KR20110012317A - Light streaks detection method, target position detection method and parking assist device using the same - Google Patents

Abstract

The present invention relates to a light streaking detection method, a target position detection method and a parking assist device using the same.

According to an embodiment of the present invention, after applying a light plane projection (Indoor Navigation) to the indoor navigation (Indoor Navigation) by detecting a light stripe (Light Stripe) from the image input through the camera, the active steering device And an apparatus for assisting parking of a vehicle by using an electronically controlled braking device, wherein a half value of a light stripe width calculated using a light stripe width function is calculated by a log filter (LOG filter: Laplacian of Gaussian Filter). The difference between the result obtained by log filtering using the constant value of and the image obtained by projecting the light plane with the light plane projector and the image obtained with the light plane projector turned off. The light streaks are obtained by obtaining an intensity histogram of the image and combining the results of detecting data having intensity above a threshold. To provide a parking assistance apparatus using the light stripe detection method characterized in that the detection.

Description

Light Stripe Detection Method, Target position Establishment Method for Indoor Navigation and Parking Assist Apparatus Using Same}

The present invention relates to a light streaking detection method, a target position detection method and a parking assist device using the same. More specifically, in the indoor navigation environment, such as automatic parking in an underground parking lot, a light streaking detection method for detecting a target position to park by properly detecting the light streaks of the light plane projection emitted from the vehicle, the target position detection A method and a parking assist device using the same.

1.1 Recognition of 3D Information by Light Plane Projection

Light Plane Projection Projects the Light Plane using a Light Plane Projector (LPP) mounted on the rear of the vehicle to obtain an image with a camera mounted on the rear of the vehicle. It is a technology that recognizes three-dimensional information by extracting stripes generated on an object from a difference image, which is a difference from.

1 is an exemplary view for explaining a three-dimensional information recognition process using light plane projection.

'O' is the optical center of the camera mounted on the rear of the vehicle, 'xy plane' is the image plane, and 'b' is between the optical center 'O' and the light plane projector in the y-axis direction of the camera. The distance P ₀ is the position of the light plane projector. Π represents the light plane produced by the light plane projector.

The light plane Π meets the Y-axis at point P ₀ (0, -b, 0), the angle between light plane Π and Y-axis is α, the angle between light plane and X-axis is ρ. Form a stripe (Laser Stripe). The corresponding point on the image plane of one point P (X, Y, Z) of the laser stripes is represented by p (x, y), and the coordinates of P are measured using the intersection point of the plane π and the straight line Op. When using a configuration where the light plane is almost parallel to the camera's X-axis, the light stripe occurs only once per image column.

1) Equation of the light side Π

When (0, 1, 0), which is the normal vector of the XZ plane, is rotated by π / 2-α about the X axis and ρ about the Z axis, the normal vector n of π is obtained. Same as

Using the normal vector n of Equation 1 and one point P ₀ on the light plane, the equation of light plane Π can be obtained as in Equation 2.

2) Find Point P on the Laser Stripe

The optical center O, a point p on the image plane, and a corresponding point P in three-dimensional space are all on a straight line. Using a perspective camera model such as Equation 3, all points Q on a straight line may be expressed as Equation 4 by using a parameter k . In this case, f denotes a focal length.

Here, one point P on the laser stripes satisfies Equations 2 and 4 at the same time because it is an intersection point of the light plane Π and a straight line Op. Therefore, by substituting Equations 4 and 1 into Equation 2, the parameter k can be derived as in Equation 5.

In addition, by substituting Equation 5 into Equation 4, the coordinates of the point P can be obtained as in Equation 6, Equation 7, and Equation 8.

In particular, when the light plane and the X axis are parallel (that is, ρ is 0), Equations 6, 7 and 8 can be simplified as in Equation 9, Equation 10 and Equation 11. Note that the distance Z between the camera and the point P on the object is in a one-to-one relationship with the y coordinate of the point on the image.

1.2 Log based Laplacian Of Gaussian

Log or Mexican Hat Wavelets are the most commonly used filters to detect line segments in the field of computer vision.

Figure 2 is an exemplary diagram graphically showing the Mexican hat wavelet function.

The log is defined as in Equation 12, and is a normalized second derivative of a Gaussian function. In addition, the log is a combination of Gaussian Low Pass Filter (LPF) and Peak Enhancement, and the magnitude of the result of convolution with the log depends on the likelihood that the input image is a line. Proportional.

Here, sigma determines the position of the zero crossing point corresponding to the line width. Using σ less than the actual line width detects edges rather than the center of the stripes. Using sigma, which is significantly smaller than the actual line width, ignores the line with the effect of a powerful low pass filter. Since there is only one light stripe per column in the light plane projector, one-dimensional LOG filtering is performed on each column to recognize the point showing the largest output as the stripe.

1.3 Composition of Radiance Map by High Dynamic Range Imaging (HDRi)

Exposure X is defined as the product of Irradiance E and Exposure Time t. Intensity Z is expressed as a non-linear function for exposure X as shown in Equation 13.

Taking Log on both sides of Equation 13 gives Equation 14, and the function g is log f ^{- 1} If it is defined as (5).

In this case, i is an index of pixel coordinates, and j is an index of exposure time during imaging. The nonlinear function that defines the relationship between exposure X and contrast Z is called the response curve of the imaging system.

Debevec defined the criteria for smoothing the response curve while minimizing the error of the equation 15 of the pixels as shown in Equation 16, and presented a method of estimating the least square method (LS Method: Least Square Method). . This means that when a scene is shot at different exposure times, the response curve of the sensor and the radiance map of the scene can be obtained.

An embodiment of the present invention is to detect a target position to park by properly detecting the light streaks of the light surface projection emitted from the vehicle in an indoor navigation environment such as automatic parking in the underground parking lot.

According to an embodiment of the present invention, after applying a light plane projection (Indoor Navigation) to the indoor navigation (Indoor Navigation) by detecting the light stripe (Light Stripe) from the image input through the camera, the active steering In a device that assists parking of a vehicle by using a device and an electronically controlled braking device, a half value of a light stripe width calculated by using a light stripe width function is a log filter (LOG filter: Laplacian of Gaussian Filter). Is a difference between the result obtained by log filtering using a constant value of), and the image obtained by projecting the light plane using the light plane projector and the image obtained when the light plane projector is turned off. Intensity Histogram of the difference image is obtained, and the result of detecting the data having intensity above the threshold is summed and the light line is added. It provides a parking assist device using a light streaks detection method characterized in that the detection of the pattern.

In addition, according to another embodiment of the present invention, by applying a light plane projection (Indoor Navigation) to detect the light stripe (Light Stripe) from the image input through the camera, the active steering device and An apparatus for assisting parking of a vehicle using an electronically controlled braking apparatus, comprising: direction normalization to level the direction of a guideline in the range data constituting the light streaks, and a depth map from the range data where the direction standardization is performed. It provides a parking assist device characterized in that for detecting the target position from the depth map by writing.

In addition, according to another embodiment of the present invention, the parking assistance device connected to the camera, the active steering device and the electronic control braking device to assist the parking of the vehicle, the light plane projection (Light Plane Projection) to the indoor navigation (Indoor Navigation) A method of detecting a light stripe from an image input through a camera by applying a method, the method comprising: (a) constructing a light stripe radiance map using an input image input from the camera ; (b) modeling a parameter of the light streaked radiance map as a function of distance from the camera to an obstacle; (c) calculating a light stripe width function using a parameter of the light stripe radiance map; (d) calculating a light stripe width using the light stripe width function; (e) detecting the first light streaks by log filtering using 1/2 the size of the light streaks width as a constant of a log filter (LOG filter: Laplacian of Gaussian Filter); (f) obtaining a difference image that is a difference between an image obtained by projecting a light plane using a light plane projector and an image obtained while the light plane projector is turned off; (g) obtaining an intensity histogram for the difference image and detecting second light streaks having intensity above a threshold; And (h) combining the first light streaks with the second light streaks.

In addition, according to another embodiment of the present invention, in the target position detection method, (a) applying a light plane projection to the indoor navigation (Indoor Navigation) by applying a light stripe from the image input through the camera (Light Stripe) Detecting; (b) performing direction normalization to level the direction of the guideline in the range data constituting the light streaks; (c) creating a depth map from the range data on which the direction normalization is performed; And (d) detecting a target position from the depth map.

Hereinafter, some embodiments of the present invention will be described in detail through exemplary drawings. In adding reference numerals to the components of each drawing, it should be noted that the same reference numerals are assigned to the same components as much as possible even though they are shown in different drawings. In describing the embodiments of the present invention, when it is determined that the detailed description of the related well-known configuration or function may obscure the gist of the present invention, the detailed description thereof will be omitted.

In addition, in describing the components of the embodiment of the present invention, terms such as first, second, A, B, (a), and (b) may be used. These terms are only for distinguishing the components from other components, and the nature, order or order of the components are not limited by the terms. If a component is described as being "connected", "coupled" or "connected" to another component, that component may be directly connected to or connected to that other component, but there may be another configuration between each component. It is to be understood that the elements may be "connected", "coupled" or "connected".

The light streaks detection method according to an embodiment of the present invention is obtained by projecting the light plane by using the light streaks and the light plane projector extracted by the model-based method and the light plane projector obtained while the light plane projector is turned off. It is obtained by combining the light streaks extracted by the histogram-based method from the difference image, which is a difference from one image.

1. Extracting light streaks by model-based method

1.1 Approximation of Camera Response Curve

3 is an exemplary diagram showing a graph of a response curve according to an experiment.

The response curve of the camera is obtained by using a high contrast ratio image (HDRi: High Dynamic Range Imaging). Since the noise generated as shown in Fig. 3 results from the experiment, the response curve modeled as in Equation 17 is used. Here, the parameter of Equation 17 is estimated by the least square method (LS Method).

1.2 Get Radiance Map of Light Stripe

4 is an exemplary diagram illustrating a process of obtaining a radiance map of light streaks.

4A shows the process of constructing a light streaked radiance map with different exposure images when the light plane projector is turned on, and 4B shows the process of constructing a light streaked radiance map with different exposure images when the light projector is turned on. 4C shows the final light streaked radiance map, 4D shows the light streaked radiance map with distortion correction, and 4E shows the transformation of the light streaked radiance map with distortion correction as an image.

As shown in the figure above, HDRi is applied to the images taken while the Light Plane Projector is turned on and the Exposure Time is changed to compose a radiance map, and when the Light Plane Projector is turned off and the exposure time is changed. HDRi is applied to the images to construct a radiance map. The difference between the two radiance maps thus obtained is the radiance map of the light stripe.

In general, since indoor navigation uses a wide-angle lens, the radial distortion parameters are estimated through a calibration process, and the radial distortion is removed using the wide-angle lens in advance. do.

1.3. 2 Dimensional Gaussian Modeling of Light Stripe Radiance Map

The radiance map of the light streaks follows a two-dimensional normal distribution (Gaussian Distribution) as shown in Equation 18.

To define a two-dimensional normal distribution, five parameters must be estimated. Here, the five parameters are amplitude K , mean value (μ _x , μ _y ), and standard deviation with respect to each axis (σ _x , σ _y ). After estimating the y-axis distribution by the least square method, the x-axis distribution can be estimated to obtain parameter values.

5 is an exemplary diagram showing a light streaked radiality map depicted by the estimated parameter.

As shown in FIG. 5, the light streaked radiance map measured by the estimated parameters is well depicted.

When the light stripe projector and the camera change the distance d to the obstacle, the mean value of the two-dimensional Gaussian parameter does not change, but K , σ _x , and σ _y are represented by Equations 19, 20, It can be modeled as a function of the distance d as shown in Equation 21.

6 is an exemplary view illustrating a result of measuring parameters measured by a mathematical equation.

6A represents measured K , 6B represents measured sigma _x , and 6C represents measured sigma _y . As shown, K , σ _x , σ _y show the results of modeling by equations (19), (20) and (21).

1.4 Light Stripe Width Function

Assuming that the light stripe width is the length of the region greater than the difference in intensity from the surroundings θ _Z , the contrast difference θ _Z is converted into an irradiance difference θ _E by equation (15). Can be. Substituting the irradiance threshold value θ _E into Equation 18, and covering the log, Equation 22 is derived.

In Light Plane Projection, Light Stripes appear only once per column of an image and follow a one-dimensional normal distribution in the y-axis direction.

Accordingly, it can be seen that the y coordinate satisfying θ _E is the y coordinate of the boundary of the light stripe and has the largest irradiance value in the mean value, as shown in Equation 22, and the distance between the y coordinate and the mean value of the boundary is doubled. It can be seen that the light stripe width (Light Stripe Width).

In this case, K, the equation model for the distance σ _x, σ _y d 19, Substituting Equation 20, Equation 21 to Equation 22, on the image x-coordinate x and the distance d as shown in Equation 23 We get the Light Stripe Width function w (x, d).

7 is an exemplary diagram showing the measured light streaks width and the calculated light streaks width.

7A shows the measured light stripe width, and 7B shows the light stripe width calculated by Equation 23.

In the light-striped projection, since distance d (Z in the world coordinate system) has a one-to-one relationship with the y-coordinate of the image, as shown in Equation 11, w (x, d) in Equation 23 is expressed as in Equation 24. The light streaking width function w (x, y) for the coordinate (x, y) on the image can be converted.

8 is an exemplary diagram illustrating a light stripe width and a calculated light stripe width measured at each pixel of an image.

8A represents the light stripe width measured at each pixel (x, y) of the image, and 8B represents the light stripe width calculated by Equation (24).

Therefore, if the configuration of the light plane projection is fixed, and the parameter function of the two-dimensional normal distribution of the light streaked radiance map for the painted wall, which is the main object of indoor navigation, is estimated in advance, Estimate the width of light streaks to appear at image coordinates (x, y)

In the case of indoor driving such as, for example, driving in an underground parking lot, the obstacles in the surroundings have relatively the same reflective characteristics as the paint walls. Therefore, in the following, it is assumed that the obstacle is a homogeneous Lambertian Surface, and that the light streaked radiance map can be modeled as a 2D normal distribution.

The amplitude, x-axis, and y-axis normal distribution of the two-dimensional Gaussian model is a function of distance, and the parameters of this function can be estimated by pre-calibration. For the light intensity threshold θ _Z that can distinguish the stripe from the background, the light stripe width function may be defined using Equation 23 and Equation 24, and the light stripe width for the pixel coordinates (x, y) may be calculated in advance. .

9 is an exemplary diagram illustrating an image of a result of calculating light streaks for all pixels of an image.

Therefore, if the light streaks are detected by using a log filter as shown in Equation 12 in applying light projection to indoor driving, the size of 1/2 of the light streaks width obtained by Equation 22 and Equation 23 is logged. By using the filter as sigma, it is possible to improve the accuracy and recognition rate of the light streaks obtained by log filtering.

2. Histogram Based Light Streaks Extraction

An intensity histogram is obtained for the data of the difference image, and data having intensity greater than or equal to a threshold on the intensity histogram is recognized as light streaks. In this embodiment, the threshold value takes a point of 6σ on the intensity histogram distribution.

FIG. 10 is a diagram illustrating an intensity histogram of data of a difference image in a method of detecting light streaks according to an embodiment of the present invention.

In FIG. 10, the x-axis is brightness and the y-axis is LOG in histogram frequency.

As shown in Fig. 10, data representing brightness above a threshold of a certain size is extracted as light streaks.

FIG. 11A is a diagram illustrating the extraction of histogram-based light streaks, FIG. 11B is a diagram illustrating the extraction of the model-based light streaks, and FIG. 11C is a final formed by combining the histogram-based light streaks and the model-based light streaks. 11D is a diagram showing light streaks, and FIG. 11D is a diagram showing final formed light streaks data (hereinafter referred to as range data) in x and y coordinates.

As described above, the accuracy of the light streaks detection can be improved by using the model-based method and the histogram-based method together in the light streaks detection.

3. Standardization of Direction

In general, parking area marking consists of two types of lines. That is, a guideline separating the road and the parking area, and a straight line drawn between the parking areas, and these two types of straight lines are perpendicular to each other.

Buildings around parking areas such as walls and columns, and the side surfaces of parked vehicles are considered to have nearly the same orientation as these two types of lines.

Evaluate the guidelines of the range data and rotate the range data so that the guidelines are horizontal. Thus, the depth map formation and free space detection described later can be made robust and simple.

A weighted orientation histogram (WOH) is used to estimate the direction of the guideline.

The direction is obtained for each pair of range data, and the direction is added to the histogram of the WOH to accumulate the angle with respect to the direction of any range data pair to form the WOH. At this time, rather than add an unconditional one, when added to the histogram Destiny in each vector forming from the origin pair range data Nation (that is, (V = P _n2 with respect to the direction of the obtained range data pairs means the P _n2 in a vector of P _n1) ) Divided by the square of the distance to the histogram of the angle. This reduces the weight of the data pairs in the long distance and increases the weight of the data pairs in the short distance, thereby reducing the influence of noise on the long range data and improving the accuracy of the direction extraction.

The algorithm for obtaining WOH is shown as pseudo-code as follows.

for n ₁ = 1: (N-1)

for n ₂ = (n ₁ +1): N

V = P _n2 -P _n1

θ = angle (V)

WOH (θ) + = 1 / (| P _n2 |) ² )

end

end

Here, N means the number of lazy data, P _n1 and P _n2 means the coordinates for the N-th range data.

FIG. 12 illustrates WOH formed from the range data of FIG. 11D through an algorithm for obtaining WOH.

If the driver selects the target location for the vehicle to the left or right, it detects the peak value between 0 ° and 90 ° on the right side and the peak value between 90 ° and 180 ° on the left side. Therefore, since the range data of FIG. 11D is a place located on the left side of the vehicle's driver basis, the range data is found in the range of 90 ° to 180 ° (indicated by x in FIG. 12). Therefore, the range data is rotated so that the guideline direction of the range data becomes horizontal according to the extracted angle, and the guideline is recognized.

FIG. 13A is a diagram showing the direction of the range data (denoted by θ), and FIG. 13B is a diagram illustrating the standardization of the direction of the range data and the recognition of a guideline. Therefore, when the WOH has a peak value in the range of 90 ° to 180 °, as shown in FIG. 13B, the range data is rotated counterclockwise by (180-θ).

In FIG. 13B, the guideline recognizes as a guideline the first y-coordinate where a certain number or more of range data is found with respect to the y-coordinate.

4. Depth Lead Forming

The depth map is formed by finding range data having the smallest y coordinate for each x coordinate of the range data whose direction is normalized as shown in FIG. 13. In the case of the direction normalized range data, several range data may exist for each x coordinate. Therefore, the smallest y coordinate can be selected and formed for the range data of each x coordinate. At this time, in order to reduce the influence of noise, the range data having an extremely short width of less than a predetermined width may be deleted during the setting of the depth map. On the other hand, two consecutive range data can be separated by a certain distance with respect to the x-axis (ie, discontinuous). In this case, the depth map can also be set to give the y coordinate interpolated between the two points.

FIG. 14 illustrates a depth map formed with respect to the range data of FIG. 13.

As shown in FIG. 14, it can be seen that a depth map represented by a red line forms a free space having a predetermined size in the range data.

5. Set target position

The free space is detected in consideration of the width and length of the vehicle, and when several free spaces are detected, the nearest free space is selected as the target position. When the target position is to the left of the vehicle, free space search is performed from left to right on the depth map. If on the right side of the vehicle, the search direction is reversed, ie from right to left of the depth map.

In detecting the free space, whether or not a particular x coordinate of the depth map forms a free space is determined by checking whether the corresponding y coordinate is dustier than the length of the target vehicle from the guideline. In this way, the method connects the points of the x coordinate of the range data when the y coordinate is farther than the length of the target vehicle from the guideline, and this method is applied to the x coordinate up to the range data having the y coordinate shorter than the vehicle length. Recognize it as an x-axis component. When the length of the detected x-axis component of the free space is wider than the width of the vehicle, the detected free space is recognized as a valid parking space, and the position of the nearest free space is set as the target position.

15A is a diagram illustrating a target position detected in the depth map of FIG. 14, and FIG. 15B is a diagram illustrating a target position converted to a vehicle coordinate system. If the target position is set in the depth map, the target position may be converted to the vehicle coordinate system as shown in FIG. 15B by applying rotation in a direction opposite to the rotation direction at the time of normalizing the direction.

On the other hand, since the x-axis field of view (FOV) of the camera mounted on the rear of the vehicle is limited, sometimes detecting both sides of the free space may fail depending on the position of the vehicle.

FIG. 16A illustrates an image obtained by a camera on a rear surface of a vehicle when a light plane projector emits light, FIG. 16B illustrates range data detected in the situation of FIG. 16A, and FIG. 16C illustrates detected range. The depth map is set after directional normalization of data.

As shown in Fig. 16C, the left side of the free space is not detected in the set depth map. Since both sides of the free space are not detected, the target position cannot be established based on the method described previously. In this situation, when the free space between two objects cannot be detected by not obtaining the x coordinate up to the range data having the y coordinate shorter than the vehicle length until the free space detection is completed, the free space detection is completed. If the length of the x-axis component of the free space detected so far is wider than the width of the vehicle, it can be set as the target position.

16D illustrates a target position set in the depth map.

As shown in Fig. 16D, the free space can be accurately detected even without range data for one side of the free space.

6. Rewards for Hidden Vehicles

 Since the side surface of the parked vehicle is assumed to be horizontal or perpendicular to the direction normalized range data, the vehicle may disappear from the depth map if the front surface of the vehicle is insufficiently detected.

17A is a view showing an image obtained by the camera on the rear of the vehicle when the light is projected to emit light when the vehicle parked on the left and right sides of the free space is correctly arranged in accordance with the direction of the parking zone, and FIG. 17B is FIG. 17A In this situation, the depth map is shown.

As shown in Fig. 17A, when the vehicles parked on the left and right sides of the free space are arranged correctly in accordance with the direction of the parking area, the range data of the side surface of each vehicle is represented by a perpendicular line as shown in Fig. 17B. To reduce the effects of noise, range data with too short a width is deleted during setting of the depth map, and range data on a right angle tends to be projected on a very narrow width, so range data on the side surfaces of parked vehicles Can be deleted during depth map setup.

In order to solve this problem, an x-axis histogram is created for the x-axis of the direction-normalized range data, and it is recognized as the side of the vehicle when it appears on the x-axis histogram on the x-axis histogram with a predetermined number or more for the x-axis having a predetermined width or less.

FIG. 17C is a diagram showing an x-axis histogram of the range data of FIG. 17B.

As shown in FIG. 17C, when the reference temperature is greater than or equal to, it may be recognized as the side of the parked vehicle. In the case of 'b' and 'c', it is already reflected in the depth map, so 'b' which is not reflected in the depth map is recognized as the side of the hidden vehicle and has a certain width and length in the direction where the depth map does not exist. It is assumed that a parked vehicle exists so that it can be reflected in the depth map.

FIG. 18A illustrates a depth map created in consideration of a hidden vehicle having a predetermined width and length, and FIG. 18B illustrates a target position set based on the depth map.

By reinforcing the depth map with the detected hidden vehicle as in FIG. 18, the loss of the parked vehicle, which may occur during parking in the event of emergency, can be avoided, and the target position can be set correctly as in FIG. 18B.

19 is a block diagram schematically illustrating a parking assist device according to an embodiment of the present invention.

Parking assistance apparatus 1920 according to an embodiment of the present invention is connected to the camera 1910, the active steering device 1930 and the electronically controlled braking device 1940, the obstacle of the surrounding environment using the image input from the camera And assists parking of the vehicle by steering and braking the vehicle using the active steering device 1930 and the electronically controlled braking device 1940.

Here, the active steering device 1930 is a steering assist means for assisting the steering of the vehicle by grasping the driving situation of the vehicle and the driver's driving intention, and includes an electronic power steering (EPS) and a motor driving steering (MDPS). The concept includes Motor Driven Power Steering (AFS) and Active Front Steering (AFS).

In addition, the electronically controlled braking device 1940 is a brake control means for converting the braking state of the vehicle, and includes an anti-lock brake system (ABS), an automatic stability control (ASC) system, and dynamic stability. This concept includes the Dynamic Stabilty Control (DSC) system.

The parking assisting apparatus 1920 according to an embodiment of the present invention applies light plane projection to indoor driving to detect light streaks from an image input through a camera, and assists parking by detecting obstacles using light streaks. .

In addition, in the parking assisting device 1920 according to an embodiment of the present invention, in detecting light streaks, the half of the light streaking width calculated using a light stripe width function is a log filter. Filter: The result obtained by log filtering using the constant value of the Laplacian of Gaussian Filter, and the image obtained by projecting the light plane with the light plane projector and the light plane projector are obtained. An intensity histogram is obtained for a difference image that is a difference from one image, and light streaks are detected by adding the results of detecting data having an intensity greater than or equal to a threshold value.

In addition, the parking assist device 1920 according to an embodiment of the present invention, when modeling a light stripe radiance map (2D normal distribution) in the light plane projection, the camera of the parameters of the light stripe radiance map It is modeled as a function of the distance from the path to the obstacle.

The parking assisting device 1920 according to an embodiment of the present invention performs direction standardization so that the direction of the guideline is horizontal in the range data constituting the light streaks, and creates a depth map from the range data where the direction standardization is performed. Detect the target location from the map.

In this case, the direction normalization applies a weighted direction histogram to calculate the direction of the guideline in the range data, and if the direction in which the vehicle is parked is the right side of the vehicle, the peak value from 0 ° to 90 ° is detected and left on the weighted direction histogram. After detecting the peak value between 90 ° and 180 °, the direction of the guideline is determined, and the range data is rotated so that the guideline is horizontal.

Meanwhile, data having a short width less than or equal to a predetermined width in the depth map may be deleted from the depth map. If it is more than the standard tap water can be set by recognizing the side of the vehicle.

20 is a flowchart illustrating a light streaks detection method according to an embodiment of the present invention.

The above-described parking assisting device 1920 detects an obstacle in an indoor driving environment such as an underground parking lot to assist in driving or parking the vehicle, and detects an obstacle by applying a light projection to indoor driving to detect light streaks. do.

That is, the parking assisting device 1920 projects the light surface to the indoor driving environment using the light surface projector mounted on the camera 1910 (S2010), and receives an input image input from the camera 1910 (S2020). A light streaked radiance map is constructed from the image (S2030).

The parking assist device 1920 that composes the light streaked radiance map models the parameter of the light streaked radiance map as a function of the distance from the camera to the obstacle (S2040), and uses the model of the light streaked radiance map to model the light. The stripe width function is calculated (S2050).

The parking assist device 1920 calculates the light streaks width by using the light streak width function, and detects the light streaks by using the 1/2 size of the calculated light streaks width as a constant of the log filter for detecting line segments, that is, σ. (S2060).

The parking assisting device 1920 obtains a difference image that is a difference between an image obtained by projecting a light plane using a light plane projector and an image obtained when the light plane projector is turned off (S2070).

The parking assisting device 1920 obtains an intensity histogram of the vehicle image and detects a second light stripe having an intensity greater than or equal to a threshold (S2080).

The parking assistance device 1920 extracts the light streaks by combining the first light streaks and the second light streaks (S2090).

21 is a flowchart illustrating a target position detection method according to an embodiment of the present invention.

In order to detect the target position, the parking assisting device 1920 applies a light plane projection to indoor navigation to detect a light stripe from an image input through a camera (S2110).

The parking assisting device 1920 performs direction normalization so that the direction of the guideline is horizontal in the range data constituting the light streaks (S2120).

The parking assisting device 1920 creates a depth map from the range data where the direction standardization is performed (S2130).

The parking assisting device 1920 detects a target position from the depth map (S2140).

As described above, according to the present invention, when parking a vehicle, it is possible to recognize a location that can be parked even in a dark parking lot, thereby improving the parking safety of the driver, thereby increasing the safety of parking.

In the above description, all elements constituting the embodiments of the present invention are described as being combined or operating in combination, but the present invention is not necessarily limited to the embodiments. That is, all of the components may operate selectively in combination with one or more of them. In addition, although all of the components may be implemented in one independent hardware, each or all of the components may be selectively combined to perform some or all functions combined in one or a plurality of hardware. It may be implemented as a computer program having a. Codes and code segments constituting the computer program may be easily inferred by those skilled in the art. Such a computer program may be stored in a computer readable storage medium and read and executed by a computer, thereby implementing embodiments of the present invention. The storage medium of the computer program may include a magnetic recording medium, an optical recording medium, a carrier wave medium, and the like.

In addition, the terms "comprise", "comprise", or "having" described above mean that the corresponding component may be embedded unless otherwise stated, and thus, other components. It should be construed that it may further include other components rather than to exclude them. All terms, including technical and scientific terms, have the same meanings as commonly understood by one of ordinary skill in the art unless otherwise defined. Terms commonly used, such as terms defined in a dictionary, should be interpreted to coincide with the contextual meaning of the related art, and shall not be construed in an ideal or excessively formal sense unless explicitly defined in the present invention.

The foregoing description is merely illustrative of the technical idea of the present invention, and various changes and modifications may be made by those skilled in the art without departing from the essential characteristics of the present invention. Therefore, the embodiments disclosed in the present invention are not intended to limit the technical idea of the present invention but to describe the present invention, and the scope of the technical idea of the present invention is not limited by these embodiments. The protection scope of the present invention should be interpreted by the following claims, and all technical ideas within the equivalent scope should be interpreted as being included in the scope of the present invention.

1 is an exemplary view for explaining a three-dimensional information recognition process using light plane projection,

2 is an exemplary diagram showing a Mexican hat wavelet function graphically;

3 is an exemplary diagram showing a graph of a response curve according to an experiment;

4 is an exemplary diagram illustrating a process of obtaining a radiance map of light streaks;

5 is an exemplary diagram showing a light streaked radiality map depicted by the estimated parameter;

6 is an exemplary view illustrating a result of measuring parameters measured by a mathematical equation;

7 is an exemplary view showing the measured light streaks width and the calculated light streaks width,

8 is an exemplary diagram illustrating a light stripe width and a calculated light stripe width measured at each pixel of an image;

9 is an exemplary diagram showing an image of a result of calculating light streaks for all pixels of an image.

FIG. 10 is a diagram illustrating an intensity histogram of data of a difference image in a method of detecting light streaks according to an embodiment of the present invention; FIG.

FIG. 11A is a diagram illustrating the extraction of histogram-based light streaks, FIG. 11B is a diagram illustrating the extraction of the model-based light streaks, and FIG. 11C is a final formed by combining the histogram-based light streaks and the model-based light streaks. FIG. 11D is a diagram showing light streaks, and FIG. 11D is a diagram showing final formed light streaks data (hereinafter, referred to as range data) in x and y coordinates.

12 illustrates a WOH formed from the range data of FIG. 11D through an algorithm for obtaining a WOH;

FIG. 13A is a diagram illustrating the direction of range data (denoted by θ), and FIG. 13B is a diagram illustrating the standardization of the direction of range data and recognition of a guideline;

14 illustrates a depth map formed with respect to the range data of FIG. 13;

15A is a diagram illustrating a target position detected in the depth map of FIG. 14, and FIG. 15B is a diagram illustrating a target position converted to a vehicle coordinate system.

FIG. 16A illustrates an image obtained by a camera on a rear surface of a vehicle when a light plane projector emits light, FIG. 16B illustrates range data detected in the situation of FIG. 16A, and FIG. 16C illustrates detected range. FIG. 16D illustrates a depth map set after normalizing data in a direction direction, FIG. 16D illustrates a target position set on a depth map;

17A is a view showing an image obtained by the camera on the rear of the vehicle when the light is projected to emit light when the vehicle parked on the left and right sides of the free space is correctly arranged in accordance with the direction of the parking zone, and FIG. 17B is FIG. 17A FIG. 17C is a diagram illustrating an x-axis histogram of the range data of FIG. 17B.

FIG. 18A illustrates a depth map created in consideration of a hidden vehicle having a predetermined width and length, and FIG. 18B illustrates a target position set based on the depth map;

19 is a block diagram schematically illustrating a parking assisting device according to an embodiment of the present invention;

12 is a flowchart illustrating a light streaking detection method for indoor driving according to an embodiment of the present invention.

<Description of Symbols for Main Parts of Drawings>

1910: camera 1920: parking assist device

1930: active steering 1940: electronically controlled brake

Claims (9)

Light Plane Projection is applied to Indoor Navigation to detect Light Stripe from the image input through the camera and to detect obstacles, and then to use Active Steering System and Electronic Control Braking System. In the device for assisting the parking of the vehicle, The result obtained by log filtering using the half value of the light stripe width calculated using the Light Stripe Width Function as a constant value of the Log Filter (Laplacian of Gaussian Filter), and the light plane projector Intensity Histogram is obtained for the difference image between the image obtained by projecting the light plane using the Plane Projector and the image obtained with the light plane projector turned off. Parking assistance device using the light streaks detection method characterized in that for detecting the light streaks by adding the result of detecting the data having. According to claim 1, The parking assist device, When modeling a light stripe radiance map as a two-dimensional normal distribution in a light projection, the parameter of the light stripe radiance map is modeled as a function of the distance from the camera to the obstacle. Parking aids. The method of claim 1, wherein the light stripe width function is Parking assistance device, characterized in that calculated using a threshold for distinguishing the light streaks from the background. Light Plane Projection is applied to Indoor Navigation to detect Light Stripe from the image input through the camera, and then parking the vehicle using the active steering and electronic brake system. In the assisting device, Directional standardization is performed to level the direction of the guideline in the range data constituting the light streaks, and a depth map is generated from the range data where the direction standardization is performed to detect a target position from the depth map. Auxiliary devices. The method of claim 4, wherein The direction standardization, In the range data, a weighting direction histogram is applied to calculate the direction of the guideline, and if the direction in which the vehicle is parked is the right side of the vehicle, a peak value between 0 ° and 90 ° is detected in the weighting direction histogram; And detecting the peak value at ˜180 ° to obtain the direction of the guideline, and then rotating the range data to level the guideline. The method of claim 4, wherein The depth map is, Parking assistance device, characterized in that the data having a short width less than a certain width is deleted from the depth map. The method of claim 6, The depth map may be configured by recognizing the side surface of the vehicle when the number of degrees for data having a short width less than or equal to the predetermined width is greater than a reference number by creating an x-axis histogram of the direction normalized data. . A parking assist device connected to a camera, an active steering device and an electronically controlled braking device to assist the parking of the vehicle is applied to the indoor navigation by applying a light plane projection to the indoor navigation. In the method of detecting (Light Stripe), (a) constructing a light stripe radiance map using an input image input from the camera; (b) modeling a parameter of the light streaked radiance map as a function of distance from the camera to an obstacle; (c) calculating a light stripe width function using a parameter of the light stripe radiance map; (d) calculating a light stripe width using the light stripe width function; (e) detecting the first light streaks by log filtering using 1/2 the size of the light streaks width as a constant of a log filter (LOG filter: Laplacian of Gaussian Filter); (f) obtaining a difference image that is a difference between an image obtained by projecting a light plane using a light plane projector and an image obtained while the light plane projector is turned off; (g) obtaining an intensity histogram for the difference image and detecting second light streaks having intensity above a threshold; And (h) combining the first light streaks with the second light streaks Light streaks detection method comprising a. In the target position detection method, (a) detecting a light stripe from an image input through a camera by applying a light plane projection to indoor navigation; (b) performing direction normalization to level the direction of the guideline in the range data constituting the light streaks; (c) creating a depth map from the range data on which the direction normalization is performed; And (d) detecting a target position from the depth map Target position detection method comprising a.

Images