection system warns the driver by detecting the distance of the obstacle in proximity to the distance sensor mounted on the front and rear bumpers of the vehicle, there is a problem that can not determine the movement of the obstacle.

In addition, the conventional obstacle detection system incorrectly recognizes an obstacle in the parking space due to the obstacle position recognition error of the sensor when the obstacle suddenly appears during the vehicle parking control, and the parking assist system does not properly guide the vehicle to the parking space. The vehicle may collide with a suddenly appearing obstacle or an obstacle in a parking space.

That is, when parking, the driver must detect moving obstacles and pass the obstacles.If other vehicles are parked on the left and right sides of the vehicle in complex or the parking space for passing the obstacles is narrow, the driver detects the movement of the obstacles. If not, the risk of colliding with obstacles is not increased due to failure to maintain a certain distance from the surrounding obstacles. As the driver recognizes the obstacles using only the image of the camera, there is a problem of collision with low obstacle recognition rate in bad weather such as night or fog.

The present invention was created in view of the above-mentioned problems, and is equipped with a plurality of sensors on the front, rear, and side of the vehicle to enlarge an obstacle movement detection area, and to find an obstacle only in an image having a change in brightness instead of finding an obstacle in the entire image. An obstacle collision avoidance apparatus and method are provided in an automatic parking assistance system that detects an area, detects obstacle movements using principal component analysis and obstacle image feature values, and expresses decision boundaries by using a small number of support vectors. Its purpose is to.

In order to achieve the above object, in the automatic parking assist system according to an aspect of the present invention, an obstacle collision avoidance device that detects an obstacle when parking a vehicle and avoids a collision searches for a parking space through an ultrasonic sensor mounted on the vehicle, An automatic parking controller configured to control a parking of the vehicle by generating a parking path to a parking target position based on the found parking space; An obstacle detection unit determining whether an obstacle is present based on the position and movement information of the vehicle and the sensing information about the parking space; An obstacle motion detection unit detecting an obstacle motion using a machine learning algorithm (SVM) when the obstacle is detected by the obstacle detection unit; And an obstacle collision avoidance control unit configured to regenerate the parking path and perform parking control of the vehicle when the obstacle movement is detected so as not to collide with the moving obstacle.

In the automatic parking assist system according to another aspect of the present invention, an obstacle collision avoidance method for detecting an obstacle when a vehicle is parked and avoiding a collision is based on a parking path to a parking target position based on a parking space searched through an ultrasonic sensor mounted on the vehicle. Generating and controlling parking of the vehicle; Determining whether an obstacle is present based on location and movement information of the vehicle and sensing information about the parking space; If it is determined that an obstacle exists in the parking space, detecting an obstacle movement by using a support vector machine (SVM); And when the obstacle movement is detected, regenerating the parking path so as not to collide with the moving obstacle to perform parking control of the vehicle.

According to the present invention, obstacle motions can be detected using principal component analysis and obstacle image feature values, thereby improving detection performance.

It can automatically detect obstacles and detect movement, so the driver does not need to operate it separately.

Obstacle movement detection and collision avoidance can be added to existing devices, minimizing device changes.

Complexity can be reduced by extracting feature vectors of data and classifying them using machine learning algorithms (SVMs).

By using the sensed value and the image information, other obstacles such as lighting or pose can be accurately detected.

By detecting collisions in real time and regenerating the parking path, the risk of collision between a vehicle to be parked and a moving obstacle can be reduced, thereby significantly reducing errors compared to the prior art.

Development costs are low because only an additional sensor and camera are required.

1 is a block diagram illustrating an obstacle collision avoidance apparatus in an automatic parking assist system according to an embodiment of the present invention.
2 is a view for explaining a sensor attached to the vehicle.
3 is a view for explaining FIG. 1 in more detail.
4 is a diagram for explaining an obstacle motion detector.
5 is a flowchart illustrating an obstacle collision avoidance method in an automatic parking assist system according to an embodiment of the present invention.

Advantages and features of the present invention and methods for achieving them will be apparent with reference to the embodiments described below in detail with the accompanying drawings. The present invention may, however, be embodied in many different forms and should not be construed as being limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art. And is intended to enable a person skilled in the art to readily understand the scope of the invention, and the invention is defined by the claims. It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. In the present specification, the singular form includes plural forms unless otherwise specified in the specification. It is noted that " comprises, " or "comprising," as used herein, means the presence or absence of one or more other components, steps, operations, and / Do not exclude the addition.

Hereinafter, an obstacle collision avoidance apparatus in an automatic parking assist system according to an embodiment of the present invention will be described with reference to FIGS. 1 to 4. 1 is a block diagram illustrating an obstacle collision avoidance apparatus in an automatic parking assist system according to an embodiment of the present invention, FIG. 2 is a diagram for explaining a sensor attached to a vehicle, and FIG. 4 is a view for explaining in detail, and FIG. 4 is a view for explaining an obstacle motion detection unit.

As shown in FIG. 1, the obstacle collision avoidance apparatus in the automatic parking assistance system of the present invention includes a camera image unit 110, an image signal processor 120, a sensor unit 130, an obstacle detector 140, and obstacle movement. The detector 150, the automatic parking controller 160, and the obstacle collision avoidance controller 170 are included.

The obstacle detecting unit 140 obtains sensing data (such as a distance to an obstacle) and image data from the camera image unit 110 from the sensor unit 130, and detects an obstacle based on the acquired data. Pass in 150.

The obstacle movement detector 150 may detect an obstacle in real time by improving the processing speed for obstacle movement detection, and may detect an obstacle movement. When the obstacle is detected by the obstacle detector 140, the automatic parking controller The obstacle movement signal is transmitted to 160.

The automatic parking control unit 160 controls the parking of the vehicle by searching the parking space through the sensor.

When the obstacle collision avoidance control unit 170 receives an obstacle motion signal from the obstacle motion detection unit 150, the obstacle collision avoidance control unit 170 regenerates the parking path in real time so that the vehicle to be parked does not collide with the obstacle to perform parking control of the vehicle.

Hereinafter, the obstacle collision avoidance apparatus will be described in detail with reference to FIGS. 2 and 3.

The camera imaging unit 110 acquires image data from the camera and transfers the acquired image data to the obstacle motion detection unit 150 to use it as learning data.

As shown in FIG. 2, the sensor unit 130 includes four front ultrasonic sensors mounted on the front of the vehicle for detecting the movement of the vehicle front obstacle, and two side ultrasonic waves mounted on the side of the vehicle for detecting the movement of the vehicle side obstacle. Sensor, four rear ultrasonic sensors mounted on the rear of the vehicle to detect the movement of the obstacle behind the vehicle, as shown in Figure 3, Steering Angle sensor (steering angle sensor, 131) for detecting the steering direction and steering angle, Wheel Pulse sensor (Wheel pulse sensor 132), Yawrate sensor (radial speed sensor, 133), Temperature sensor (temperature sensor, 134) and Ultrasonic sensor (ultrasound sensor, 135) for obstacle detection.

The steering wheel sensor 131 measures a signal due to the steering angle of the handle, and transmits the measured value to the obstacle detector 140.

The ultrasonic sensor 135 detects side, front and rear obstacles of the vehicle, and transmits the measured value to the obstacle detection unit 140.

The obstacle detecting unit 140 detects the position and the movement of the vehicle in real time through the value measured by the steering wheel sensor 131, and determines the existence of the obstacle through the value measured by the ultrasonic sensor 135. .

When the image signal is input, the obstacle motion detector 150 extracts a feature vector from the input image, determines whether there is a change in brightness in the input image signal based on the extracted feature vector, and determines that there is a change in brightness. The motion detection signal is output after detecting the obstacle area by using a decision boundary of a support vector machine (SVM) that distinguishes the non-obstacle area.

The automatic parking control unit 160 searches the parking space through the sensor and controls the parking of the vehicle by generating a parking path to the parking target position based on the found parking space.

When the motion detection signal is input from the obstacle motion detection unit 150, the obstacle collision avoidance control unit 170 regenerates the parking path so as not to collide with the obstacle to perform parking control of the vehicle.

Hereinafter, the operation of the obstacle motion detection unit will be described in more detail with reference to FIG. 4.

As shown in FIG. 4, the obstacle motion detector 150 using the obstacle feature includes a frame memory 151, a feature value extractor 152, a first threshold value setter 153, a comparator 154, and a value. It includes a fire unit 155, an adder 156, a motion discriminator 157, a motion detector 158 and a subtractor 159.

The frame memory 151 stores the image acquired from the camera image unit 130.

The subtractor 159 obtains a difference image of a brightness value between the stored previous image and the currently input image.

The feature value extractor 152 extracts a feature value for obstacle movement detection.

Feature values here include mean, standard deviation, normal distribution, histogram, and so on.

The first threshold value setting unit 153 sets a pixel boundary value that assumes that there is an obstacle movement from the feature value extraction unit 152.

The comparator 154 compares the difference image of the brightness value of the previous image and the current input image obtained by the subtractor 159 with the first threshold value.

For example, the comparison unit 154 determines that the difference image is greater than the first threshold value when the difference image is greater than the first threshold value, and that the difference image is less than the first threshold value, the pixel having no obstruction movement Judging by.

The binarizing unit 155 allocates +1 and -1 values depending on whether the brightness value changes.

For example, the binarization unit 155 binarizes an image by assigning a +1 value to a pixel having a change in brightness and a -1 value to a pixel having no change.

The adder 156 obtains the total sum of pixels with obstacle movement in the binarized image.

The motion determining unit 157 calculates whether there is a change in brightness value in the entire input image and then determines whether there is an obstacle movement area according to the change.

For example, the motion determining unit 157 determines that there is an obstacle movement in the current input image when the total sum of the moving pixels is greater than the threshold value.

When the motion detector 158 determines that there is an obstacle movement area, the motion detection unit 158 determines whether there is an obstacle characteristic in the portion having the movement area, and if there is an obstacle, outputs an obstacle motion detection signal, and detects the obstacle image from the primarily determined image. If there is an obstacle image, the obstacle movement detection signal is output because there is an obstacle movement finally.

The motion detector 158 includes an M-grid Gabor Wavelet-based candidate region detector (not shown), a low-resolution SVM-based candidate region detector (not shown), and a high-resolution SVM-based candidate region detector (not shown).

The M-grid Gabor Wavelet-based candidate region detection unit (not shown) detects a candidate region in which an obstacle may exist by applying a specific pattern to the region determined to have obstacle movement.

For example, when a camera image sequence is input in units of frames, the M-grid Gabor Wavelet-based candidate region detector (not shown) matches a M like shape grid at all possible positions, extracts a feature vector, and calculates a distance from the average vector. If the distance to the average vector is smaller than the maximum distance obtained by the previous learning, it is determined as an obstacle area and a candidate area in which an obstacle exists is detected.

The low-resolution SVM-based candidate region detector (not shown) performs principal component analysis (PCA, PrincipleComponentAnalysis) on a plurality of obstacle images (NXN pixel sizes) normalized in the previous learning stage to extract feature vectors using N eigen vectors. do.

For example, a low-resolution SVM-based candidate region detector (not shown) reduces the dimension by linearly projecting the feature space in a direction to maximize the variance represented by the image feature vector using the PCA, and obtains the eigenvalue and the feature vector by using the variance. Then, the feature vectors of the desired dimension are extracted by sequential ordering of the eigenvalues, and k is selected whose cumulative contribution (cumulative variance) makes up 99% of the sum of the eigenvalues.

The low-resolution SVM-based candidate region detector (not shown) applies extracted obstacle feature vectors and non-obstacle feature vectors to the SVM to obtain a decision boundary that can distinguish the obstacle class from the non-obstacle class. The grid gabor wavelet-based candidate region detection unit (not shown) searches all possible partial images around the candidate points determined as obstacles and finds the obstacle candidate regions by the previously learned decision boundary.

The high-resolution SVM-based final obstacle detector (not shown) uses a 2N X 2N pixel-size obstacle image in the learning step in a similar manner to the learning and detection step of the low resolution SVM-based candidate area detector (not shown), and uses 2N eigen vectors. In the detection step, a low resolution SVM-based candidate region detector (not shown) is used to detect an obstacle finally using a high resolution SVM around the candidate region determined to be a detected obstacle.

As described above, according to the present invention, it is possible to detect obstacle movement by using the principal component analysis method and the obstacle image feature value to improve the detection performance, and to detect the obstacle motion by using the obstacle image feature value. Intelligent detection can improve auto parking performance, and can store only the necessary images in the future, making efficient use of video storage capacity. In addition, instead of searching for obstacles in the entire image, the obstacle region can be detected only for the image with the change in brightness, which reduces the computation of obstacle detection, and the use of SVM effectively expresses the decision boundary through a small number of support vectors. In this way, the speed of classification of obstacle detection can be improved and high reliability can be guaranteed by obtaining a larger margin.

In the above, the obstacle collision avoidance apparatus has been described in the automatic parking assist system according to an embodiment of the present invention with reference to FIGS. 1 to 4, and the automatic parking assist system according to an embodiment of the present invention is described below with reference to FIG. 5. Explains how to avoid obstacle collision. 5 is a flowchart illustrating an obstacle collision avoidance method in the automatic parking assistance system of the present invention.

As shown in Figure 5, the obstacle collision avoidance method in the automatic parking assistance system of the present invention, when the vehicle position value, obstacle position value, obstacle image data is input from the sensor and the camera (S500), there is an obstacle based on the input information It is determined whether or not (S501).

As a result of the determination, if there is an obstacle, a difference image of the brightness values of the pre-stored image and the input image is obtained (S502), the brightness value change is calculated from the entire input image, and the movement of the obstacle based on the calculated brightness value change. It is determined whether an area exists (S503).

For example, after calculating whether there is a change in brightness value in the entire input image, it is determined whether there is an obstacle movement area according to the change, and if the total sum of the moving pixels is larger than the threshold, there is an obstacle movement in the current input image. Determine.

As a result of the determination, when there is a movement region of the obstacle, it is determined whether there is a characteristic of the obstacle in the entire input image in which the movement region exists, and when it is determined that there is an obstacle characteristic, the movement of the obstacle is detected (S504).

Meanwhile, the parking space is searched through the detected information and the sensor to generate a parking path to the parking target position through the first path calculation (S505), and the steering wheel is moved to move forward or backward based on the generated parking path. Control (S506).

While controlling the vehicle, obstacle movement is detected in real time (S507), and if a collision risk due to the obstacle movement is detected, the final parking target position is extracted through the second route calculation and the final route is calculated to control the obstacle collision risk. Parking control is performed (S508).

Although the configuration of the present invention has been described in detail with reference to the preferred embodiments and the accompanying drawings, this is only an example, and various modifications are possible within the scope without departing from the spirit of the present invention. Therefore, the scope of the present invention should not be limited by the illustrated embodiments, but should be determined by the scope of the appended claims and equivalents thereof.

110: camera video unit 120: video signal processing unit
130 sensor unit 140 obstacle detection unit
150: obstacle movement detection unit 160: automatic parking control unit
170: obstacle avoidance control

Claims (6)

In the obstacle collision avoidance device for avoiding a collision by detecting an obstacle when the vehicle parking in the automatic parking assist system,
An automatic parking control unit for searching a parking space through an ultrasonic sensor mounted on the vehicle, and controlling a parking of the vehicle by generating a parking path to a parking target position based on the found parking space;
An obstacle detection unit determining whether an obstacle is present based on the position and movement information of the vehicle and the sensing information about the parking space;
An obstacle motion detection unit detecting an obstacle motion using a machine learning algorithm (SVM) when the obstacle is detected by the obstacle detection unit; And
When the obstacle movement is detected, an obstacle collision avoidance control unit regenerating the parking path so as not to collide with a moving obstacle to perform parking control of the vehicle.
Obstacle collision avoidance device in an automatic parking assist system comprising a. The method of claim 1,
The obstacle motion detector detects an obstacle motion region by using a decision boundary of the machine learning algorithm only in an image signal having a brightness change by checking a brightness change in an image signal input from the camera of the vehicle.
Collision avoidance device in an automatic parking assist system. 3. The method of claim 2,
The obstacle motion detector calculates a change in brightness value from the input image signal, determines an image signal in which there is a brightness change based on the calculated change in brightness value, and the total sum of the pixels in the brightness change area is greater than a preset threshold. If large, discriminating that there is an obstacle movement in the video signal in which there is a change in brightness
Collision avoidance device in an automatic parking assist system. In the obstacle parking avoidance method for avoiding a collision by detecting an obstacle when the vehicle parking in the automatic parking assist system,
Controlling the parking of the vehicle by generating a parking path to a parking target position based on the parking space searched by the ultrasonic sensor mounted on the vehicle;
Determining whether an obstacle is present based on location and movement information of the vehicle and sensing information about the parking space;
If it is determined that an obstacle exists in the parking space, detecting an obstacle movement by using a support vector machine (SVM); And
If the obstacle movement is detected, regenerating the parking path so as not to collide with a moving obstacle to perform parking control of the vehicle;
Obstacle collision avoidance method in an automatic parking assist system comprising a. The method of claim 4, wherein the detecting of the obstacle movement comprises:
Checking a change in brightness in an image signal input from the camera of the vehicle; And
Detecting an obstacle motion region by using a decision boundary of the machine learning algorithm only in an image signal having a change in brightness in the image signal input based on the verification result;
Collision avoidance method in automatic parking assist system. The method of claim 4, wherein the detecting of the obstacle movement comprises:
Calculating a change in brightness value in the input image signal;
Determining an image signal having a change in brightness based on the calculated change in brightness value; And
If the total sum of the pixels in the brightness change region of the image signal in which there is a brightness change is greater than a predetermined threshold, determining that there is an obstacle movement in the video signal in which the brightness change is present;
Collision avoidance method in automatic parking assist system.

Images