JP4821348B2 - Road section line detection apparatus and method, and program - Google Patents

Description

  The present invention relates to a road section line detection apparatus, method, and program for detecting a road section line.

2. Description of the Related Art Conventionally, an imaging unit that is mounted on a vehicle and images the front of the vehicle, a first lane marker recognition unit that performs a lane marker recognition process based on a change in luminance of an image captured by the imaging unit, and an image that is captured by the imaging unit A second lane marker recognizing unit that performs a lane marker recognizing process by pattern matching between the generated image and a predetermined template, a signal transmitting unit that transmits an optical signal to the front of the vehicle, and an optical signal transmitted from the signal transmitting unit. One of the first lane marker recognizing means and the second lane marker recognizing means is selected based on the signal receiving means for receiving the reflected signal and the received amount of the reflected signal received by the signal receiving means. A lane marker recognition device including a selection unit is known (for example, see Patent Document 1).
JP 2004-139338 A

  By the way, in general, the road section line is the difference between the signal level of the pixel related to the road marking line included in the image and the signal level of the pixel in the other region (for example, luminance) as in the above-described conventional technology. Is detected based on the difference.

  However, there is a possibility that the influence of noise cannot be appropriately removed simply by focusing on only the difference in signal level.

  Further, when a landscape including a road surface is imaged from a vehicle and an ambient image is acquired, the signal level of the obtained ambient image is affected by the ambient environment of the vehicle such as ambient light. For this reason, depending on the surrounding environment of the vehicle, the difference between the signal level of the pixel related to the road lane marking and the signal level of the pixel in other areas is small, and the signal level difference necessary for detection of the road section line can be obtained. There may not be.

  Therefore, an object of the present invention is to provide a road section line detection apparatus, method, and program capable of detecting a road section line with high accuracy in a manner that is less susceptible to the influence of the surrounding environment and noise.

In order to achieve the above object, a road section line detection device according to a first aspect of the present invention is an imaging unit that captures a landscape including a road surface from a vehicle and acquires a surrounding image;
The degree of signal level variation representing luminance, hue, or saturation in the region of interest of the surrounding image is calculated, and the feature points of the road section line are calculated in the region of interest based on whether the calculated degree of variation is greater than a threshold. Determining means for determining whether or not exists,
Threshold value setting means for calculating a variation level of a signal level representing luminance, hue or saturation in a region larger than the region of interest, and setting the threshold value based on the calculated variation level;
The determination means all feature points in the target area where the feature point is determined not to exist in the road section line while discarded by the features of interest within the region feature point of the road section wire is determined to exist by said judgment means Feature point detecting means for detecting feature points of road section lines based only on the points,
A road segment line is detected based on a feature point of the road segment line detected by the feature point detection means.

The second invention is the road section line detection device according to the first invention,
A plurality of the regions of interest are set, and the determination unit performs the determination for each region of interest.

A third aspect of the present invention is the road section line detection device according to the second aspect of the present invention,
The threshold setting means sets the threshold based on an average value of the degree of variation calculated for each region of interest.

4th invention is the road segment line detection apparatus which concerns on either 1st-3rd invention,
The threshold value setting means sets the threshold value based on the number of feature points in a region of interest determined that a feature point on a road section line exists.

5th invention is the road section line detection apparatus which concerns on either invention of 1-4,
The surrounding image is processed for each frame,
The threshold value set by the threshold value setting unit is used for determination by the determination unit for the surrounding image of the next frame.

6th invention is the road section line detection apparatus which concerns on either invention of 1-5,
The degree of variation in the signal level is calculated based on a surrounding image after morphological operation processing.

In a seventh aspect of the present invention, in the road section line detection device according to any one of the first to sixth aspects of the invention,
The road section line is a kind of road demarcation line that is provided in the form of dots at intervals.

In order to achieve the above object, a road section line detection method according to an eighth aspect of the invention acquires a surrounding image including a road surface imaged by an imaging means from a vehicle;
The degree of signal level variation representing luminance, hue, or saturation in the region of interest of the surrounding image is calculated, and the feature points of the road section line are calculated in the region of interest based on whether the calculated degree of variation is greater than a threshold. A determination step of determining whether or not exists,
A threshold setting step of calculating a variation level of a signal level representing luminance, hue or saturation in a region larger than the region of interest, and setting the threshold based on the calculated variation level;
Wherein while all feature points in the target area where the feature point is determined not to exist in the road section line in the determination step discarded, characterized in interest in the region feature point of the road section wire is determined to exist in said determining step A feature point detecting step for detecting a feature point of a road section line based only on the point;
And a step of detecting a road section line based on the feature points of the road section line in the region of interest detected in the feature point detecting step.

In order to achieve the above object, a computer readable program according to a ninth aspect of the invention is a computer, and (1) a process of acquiring a surrounding image including a road surface imaged by an imaging means from a vehicle,
(2) The degree of variation of the signal level representing the luminance, hue, or saturation in the region of interest of the surrounding image is calculated, and the road segment line is included in the region of interest based on whether the calculated degree of variation is greater than a threshold value. A process for determining whether or not a feature point exists,
(3) Discarding all the feature points in the target area determined that the feature points of the road section line do not exist, and only based on the feature points in the target area determined to have the feature points of the road section line , Processing for detecting feature points of road section lines , and
(4) A process for detecting a road section line is executed based on the detected feature point of the road section line in the region of interest .

  ADVANTAGE OF THE INVENTION According to this invention, the road segment line detection apparatus, method, and program which can detect a road segment line with high precision in the aspect which is hard to receive to the influence of surrounding environment and noise are obtained.

  The best mode for carrying out the present invention will be described below with reference to the drawings.

  FIG. 1 is a system configuration diagram including an embodiment of a road marking line detection apparatus according to the present invention. A camera 10 that captures a landscape including a road surface from a vehicle is connected to the road marking line detection device 100 of the present embodiment. The camera 10 obtains a surrounding image including a road surface having 5 million pixels, for example, by an imaging element such as a charge-coupled device (CCD) or a complementary metal oxide semiconductor (CMOS). The camera 10 may be attached to a rearview mirror, for example, and may capture a road surface in front of the vehicle, or may be attached to a side mirror, for example, to capture a road surface on the side of the vehicle, It may be attached to the back door and image the road surface behind the vehicle.

  The camera 10 preferably acquires a surrounding image in real time, and the acquired surrounding image is supplied to the road marking line detection device 100. In this case, the surrounding image may be supplied to the road marking line detection apparatus 100 in a stream format with an appropriate frame rate.

  As will be described in detail below, the road marking line detection device 100 detects road marking lines around the vehicle based on the surrounding image supplied from the camera 10. For example, in the case of the camera 10 that images the road surface ahead of the vehicle traveling direction, the road lane marking detection device 100 detects a road lane marking ahead of the vehicle. Road lane markings include white lines on the road, Botts Dots, cat's eyes, and the like. The white line may include any line that divides the road, such as a line of an arbitrary color such as a yellow line, a solid line, a broken line, a dotted line, and a double line.

  FIG. 2 is an explanatory diagram of botsdots and cat's eyes. The botsdots 51 is a lane marking mainly used in North America, and is formed by arranging a plurality of solid disks having a diameter of about 100 mm made of ceramic, for example, and embedded in the road surface (FIG. 2A). On the other hand, the cat's eye 53 is a dividing line in which a plurality of reflectors formed in a substantially rectangular shape are arranged, and has a characteristic of reflecting incident light in the same direction (FIG. 2B). For example, in Japan, a lane marking composed only of the cat's eye 53 is used on a curved road other than a highway, and in North America, it is used not only on a curved road but also on a straight road. Both the bots dot 51 and the cat's eye 53 are arranged in a state of slightly protruding from the road surface.

  The detection result of the road lane marking by the road lane marking detection apparatus 100 can be used for various in-vehicle systems that support vehicle travel along the lane.

  For example, in the example illustrated in FIG. 1, the road lane marking detection apparatus 100 is connected to a travel support ECU 30 that supports travel of the vehicle via a communication line. The road marking line detection device 100 transmits the road marking line detection result to the travel support ECU 30 via the communication line.

  The driving support ECU 30 is connected to a communication line 33 such as a vehicle LAN (Local Area Network). The vehicle LAN 33 is connected to an electric power steering ECU (EMPS-ECU) 35 that controls the assist force applied to the steering. Furthermore, an engine ECU (EFI-ECU) 37 that controls the engine in an integrated manner and a meter ECU 39 that controls display of a meter panel and the like are connected to the vehicle LAN 33. These travel support ECU 30, EMPS-ECU 35, EFI-ECU 37, and meter ECU 39 perform bidirectional data communication based on the CAN protocol or the like via the vehicle LAN 33.

  The EMPS-ECU 35 is connected to a steering actuator 35a such as an electric motor that applies assist force to the steering. The EMPS-ECU 35 drives and controls the steering actuator 35a based on the torque value (current value) transmitted from the travel support ECU 30.

  A vehicle speed sensor 37a for detecting the vehicle speed and a steering angle sensor 37b for detecting the steering angle of the steering are connected to the EFI-ECU 37. The vehicle speed detected by the vehicle speed sensor 37 a and the steering angle detected by the steering angle sensor 37 b are transmitted to the travel support ECU 30 via the EFI-ECU 37 and the vehicle LAN 33.

  The driving assistance ECU 30 predicts the position of the subsequent vehicle based on the road lane line detection result transmitted from the road lane line detection device 100 via the communication line and the driving situation of the driver, and the own vehicle is When it is determined that there is a possibility of deviating from the above, an alarm signal is transmitted to the meter ECU 39 (lane departure control). When the meter ECU 39 receives an alarm signal from the driving support ECU 30, the meter ECU 39 generates a warning sound in the buzzer 39a and / or turns on the warning lamp 39b.

  Further, the driving support ECU 30 calculates the estimated position of the host vehicle based on the detection result of the road marking line transmitted from the road marking line detection device 100 via the communication line. Further, the driving support ECU 30 determines that the own vehicle is an abbreviation of the lane marking based on the calculated estimated position of the own vehicle, the vehicle speed detected by the vehicle speed sensor 37a, and the steering angle detected by the steering angle sensor 37b. The assist steering force (current value) required for traveling in the center is calculated and transmitted to the EMPS-ECU 35 (lane keeping control). The EMPS-ECU 35 controls the steering actuator 35a based on the assist steering force transmitted from the travel support ECU 30. The detection result of the road marking line can be used for lane departure control, lane maintenance control, and the like. According to the present embodiment, road lane markings can be detected with high accuracy as will be described later, so that each of these controls can be performed with high accuracy.

  FIG. 3 is a functional block diagram showing the main part of the road marking line detection apparatus 100 according to the present embodiment. A hardware configuration of the road marking line detection apparatus 100 is configured around a microcomputer. Accordingly, the road lane marking detection apparatus 100 includes a CPU that performs various processes according to a given execution program, a ROM that stores the execution program of the CPU, a readable / writable RAM (Random Access Memory) that stores image data, calculation results, and the like. ), Timer, counter, input / output interface, etc. The CPU, ROM, RAM, and input / output interface are connected to each other by a data bus. In addition, each part of the road marking line detection apparatus 100 demonstrated below is implement | achieved by the program run by CPU.

The road marking line detection apparatus 100 includes a feature point extraction unit 102, a noise removal unit 104, a road marking line detection unit 106, and a threshold setting unit 108, as shown in FIG. 3 is executed for each frame of the surrounding image. That is, in principle, it is executed independently for each frame of the surrounding image that is supplied or read as needed (however, as will be described later, the method of setting the standard deviation threshold value _SDTH is correlated between the frames). Can be.) Hereinafter, the processing of each unit will be described in detail.

  The feature point extraction unit 102 extracts feature points in the surrounding image. Although there are various feature point extraction methods, the feature point extraction unit 102 preferably extracts feature points by morphological operations. According to the morphological operation, only elements having a specific geometric structure can be selectively extracted from the original image (grayscale image) by a set-theoretic operation using a preset structural element B. In addition, it is possible to relatively easily detect the dotted road segment lines such as the above-described botsdots and cat's eyes that are difficult to distinguish from noise.

  Here, as a preferred embodiment, a case will be described in which feature point extraction processing is executed by morphological operations. The morphological operation is described in detail in “Morphology” (Corona).

  The feature point extraction unit 102 receives the surrounding image acquired as described above. At this time, the surrounding image may be grayscaled after being subjected to shading correction in order to remove luminance unevenness caused by the optical system. For example, it may be a grayscale image (grayscale image) with 256 gradations. Hereinafter, the grayscale image of the surrounding image thus input is referred to as an “input image”.

  Based on a predetermined structural element B, the feature point extraction unit 102 generates a contracted image by performing contraction processing (Erosion) on the input image, and performs expansion processing (Dilation) on the generated contracted image To generate an Opening image. Next, the feature point extraction unit 102 generates a difference image obtained by subtracting the opening image from the input image by the top hat conversion process (see FIG. 5).

  The structural element B used for the morphological calculation is appropriately set according to the road segment line to be detected. That is, the size of the structural element B is determined so that only the road segment line to be detected and noise smaller than that remain in the obtained difference image.

  Here, as an example of detecting the dotted road segment lines such as the above-mentioned botsdots and cat's eyes, which are difficult to distinguish from noise, a case where the detection target is a botsdots will be described. Since both the botsdots and cat's eyes are very small, they appear to have substantially the same characteristics on the input image. Accordingly, dotted road segment lines such as botsdots and cat's eyes can be detected in a similar manner.

  The size of the botsdots apparently changes from “a state like a point” to a case of “having a width of several pixels in the horizontal direction”. The nature of the top-hat transformation is basically to “leave something smaller than the structural element B”. Therefore, if the size of the structural element B is 10 cm + α, which is the size of the botsdots, it will apparently be 10 cm or less. If you can see it, you can deal with it. The size of α is set by actual experiment, but it is preferable to set it to an appropriate size so as not to leave a feature point related to an unnecessary object such as noise.

  FIG. 4 is a conceptual diagram for explaining the morphological operation processing, and FIG. 4A shows a luminance distribution on one arbitrary line (scanning line) in the horizontal direction of the input image. That is, FIG. 4A shows the luminance distribution signal f (x) of the pixel row of the input image (x indicates the pixel position in the horizontal direction). In this example, as shown in FIG. 4A, it is assumed that a white line having a width of 6 pixels, a noise of 2 pixels (pixel), and a botsdot of 3 pixels are included on the scanning line. The structural element B has a height of 1 pixel and a horizontal width of 5 pixels, for example.

  In FIG. 4B, the luminance distribution on the scanning line of the contracted image is indicated by a broken line, and the luminance distribution signal f (x) of the input image is indicated by a solid line for reference. That is, FIG. 4B shows the luminance distribution signal after the contraction process by a broken line. The contraction process is a process of searching for a minimum value within a range of a mask size width (a value determined according to the structural element B, which is 2.5 pixels in this example) with the target pixel as the center. Therefore, noise and botsdots smaller than the mask size (5 pixels) are eliminated by the shrinking process, and the width representing the white line is shrunk.

  In FIG. 4C, the luminance distribution on the scanning line of the opening image is indicated by a broken line, and the luminance distribution signal f (x) of the input image is indicated by a solid line for reference. The opening process is a process of performing an expansion process after the contraction process, that is, a process of searching for the maximum value after searching for the minimum value. Therefore, although the width representing the white line is restored by the expansion process, the noise and the botsdots remain erased and are not restored.

  FIG. 4D shows a luminance distribution on the scanning line of the difference image by a solid line. That is, FIG. 4C shows the luminance distribution signal after the top hat conversion process by a solid line. When the luminance distribution signal after the opening process is subtracted from the luminance distribution signal f (x) of the input image, the luminance distribution signal related to the white line (dotted line) larger than the structural element B is deleted as shown in FIG. It can be seen that a luminance distribution signal related to only noise and botsdots smaller than the structural element B can be obtained. The feature point extraction unit 102 extracts pixel points having a predetermined luminance level or higher in the difference image as feature points. As a result, feature points related to an object larger than the structural element B such as a white line are removed, and only feature points related to noise and botsdots are extracted.

  In the above description, the one-dimensional structural element B having a height of 1 pixel and a horizontal width of 5 pixels is used. For example, a circular structural element (for example, a diameter of 4 pixels) larger than Botsdots (two-dimensional structure) Element) B may be used. When one-dimensional structural elements are used, even if there are repair marks, wrinkle streaks, rain streaks, etc. equal to the width of the botsdots, they may not be removed even though their shapes are completely different. is there. Furthermore, since the horizontal width is the same, these unnecessary items remain even if the top hat conversion process is performed. However, even in such a situation, the use of the two-dimensional structural element B makes it possible to easily distinguish the above unnecessary items.

  Further, the direction of the scanning line is not necessarily a horizontal direction, and may be a vertical direction or an oblique direction. In addition, the morphological calculation process may be executed for all scan lines in the recognition target area with the entire area of the input image as the recognition target area, but the predictable road surface area is limited as the recognition target area. May be. In this case, the difference image has fewer total pixels than the input image. Moreover, you may limit the area | region near a vehicle among road surface areas as a recognition object area | region. This is an area where a 10 cm bots dot has a pixel size (for example, 2 pixels or less) that is difficult to distinguish from noise, considering that the bots dot is only 10 cm in diameter and is almost invisible in a distant area. This is to eliminate unnecessary processing.

  The noise removing unit 104 removes noise included in the feature points extracted by the feature point extracting unit 102. That is, the noise removing unit 104 removes components (mainly noise) components other than the bots dot included in the difference image.

Specifically, the noise removing unit 104 divides the difference image (recognition target region) into several regions of interest, calculates the degree of variation in brightness level (degree of dispersion) for each region of interest, and calculates the degree of variation there is judged based on whether a larger standard deviation threshold SD _TH, the remarked region, whether the feature point of the road section line exist.

  FIG. 5 shows an example of how to set a region of interest for a difference image. In the example shown in FIG. 5, the recognition target area is set in a substantially parallelogram shape corresponding to a road extending diagonally to the right, and the recognition target area includes both botsdots and noise as described above. ing. The region of interest may be set as a rectangular area as shown in FIG. 5, or may be set as an area of another shape. Further, the region of interest may be a region on one scanning line. However, it is desirable that the region of interest has a size larger than the number of pixels representing the botsdots so that the degree of variation is appropriately evaluated. The number of pixels used for calculating the degree of variation (the number of samples) depends on the total number of pixels of the difference image, the number of pixels representing botsdots, and the like, but may be about 100 pixels, for example.

  In this example, the variation degree is calculated as a standard deviation of the luminance level. However, statistics such as variance and average deviation may be used instead of the standard deviation. The degree of variation is preferably derived based on the luminance levels of all the pixels included in the region of interest, but may be derived after appropriately thinning out some of the pixels included in the region of interest.

For example, when a certain region of interest has N pixels i (i = 1,... N) and the luminance level Li of each pixel i (i = 1,... N), The variance Φ for the region of interest is obtained by using the average value Lav of the luminance levels Li of the respective pixels i (i = 1,... N) of the region of interest.
Φ = 1 / N {(L1−Lav) ² + (L2−Lav) ² +. . . + (LN−Lav) ² }. Therefore, the standard deviation SD for one region of interest is expressed as SD = Φ1 ^{/ 2} . Hereinafter, assuming that there are M regions of interest, the standard deviations in each region of interest R _i (i = 1,... M) are represented by symbols SD _i (i = 1,... M), respectively.

Here, as in the region of interest R ₂ shown in FIG. 5, if it contains pixels representing the Botts' dots to the target region, as the region of interest R ₁ shown in FIG. 5, compared with a case that contains only the noise region of interest Therefore, the standard deviation is significantly increased. That is, SD ₂ > SD ₁ . In the present embodiment, paying attention to this point, an appropriate standard deviation threshold value SD _TH is set to discriminate a region of interest that includes botts and dots and a region of interest that does not include bots. The standard deviation threshold value SD _TH is appropriately determined in consideration of the resolution of the difference image, the size (number of pixels) of the bots dot, and the like. The standard deviation threshold SD _TH is set by a threshold setting unit 108 described later. The standard deviation threshold SD _TH may be fixed, but is preferably varied by a threshold setting unit 108 described later as described later.

For the attention area R _i in which the standard deviation SD _i is equal to or less than the standard deviation threshold SD _TH (for example, the attention area R ₁ shown in FIG. 5), the noise removing unit 104 includes only a noise component in the attention area R _i. If it is determined that it is included, all feature points in the region of interest R _i are removed and discarded. The noise removing unit 104, for the standard deviation SD _i standard deviation threshold _{SD TH} larger region of interest than _{R i} (e.g. target region _{R 2} shown in FIG. 5), the said region of interest _{R i} to Botts' dots It is determined that such a feature point is included, and the feature point in the region of interest R _i is maintained.

FIG. 6 shows the difference image after the above-described noise removal processing by the noise removal unit 104. As shown in FIG. 6, after the noise removal processing, all feature points in the target area small standard deviation of such luminance levels as region of interest R ₁ shown in FIG. 5 is removed. Accordingly, after the noise removal process, only the feature points in the target region standard deviation is large brightness level, such as region of interest R ₂ shown in FIG. 5 is left. That is, only the feature points in the region of interest determined to include botsdots are left.

The road marking line detection unit 106 detects the feature points related to the botsdots based on the feature points of the difference image after the noise removal processing by the noise removal unit 104 as described above. That is, the road marking line detection unit 106 detects a feature point related to the bots dot in the region of interest R _i in which the standard deviation SD _i is larger than the standard deviation threshold SD _TH . Here, as described above, even in the region of interest R _i in which the standard deviation SD _i is larger than the standard deviation threshold SD _TH , noise components can still remain in addition to the botsdots. For example, the noise component can be easily removed by a predetermined filter threshold. On the other hand, when the brightness of noise is higher than the brightness of botsdots, only feature points related to botsdots can be easily detected by Hough transform or the like that detects only a linearly arranged pattern.

  The road lane marking detection unit 106 generates and outputs a feature point list based on the feature points detected as the feature points related to botsdots. Although the feature point list is not described in detail here, the position and direction (lane direction) of the botsdots can be grasped based on the feature point list. Become.

The threshold setting unit 108 sets the standard deviation threshold SD _TH used in the threshold determination processing based on the standard deviation SD _i described above by the noise removing unit 104. In this example, the threshold setting unit 108 sets the standard deviation threshold SD _TH for the next frame, paying attention to the fact that the appearance mode of the noise component on the input image does not change greatly between one frame. An appropriate initial value (for example, 6) may be set as the standard deviation threshold value SD _TH for the input image of the first frame.

Specifically, the threshold setting unit 108 determines and sets the standard deviation threshold SD _TH by the following formula using the average value SD _AV of the standard deviation of the luminance level in the entire recognition target region.
SD _TH = SD _AV + β
Here, β is a correction term and may be a fixed value, or may be a variable value as will be described later. The average value SD _AV of the standard deviation of the entire recognition target area is expressed by SD _AV = 1 / M (SD ₁ + SD ₂ + ... + SD _M ). Note that the average value SD _AV of the standard deviation of the entire recognition target area may be a standard deviation based on the luminance levels of all pixels included in the entire recognition target area.

By the way, in a dark scene with little ambient light, the luminance level of the pixels related to the botsdots decreases, and accordingly, the standard deviation of the region of interest including the botsdots pixels decreases. However, in such a case, the average value SD _AV of the standard deviation of the entire recognition target area becomes small, and accordingly, the standard deviation threshold value SD _TH that is set also becomes a small value.

As described above, according to the present embodiment, the standard deviation threshold value SD _TH is set using the degree of variation in the luminance level of the entire input image (difference image), so that the luminance level of the entire input image) depends on the surrounding environment of the vehicle. be wholly if changed, it is possible to accurately detect a feature point according to Botts' dots by suitable standard deviation threshold SD _TH.

In the present embodiment, the correction term β used to set the standard deviation threshold value SD _TH may be a variable value that dynamically changes according to the state of the input image. For example, the correction term β may be changed according to the number of feature points included in the difference image after the noise removal processing by the noise removal unit 104. In this case, when the number of feature points is equal to or greater than a predetermined number (when there are many feature points), the threshold setting unit 108 determines that the standard deviation threshold SD _TH is smaller than the optimum value, and increases the correction term β. On the other hand, when the number of feature points is smaller than the predetermined number (when there are few feature points), the threshold setting unit 108 determines that the standard deviation threshold SD _TH is larger than the optimum value, and decreases the correction term β. In this way, by changing the correction term β according to the state of the input image, it is possible to always set the optimum standard deviation threshold value SD _TH, and as a result, more appropriate noise removal by the noise removing unit 104 is realized. can do.

  The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above-described embodiments, and various modifications and substitutions can be made to the above-described embodiments without departing from the scope of the present invention. Can be added.

For example, in the above-described embodiment, the standard deviation threshold value SD _TH set based on the input image of the current frame is used for the noise removal processing related to the next frame, but based on the input image of the current frame. It is also possible to use the set standard deviation threshold value SD _TH for the noise removal processing related to the current frame. That is, the standard deviation threshold value SD _TH (average value SD _AV ) is calculated and set before the noise removing process by the noise removing unit 104, and the noise removing process (threshold value) by the noise removing unit 104 using the standard deviation threshold value SD _TH. Determination processing) may be executed. Alternatively, the standard deviation based on the luminance levels of all the pixels included in the entire recognition target area of the difference image is calculated before the noise removing process by the noise removing unit 104, and the calculated standard deviation is used as the average value SD _AV. It is good.

In the above-described embodiment, the standard deviation threshold value SD _TH is determined based on the average value SD _AV of the standard deviation of the entire recognition target area. However, the recognition target area excluding the target area determined to have botts The standard deviation threshold value SD _TH may be determined based on the average value SD ′ _AV of the whole standard deviation. For example, if it is determined that the Botts 'dots in the target region SD ₂ are present, based on the average value of the standard deviation represented by the _{SD' AV = 1 / M (} SD 1 + SD 3 + ... + SD M), standard deviation A threshold value SD _TH may be determined.

  Further, as described above, the above-described embodiment is to detect botsdots with high accuracy while appropriately removing white lines and noise on a road in which an object larger than bottsdots (for example, white line) and botsdots are mixed. However, if it is on a road where only the same type of road section line (for example, a white line) exists, an appropriate structural element corresponding to the road section line is set, and a similar noise removal process is performed. It is possible to detect a road section line with high accuracy.

  In addition, the embodiment described above relates to a process suitable for detecting a road section line having a size (width) smaller than a normal white line, such as botsdots. It is possible to detect with high accuracy by setting appropriate structural elements accordingly.

  In the above-described embodiment, the feature points related to the object larger than the botsdots are appropriately removed by the morphological operation, but may be removed by other methods. For example, when a white line can be recognized by pattern matching or line edge processing, the white line is removed based on the recognition result, and the above-described noise removal processing is performed on the image from which the white line has been removed. While properly removing white lines and noise, botsdots can be detected with high accuracy.

Further, in the above-described embodiment, botsdots and the like are detected with high accuracy using luminance level variations, but a color image is used as an input image, and variations in signal levels representing hue and saturation are used. It is also possible to detect botsdots and the like with high accuracy. For example, when the standard deviation of the hue level is used, an appropriate standard deviation threshold SD _TH for the standard deviation of the hue level may be set in the same manner.

1 is a system configuration diagram including an embodiment of a road marking line detection device according to the present invention. It is explanatory drawing of a bots dot and a cat's eye. It is a functional block diagram which shows the principal part of the road marking line detection apparatus 100 by a present Example. It is a conceptual diagram for demonstrating morphological operation processing. It is a figure which shows an example of the setting aspect of the attention area | region with respect to a difference image. It is a figure which shows the difference image after the above-mentioned noise removal process by the noise removal part 104. FIG.

Explanation of symbols

10 Camera 30 Driving support ECU
DESCRIPTION OF SYMBOLS 100 Road marking line detection apparatus 102 Feature point extraction part 104 Noise removal part 106 Road marking line detection part 108 Threshold setting part

Claims (9)

An imaging means for capturing an image of a landscape including a road surface from a vehicle and acquiring a surrounding image;
The degree of signal level variation representing luminance, hue, or saturation in the region of interest of the surrounding image is calculated, and the feature points of the road section line are calculated in the region of interest based on whether the calculated degree of variation is greater than a threshold. Determining means for determining whether or not exists,
Threshold value setting means for calculating a variation level of a signal level representing luminance, hue or saturation in a region larger than the region of interest, and setting the threshold value based on the calculated variation level;
The features in the target area determined by the determining means that the feature points of the road section line existed while discarding all the feature points in the target area determined that the road section line feature points do not exist by the determining means. Feature point detecting means for detecting feature points of road section lines based only on the points,
A road segment line detection apparatus that detects a road segment line based on a feature point of the road segment line detected by the feature point detection means.   The road segment line detection device according to claim 1, wherein a plurality of the regions of interest are set, and the determination unit performs the determination for each region of interest.   The road segment line detection device according to claim 2, wherein the threshold value setting means sets the threshold value based on an average value of the degree of variation calculated for each region of interest.   The road section line according to any one of claims 1 to 3, wherein the threshold value setting means sets the threshold value based on the number of feature points in a region of interest determined that a feature point of a road section line exists. Detection device. The surrounding image is processed for each frame,
The road segment line detection apparatus according to any one of claims 1 to 4, wherein the threshold set by the threshold setting unit is used for determination by the determination unit for an image surrounding the next frame.   The road segment line detection device according to claim 1, wherein the degree of variation in the signal level is calculated based on a surrounding image after the morphological calculation processing.   The road section line detection device according to any one of claims 1 to 6, wherein the road section line is a kind of road lane marking provided at intervals. Obtaining a surrounding image including a road surface imaged by the imaging means from the vehicle;
The degree of signal level variation representing luminance, hue, or saturation in the region of interest of the surrounding image is calculated, and the feature points of the road section line are calculated in the region of interest based on whether the calculated degree of variation is greater than a threshold. A determination step of determining whether or not exists,
A threshold setting step of calculating a variation level of a signal level representing luminance, hue or saturation in a region larger than the region of interest, and setting the threshold based on the calculated variation level;
While discarding all the feature points in the attention area determined that the road section line feature points do not exist in the determination step, the features in the attention area determined in the determination step that the road section line feature points exist A feature point detecting step for detecting a feature point of a road section line based only on the point;
Detecting a road segment line based on the feature points of the road segment line in the region of interest detected in the feature point detecting step. A computer-readable program that causes a computer to execute a road section line detection process including the following processes (1) to ( 4 ).
(1) A process for acquiring a surrounding image including a road surface imaged by an imaging unit from a vehicle,
(2) The degree of variation of the signal level representing the luminance, hue, or saturation in the region of interest of the surrounding image is calculated, and the road segment line is included in the region of interest based on whether the calculated degree of variation is greater than a threshold value. A process for determining whether or not a feature point exists,
(3) Discarding all the feature points in the target area determined that the feature points of the road section line do not exist, and only based on the feature points in the target area determined to have the feature points of the road section line , Processing for detecting feature points of road section lines, and
(4) A process of detecting a road segment line based on the detected feature points of the road segment line in the detected region of interest.

Images