KR102586899B1 - Sliding window correction method and device for platooning lane detection - Google Patents

Abstract

The present invention relates to a sliding window double correction method and device, the method comprising: generating a binary image based on an original image for a current frame of a driving image including at least a lane; determining a base point based on the x-axis on the binary image; generating a first sliding window of a preset size centered on the base point; and generating, based on the first sliding window, a second sliding window adjacent to the first sliding window and accumulating in the y-axis direction on the binary image.

Description

Sliding window double correction method and device for platooning lane recognition {SLIDING WINDOW CORRECTION METHOD AND DEVICE FOR PLATOONING LANE DETECTION}

The present invention relates to a sliding window double correction technology, and more specifically, to a sliding window double correction technology that corrects positioning errors caused by an empty sliding window through the previous sliding window of the corresponding frame or the normal sliding window of the previous frame. .

Research on truck platooning systems, a field of self-driving cars, is steadily progressing. Platoon driving is receiving a lot of attention for reasons such as reduced fuel efficiency, reduced carbon dioxide emissions, and improved safety, and is attracting attention as an alternative to solving the problem of manpower shortage in the truck industry.

Truck platooning allows multiple trucks to drive while maintaining a narrow distance based on V2X communication, and when a 0.5-second time-gap is applied based on the maximum speed of highway trucks at 90 km/h, the inter-vehicle distance can be maintained at 12.5 m. At this time, the FV (Following Vehicle) may have a lot of visibility blocked by the LV (Leading Vehicle). This can make the autonomous driving cognitive process more difficult, and can especially have a significant impact on the lane cognitive process required for lateral control. General lane recognition can be done based on data from the image received through the camera sensor to the vanishing point. However, in the case of platooning, most lanes are obscured by the LV, and as a result, lane detection can be performed using only limited data.

Many of the lanes on highways where platooning occurs are dotted lines, and the dotted line section can have a length of 8m and an interval between dotted lines of 12m. In a platooning situation where the distance between vehicles is reduced to 12.5m, situations where no lanes occupy most of the obtained data may continue to occur. Suboptimal recognition using only limited data like this can ultimately lead to a decrease in accuracy. The lane recognition process for platooning can be accomplished through the same process as general lane recognition, and positioning errors can occur during the sliding window process among the various lane recognition processes.

Korean Patent Publication No. 10-2016-0072659 (2016.06.23)

One embodiment of the present invention seeks to provide a sliding window double correction technology that corrects positioning errors caused by an empty sliding window through the previous sliding window of the corresponding frame or the normal sliding window of the previous frame.

Among embodiments, the sliding window dual correction method includes generating a binary image based on an original image for a current frame of a driving image including at least a lane; determining a base point based on the x-axis on the binary image; generating a first sliding window of a preset size centered on the base point; and generating, based on the first sliding window, a second sliding window adjacent to the first sliding window and accumulating in the y-axis direction on the binary image.

Generating the binary image includes removing noise from the original image; setting a region of interest regarding the lane on the original image; converting the image of the region of interest into a bird's eye view image; and converting the bird's eye view image into the binary image.

Determining the base point may include determining the base point as a reference for each lane through a histogram representing the number of lane pixels at each coordinate on the binary image.

The step of generating the second sliding window may include generating a second sliding window of the preset size centered on the average value of x-axis coordinates of lane data existing in the first sliding window.

Generating the second sliding window may include determining the center of the second sliding window through first sliding window correction when the lane data does not exist in the first sliding window.

Generating the second sliding window may include performing first sliding window correction when the lane data existing within the first sliding window is less than a preset threshold.

The step of generating the second sliding window includes calculating an average value of x-axis coordinates of the lane data for each of the first sliding windows created in the previous step; calculating a slope according to a change in the average value; and determining the center of the second sliding window by applying the tilt to the center of the most recent first sliding window.

The step of generating the second sliding window may include determining the center of the second sliding window through second sliding window correction when the number of first sliding windows created in the previous step is less than 2.

The step of generating the second sliding window is performed when the number of first sliding windows created in the previous step is 2 and the lane data does not exist for one of the first sliding windows or the lane data is preset. If it is less than the threshold, it may include performing a second sliding window correction.

Generating the second sliding window may include resetting the center and the slope of the first sliding window based on data values stored in a previous frame of the current frame; and determining the center of the second sliding window by applying the reset slope to the center of the reset first sliding window.

Among embodiments, a sliding window dual correction device includes a frame image extractor that generates a binary image based on an original image for a current frame of a driving image including at least a lane; a base point determination unit that determines a base point based on the x-axis in the binary image; a sliding window generator that generates a first sliding window of a preset size centered on the base point; and a sliding window accumulator that generates, based on the first sliding window, a second sliding window adjacent to the first sliding window and accumulating in the y-axis direction on the binary image.

The sliding window accumulator may generate a second sliding window of the preset size centered on an average value of x-axis coordinates of lane data existing in the first sliding window.

When the lane data does not exist within the first sliding window, the sliding window accumulator may determine the center of the second sliding window through first sliding window correction.

If the number of first sliding windows created in the previous step is less than 2, the sliding window accumulator may determine the center of the second sliding window through second sliding window correction.

The disclosed technology can have the following effects. However, since it does not mean that a specific embodiment must include all of the following effects or only the following effects, the scope of rights of the disclosed technology should not be understood as being limited thereby.

The sliding window double correction method and device according to an embodiment of the present invention can correct positioning errors caused by an empty sliding window through the previous sliding window of the corresponding frame or the normal sliding window of the previous frame.

The sliding window double correction method and device according to an embodiment of the present invention can improve lane recognition accuracy by solving lane recognition problems that may occur in curved sections.

1 is a diagram illustrating a sliding window double correction system according to the present invention.
FIG. 2 is a diagram explaining the system configuration of the dual correction device of FIG. 1.
FIG. 3 is a diagram explaining the functional configuration of the dual correction device of FIG. 1.
Figure 4 is a flowchart explaining the sliding window double correction method according to the present invention.
FIG. 5 is a diagram illustrating an example of a position setting error related to a sliding window.
Figure 6 is a diagram explaining the sliding window double correction process according to the present invention.
Figure 7 is a diagram explaining an experimental example related to the present invention.

Since the description of the present invention is only an example for structural or functional explanation, the scope of the present invention should not be construed as limited by the examples described in the text. In other words, since the embodiments can be modified in various ways and can have various forms, the scope of rights of the present invention should be understood to include equivalents that can realize the technical idea. In addition, the purpose or effect presented in the present invention does not mean that a specific embodiment must include all or only such effects, so the scope of the present invention should not be understood as limited thereby.

Meanwhile, the meaning of the terms described in this application should be understood as follows.

Terms such as “first” and “second” are used to distinguish one component from another component, and the scope of rights should not be limited by these terms. For example, a first component may be named a second component, and similarly, the second component may also be named a first component.

When a component is referred to as being “connected” to another component, it should be understood that it may be directly connected to the other component, but that other components may also exist in between. On the other hand, when a component is referred to as being “directly connected” to another component, it should be understood that there are no other components in between. Meanwhile, other expressions that describe the relationship between components, such as "between" and "immediately between" or "neighboring" and "directly neighboring" should be interpreted similarly.

Singular expressions should be understood to include plural expressions unless the context clearly indicates otherwise, and terms such as “comprise” or “have” refer to implemented features, numbers, steps, operations, components, parts, or them. It is intended to specify the existence of a combination, and should be understood as not excluding in advance the possibility of the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

For each step, identification codes (e.g., a, b, c, etc.) are used for convenience of explanation. The identification codes do not explain the order of each step, and each step clearly follows a specific order in context. Unless specified, events may occur differently from the specified order. That is, each step may occur in the same order as specified, may be performed substantially simultaneously, or may be performed in the opposite order.

The present invention can be implemented as computer-readable code on a computer-readable recording medium, and the computer-readable recording medium includes all types of recording devices that store data that can be read by a computer system. . Examples of computer-readable recording media include ROM, RAM, CD-ROM, magnetic tape, floppy disk, and optical data storage devices. Additionally, the computer-readable recording medium can be distributed across computer systems connected to a network, so that computer-readable code can be stored and executed in a distributed manner.

All terms used herein, unless otherwise defined, have the same meaning as commonly understood by a person of ordinary skill in the field to which the present invention pertains. Terms defined in commonly used dictionaries should be interpreted as consistent with the meaning they have in the context of the related technology, and cannot be interpreted as having an ideal or excessively formal meaning unless clearly defined in the present application.

1 is a diagram illustrating a sliding window double correction system according to the present invention.

Referring to FIG. 1, the sliding window double correction system 100 may include a user terminal 110, a double correction device 130, and a database 150.

The user terminal 110 may correspond to a computing device operated by a user, and may be implemented as a terminal installed in a running vehicle to collect images of the front area including lanes on the road, but is not necessarily limited thereto. It can also be implemented with various devices such as smartphones, laptops, computers, or tablet PCs. The user terminal 110 may be implemented to include a camera sensor for image capture, and may be implemented to operate in conjunction with various camera modules installed in a vehicle.

Additionally, the user terminal 110 may be connected to the double correction device 130 through a network, and a plurality of user terminals 110 may be connected to the double correction device 130 at the same time. Additionally, the user terminal 110 may install and execute a dedicated program or application for linking with the double correction device 130.

The double correction device 130 may be implemented as a system that executes the sliding window double correction method according to the present invention, or as a server corresponding thereto. The double correction device 130 can be connected to the user terminal 110 through a network and exchange related data. Additionally, the dual correction device 130 may operate in conjunction with at least one external system. For example, external systems can implement various functions and services by linking with a cloud server for cloud services, an artificial intelligence server for deep learning learning, and an authentication server for user authentication.

In one embodiment, the dual correction device 130 may store and manage various data necessary in the process of recognizing a lane from a vehicle driving image collected from the user terminal 110 in conjunction with the database 150. Additionally, the double correction device 130 may be implemented including a processor, memory, user input/output unit, and network input/output unit, which will be described in more detail in FIG. 2.

The database 150 may correspond to a storage device that stores various information required during the operation of the double correction device 130. For example, the database 150 can store information about driving images captured while driving a vehicle through the user terminal 110, and contains information about the algorithm for lane recognition and various operations performed in the lane recognition process. It is not necessarily limited to this, and information collected or processed in various forms during the sliding window double correction process for platooning lane recognition can be stored.

FIG. 2 is a diagram explaining the system configuration of the dual correction device of FIG. 1.

Referring to FIG. 2, the double correction device 130 may be implemented including a processor 210, a memory 230, a user input/output unit 250, and a network input/output unit 270.

The processor 210 may execute a procedure that processes each step in the process of operating the double correction device 130, and manage the memory 230 that is read or written throughout the process. ) can schedule the synchronization time between volatile and non-volatile memory. The processor 210 can control the overall operation of the double correction device 130 and is electrically connected to the memory 230, the user input/output unit 250, and the network input/output unit 270 to control the data flow between them. You can. The processor 210 may be implemented as a central processing unit (CPU) of the dual correction device 130.

The memory 230 may be implemented as a non-volatile memory such as a solid state drive (SSD) or a hard disk drive (HDD) and may include an auxiliary memory used to store all data required for the dual correction device 130, It may include a main memory implemented as volatile memory such as RAM (Random Access Memory).

The user input/output unit 250 may include an environment for receiving user input and an environment for outputting specific information to the user. For example, the user input/output unit 250 may include an input device including an adapter such as a touch pad, touch screen, on-screen keyboard, or pointing device, and an output device including an adapter such as a monitor or touch screen. In one embodiment, the user input/output unit 250 may correspond to a computing device connected through a remote connection, in which case the dual correction device 130 may be performed as a server.

The network input/output unit 270 includes an environment for connecting with external devices or systems through a network, for example, Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), and VAN ( It may include an adapter for communication such as a Value Added Network).

FIG. 3 is a diagram explaining the functional configuration of the dual correction device of FIG. 1.

Referring to FIG. 3, the double correction device 130 includes a frame image extraction unit 310, a base point determination unit 330, a sliding window creation unit 350, a sliding window accumulation unit 370, and a control unit 390. It can be included.

The frame image extractor 310 may generate a binary image based on the original image for the current frame of the driving image including at least a lane. The frame image extractor 310 may be connected to the user terminal 110 installed and operating in the vehicle, and may receive driving images captured by the user terminal 110 and store them in the database 150. At this time, the driving image may correspond to a front image of the vehicle, and in particular, may correspond to a driving image captured to include lanes marked on the road.

Additionally, the frame image extractor 310 can analyze the driving image to selectively extract a driving image including a lane, and generate a plurality of image frames by decomposing the extracted driving image into frames. The frame image extractor 310 may specify and extract an area containing a lane in order to increase computational performance for one image frame. That is, the frame image extractor 310 may extract the image area containing the lane in the previous step for lane recognition, generate a binary image based on this, and transmit it to the lane recognition step.

In one embodiment, the frame image extractor 310 removes noise from the original image, sets a region of interest regarding the lane on the original image, and converts the image of the region of interest into a bird's eye view image, A bird's eye view image can be converted to a binary image. That is, the frame image extractor 310 may perform a preprocessing operation on the original image to derive recognition results with higher accuracy.

More specifically, the frame image extractor 310 can set a region of interest (ROI) for the entire image to remove noise and improve processing speed. Thereafter, the frame image extractor 310 may convert the image selected through ROI into a bird's eye view image for effective lane detection. Additionally, the frame image extractor 310 can convert a bird's eye view image into a binary image through edge detection and comparing color saturation differences to specify a lane in the converted image.

The base point determination unit 330 may determine the base point based on the x-axis in the binary image. Here, the base point may correspond to a reference point for creating a sliding window. For example, the base point determination unit 330 may determine the base point through a sliding window technique and a histogram.

In one embodiment, the base point determination unit 330 may determine the base point as a reference for each lane through a histogram representing the number of lane pixels at each coordinate in the binary image. The sliding window generator 350 may generate a first sliding window of a preset size centered on the base point. That is, the sliding window generator 350 may generate a first sliding window that is spaced apart from a base point by a predetermined distance in the left and right directions.

At this time, the first sliding window may correspond to a sliding window created before the second sliding window. That is, if there is a third sliding window created earlier than the first sliding window, the first sliding window may correspond to the second sliding window based on the third sliding window. That is, here, the distinction between the first and second sliding windows can be relatively defined according to the temporal order of creation time.

The sliding window accumulator 370 may generate a second sliding window adjacent to the first sliding window based on the first sliding window and accumulated in the y-axis direction on the binary image. Here, the accumulation direction of the second sliding window may correspond to the driving direction of the vehicle. The sliding window accumulator 370 sets the first sliding window and then repeatedly performs the process of stacking a second sliding window of the same size as the first sliding window in the y-axis direction to detect all lanes in one binary image. can do.

In one embodiment, the sliding window accumulator 370 may generate a second sliding window of a preset size centered on the average value of x-axis coordinates of lane data existing in the first sliding window. Here, the lane data may correspond to pixel data of an area corresponding to the lane within the image and may include x-axis coordinates and y-axis coordinates defined on image coordinates. Basically, the sliding window accumulator 370 can utilize information about the first sliding window created in the previous step to create the second sliding window. That is, when the x-axis center of the lane existing within the first sliding window is determined, the coordinates of a point moved by a predetermined distance from the x-axis center in the y-axis direction may be determined as the center coordinates for creating the second sliding window. As a result, the second sliding window may correspond to a sliding window created as a result of defining a preset size based on the corresponding center coordinate.

In one embodiment, the sliding window accumulator 370 may determine the center of the second sliding window through first sliding window correction when no lane data exists in the first sliding window. Here, the first sliding window correction may correspond to one of the correction methods according to the present invention. In other words, position correction of the sliding window can be done by utilizing lane information existing in the previously created sliding window, and if lane information does not exist in the sliding window of the previous step, it is extended to the previously created sliding windows. This can be done based on lane information of the corresponding sliding windows.

In one embodiment, the sliding window accumulator 370 may perform first sliding window correction when lane data existing within the first sliding window is less than a preset threshold. Additionally, the first sliding window correction can be performed not only when there is no lane information within the sliding window, but also when there is only a small amount of lane data that is not sufficient to determine the correct lane position. For example, the sliding window accumulator 370 may perform first sliding window correction when the number of lane data, that is, lane pixels, within the first sliding window is less than a preset threshold. At this time, the threshold may be determined in advance through repeated experiments.

In one embodiment, the sliding window accumulator 370 calculates the average value of the x-axis coordinates of the lane data for each of the first sliding windows created in the previous step, calculates the slope according to the change in the average value, and calculates the most recent 1 The center of the second sliding window can be determined by applying a tilt to the center of the sliding window. That is, the first sliding window correction process may correspond to a process of correcting the position of the next sliding window to the correct position when an empty sliding window is detected by using the previous sliding window with lane data. If the sliding window accumulator 370 determines that the sliding window is empty, it can calculate the average x-coordinate of the previously valid lane data pixels for each sliding window and calculate the slope through the average value. In other words, the slope can represent the change in the average value of x-coordinates.

Additionally, when the center position of the second sliding window is determined according to the application of the tilt, the sliding window accumulator 370 may generate a second sliding window that has a preset size based on the center position. At this time, when empty sliding windows are continuously detected, the sliding window accumulator 370 can repeatedly create and accumulate sliding windows based on the previously obtained slope, and this process continues until lane data appears in the sliding window. It can continue.

In one embodiment, the sliding window accumulation unit 370 may determine the center of the second sliding window through second sliding window correction when the number of first sliding windows created in the previous step is less than 2. Here, the second sliding window correction may correspond to another one of the correction methods according to the present invention. In other words, position correction of the sliding window can be done by utilizing the previously created sliding window within the same frame or lane information existing within the sliding window created in the previous frame. If the number of sliding windows created in the previous step is not sufficient, Therefore, if it is difficult to calculate the slope for the first sliding window correction, the second sliding window correction may be performed by extending to the previous frame and using the associated lane information.

In one embodiment, the sliding window accumulator 370 is configured to operate when the number of first sliding windows generated in the previous step is 2 and no lane data exists for one of the first sliding windows or lane data is preset. If it is less than the threshold, a second sliding window correction may be performed. That is, even if there are two sliding windows created in the previous step, if normal lane data does not exist in one or more sliding windows, the second sliding window correction can be performed.

For example, if a portion without a lane is located from the first sliding window, it may be difficult to apply the first sliding window correction. In addition, the standard for the first sliding window can be calculated through the entire y-axis histogram, and if the lower lane is very empty in a lane slanted to the left or right, only the biased upper lane data comes in, and then the standard is applicable. An error may occur where the window is moved in the wrong direction and is set to the wrong position from the first sliding window.

Additionally, lane recognition can be made based on images received through a camera sensor. Assuming that images come in at a 30fps cycle in the case of a camera sensor, the algorithm that detects lanes in the image can operate at 10fps. In other words, the time interval between successively processed frames may be 0.1 seconds, indicating that the image difference between each frame is not large. Therefore, there may not be a significant difference between the lane position in the previous frame and the lane position in the next frame. Accordingly, when the sliding window accumulation unit 370 determines that the lane data in the portion where the sliding window is first accumulated is insufficient, it can perform a second sliding window correction process using the sliding window data of the previous frame.

In one embodiment, the sliding window accumulator 370 resets the center and slope of the first sliding window based on the data values stored in the previous frame of the current frame, and places the reset slope at the center of the reset first sliding window. By applying this, the center of the second sliding window can be determined. That is, the sliding window accumulator 370 can obtain the reference point and slope of the first sliding window from the previous frame and use it for position correction, and for this purpose, the corresponding data can be transmitted together every time a frame is passed.

Specifically, the sliding window accumulator 370 can apply the second sliding window correction if any of the first and second sliding windows has a sliding window without suboptimal data, and sets the position of the sliding window as the default data of the previous frame. and tilt can be reset. Additionally, the sliding window accumulation unit 370 may repeat the process of stacking sliding windows until a sliding window containing lane data appears based on the reset slope. Afterwards, the sliding window accumulator 370 may initialize the repetition process when a lane within the sliding window is detected and set the position of the sliding window again based on the corresponding lane data.

The control unit 390 controls the overall operation of the double correction device 130, and controls the operation between the frame image extractor 310, the base point determination unit 330, the sliding window creation unit 350, and the sliding window accumulation unit 370. It can manage control flow or data flow.

Figure 4 is a flowchart explaining the sliding window double correction method according to the present invention.

Referring to FIG. 4, the dual correction device 130 may generate a binary image based on the original image for the current frame of the driving image including at least a lane through the frame image extractor 310 (step S410). . The double correction device 130 may determine the base point based on the x-axis on the binary image through the base point determination unit 330 (step S430).

Additionally, the double correction device 130 may generate a first sliding window of a preset size centered on the base point through the sliding window generator 350 (step S450). The double correction device 130 may generate a second sliding window that is adjacent to the first sliding window and accumulated in the y-axis direction on the binary image based on the first sliding window through the sliding window accumulator 370. (Step S470).

FIG. 5 is a diagram illustrating an example of a position setting error related to a sliding window.

Referring to FIG. 5, because platooning vehicles drive while maintaining a narrow distance, a significant portion of the lane image received through the sensor may be without a lane. Each sliding window 530 may be performed by stacking the next sliding window 530 based on the x-axis average value of the lane data 510 within the previous sliding window 530, where the previous sliding window 530 is the lane data 510. If it is located in a part where there is no line, the lane data 510 required for calculating the average may not be present within the corresponding sliding window 530. In this case, the general method is to stack the next sliding window 530 in the same position as the existing sliding window 530, and as shown in Figure 5, the sliding window 530 as long as the empty section is straight in the y-axis direction. can be accumulated.

The method of stacking the sliding windows 530 in a straight line may cause a lane detection error as shown in FIG. 5 with a dotted line in a curved section of the road. In other words, a problem may occur where the sliding window 530 sets an incorrect position in a section where a lane appears again after a section where there is no lane, and as a result, the accuracy of lane recognition may decrease.

Figure 6 is a diagram explaining the sliding window double correction process according to the present invention.

Referring to FIG. 6, the dual correction device 130 can resolve positioning errors caused by an empty sliding window in which no lane data exists through first and second sliding window correction processes.

Error in positioning of the sliding window may occur due to the phenomenon of the next sliding window being piled up by the empty sliding window. In the first correction (first sliding window correction), the empty sliding window is created by utilizing the previous sliding window with suboptimal data. This may correspond to the process of correcting the position of the next stacked sliding window when it comes out.

The first sliding window correction process can be performed as follows. First, as in step S610 of FIG. 6, it can be determined whether it is a sliding window that requires correction. If it is determined to be a normal sliding window with suboptimal data, a general sliding window process such as step S620 of FIG. 6 may be performed. In addition to a sliding window without any lane data, even when only a small amount of lane data is input to determine the correct lane position, it is classified as an empty sliding window and a correction process based on previous valid lane data can be performed. That is, the x-coordinate average of the previously valid lane data pixels can be calculated for each sliding window, and then the slope can be calculated using the average value. Through the slope calculated as in step S640 of FIG. 6, the reference point of the next sliding window at an appropriate location rather than the same x coordinate can be determined, and the sliding window can be set around the reference point.

Whether or not to apply the second sliding window correction process may be determined based on the minimum two sliding windows required when calculating the slope, as in step S630 of FIG. 6. That is, the second sliding window correction process can be applied if any of the first and second sliding windows exist a sliding window without suboptimal data, and the position of the sliding window is set to the data default value of the previous frame as in step S650 of FIG. 6. and the slope can be reset. Thereafter, the process of stacking sliding windows may be repeatedly performed until a sliding window containing lane data appears based on the reset slope.

Figure 7 is a diagram explaining an experimental example related to the present invention.

Referring to FIG. 7, this may correspond to an experiment result in a scale truck environment. Here, the scale truck environment may consist of a scale truck and a test track. The scale truck may correspond to a 1/14 scale experimental truck for experiments. The truck can be configured with a 148-degree wide-angle camera, sensors such as RPLidar A3, Nvidia Jetson Xavier, and OpenCR boards. In addition, the test track has a total size of 40 m wide and 12 m long, and the lane width can be 33 cm. The experiment can be run in a highway environment with a curve radius of 5 m and on a track with radii of 3 and 4 m, and thus the performance differences depending on curvature can be analyzed.

In addition, curved driving images can be used so that results can be easily compared among all driving images, and experiments can be performed based on a total of 800 images, 200 of each. Changes in accuracy can be derived by comparing GT (Ground Truth) images created based on Labelme and the resulting images before and after applying the algorithm. Changes in accuracy can be derived by converting the GT image and the resulting image into a bird's eye view image, then overlapping the images and comparing the differences in distribution. Although the value derived from the distribution difference may be somewhat different from the absolute accuracy value, the degree of improvement in the accuracy of the exact value can be confirmed through the change in the distribution difference between images. The experiment can be divided into three stages: before applying the algorithm, applying the first correction, and applying the first and second corrections.

In experiments in a scale truck environment, results such as those shown in FIG. 7 can be derived. First, in the first experiment results on a track with a radius of 5 m, the difference in distribution between the experimental results before applying the algorithm and GT may be 90.24% on average for 200 sheets, and the difference in distribution when only the first correction algorithm is applied can be 94.94%. . And when all of the second correction algorithms are applied, the result can be 97.75%. In the second experiment conducted at a radius of 4 m, the distribution difference was 87.28% before applying the algorithm, 92.96% after applying the first correction algorithm, and 94.28% when applying the second correction algorithm. Lastly, in the third experiment conducted at a radius of 3 m, the results were 73.44% before applying the algorithm, 78.96% after applying the first correction algorithm, and 84.08% when applying the second correction algorithm.

As a result, in experiments in a scale truck environment, it can be seen that the accuracy is improved by about 8% on average by applying the algorithm. And, by comparing the accuracy according to the change in curvature, it can be seen that the average accuracy decreases in the 3rd experiment with a large curvature compared to the 1st experiment with a small curvature. As the curvature increases, the average accuracy decreases, but the accuracy is improved through the application of the algorithm. You can see that the width is larger.

Although the present invention has been described above with reference to preferred embodiments, those skilled in the art may make various modifications and changes to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that you can do it.

100: Sliding window double compensation system
110: User terminal 130: Double correction device
150: database
210: Processor 230: Memory
250: user input/output unit 270: network input/output unit
310: frame image extraction unit 330: base point determination unit
350: sliding window creation unit 370: sliding window accumulation unit
390: Control unit

Claims (14)

generating a binary image based on an original image for a current frame of a driving image including at least a lane;
determining a base point based on the x-axis on the binary image;
generating a first sliding window of a preset size centered on the base point; and
Generating a second sliding window adjacent to the first sliding window based on the first sliding window and accumulating in the y-axis direction on the binary image;
The step of creating the second sliding window is
When no lane data exists in the first sliding window, determining the center of the second sliding window through first sliding window correction using lane data in the first sliding windows generated in the previous step,
If the number of first sliding windows generated in the previous step is less than 2, determining the center of the second sliding window through second sliding window correction using data values stored in the previous frame of the current frame. Features a sliding window double correction method.
The method of claim 1, wherein generating the binary image comprises:
removing noise from the original image;
setting a region of interest regarding the lane on the original image;
converting the image of the region of interest into a bird's eye view image; and
A sliding window double correction method comprising converting the bird's eye view image to the binary image.
The method of claim 1, wherein determining the base point comprises
A sliding window double correction method comprising determining the base point as a reference for each lane through a histogram representing the number of lane pixels at each coordinate on the binary image.
The method of claim 1, wherein generating the second sliding window comprises:
A sliding window double correction method comprising the step of generating a second sliding window of the preset size centered on an average value of x-axis coordinates of lane data existing within the first sliding window.
delete The method of claim 4, wherein generating the second sliding window comprises:
A sliding window double correction method comprising performing first sliding window correction when the lane data existing within the first sliding window is less than a preset threshold.
The method of claim 1, wherein generating the second sliding window comprises:
calculating an average value of x-axis coordinates of the lane data for each first sliding window created in the previous step;
calculating a slope according to a change in the average value; and
A sliding window double correction method comprising determining the center of the second sliding window by applying the tilt to the center of the most recent first sliding window.
delete The method of claim 7, wherein generating the second sliding window comprises:
If the number of first sliding windows generated in the previous step is 2 and the lane data does not exist for one of the first sliding windows or the lane data is less than a preset threshold, second sliding window correction is performed. A sliding window double correction method comprising the steps of:
The method of claim 7, wherein generating the second sliding window comprises:
resetting the center and the slope of the first sliding window based on data values stored in a frame previous to the current frame; and
A sliding window double correction method comprising the step of applying the reset slope to the center of the reset first sliding window to determine the center of the second sliding window.
a frame image extraction unit that generates a binary image based on the original image of the current frame of the driving image including at least the lane;
a base point determination unit that determines a base point based on the x-axis in the binary image;
a sliding window generator that generates a first sliding window of a preset size centered on the base point; and
A sliding window accumulator for generating a second sliding window adjacent to the first sliding window based on the first sliding window and accumulating in the y-axis direction on the binary image,
The sliding window accumulator is
If no lane data exists in the first sliding window, determine the center of the second sliding window through first sliding window correction using lane data in the first sliding windows generated in the previous step,
When the number of first sliding windows created in the previous step is less than 2, the center of the second sliding window is determined through second sliding window correction using data values stored in the previous frame of the current frame. Windows double compensator.
The method of claim 11, wherein the sliding window accumulation unit
A sliding window double correction device, characterized in that it generates a second sliding window of the preset size centered on the average value of x-axis coordinates of lane data existing in the first sliding window.
delete delete