JP6515704B2 - Lane detection device and lane detection method - Google Patents

Description

  The present invention relates to, for example, a lane detection device that detects a lane and a lane detection method.

  Research on technology to detect various objects or road markings around the vehicle from images obtained by photographing the surroundings of the vehicle with a camera mounted on the vehicle for the purpose of assisting the driver of the vehicle etc. It is done.

  For example, a technique for detecting a lane in which the vehicle is traveling by detecting a lane line, a road center line, or a road outside line that represents a lane or road boundary from an image of the surroundings of the vehicle. Is being studied. The detection result is used, for example, to detect that the vehicle has run out of the lane in which it is traveling and to warn the driver. Hereinafter, such a dividing line is referred to as a lane dividing line for the sake of convenience.

  However, in parallel with the lane markings, the road may be provided with an auxiliary line for promoting deceleration. Such an auxiliary line is drawn, for example, as a broken line (hereinafter referred to as a short broken line) in which each block is shorter than a lane marking in the case of drawing the broken line. If an auxiliary line is provided, a plurality of parallel lines will be present along the extension direction of the road. Therefore, techniques have been proposed that can detect lane markings even when such multiple lines exist (see, for example, Patent Documents 1 to 3).

  For example, when a plurality of road division line candidates are detected on one side of the lane on which the vehicle travels by edge extraction, the image processing apparatus for a vehicle disclosed in Patent Document 1 compares the edge cycles of the respective candidates. The candidate closest to the continuous line is determined as the road division line.

  Further, the lane displacement correction system disclosed in Patent Document 2 calculates at least one lane width type based on the recognized lane width calculated from the recognized lane position, and according to the appearance frequency of the lane width type. Identify the main lane width. The lane displacement correction system determines the positional relationship between the main lane and the auxiliary lane based on the recognition lane width, the at least one lane width type, and the main lane width when the lanes on both sides are recognized. And correct the lane position according to the judgment result.

  Furthermore, the road shape estimation device disclosed in Patent Document 3 detects a plurality of edge lines from an image obtained by photographing the road surface of the road, and connects one or more edge lines among those edge lines to form a composite edge line. Generate The road shape estimation apparatus obtains road shape parameters of the road including the lateral offset of the camera that has photographed the road surface and the road width based on the information of the inner composite edge line closest to the center of the road. Then, the road shape estimation device obtains a correction amount based on the distance between the inner composite edge line and the outer composite edge line located outside the inner composite edge line, and calculates the lateral offset and the road width as the correction amount. Use to correct.

JP, 2004-310522, A Japanese Patent Application Laid-Open No. 2002-163642 JP, 2013-120458, A

  In the technique described in Patent Document 1, in order to accurately identify a lane line, the lane line and the auxiliary line can be distinguished on the image, and the edge period difference between the lane line and the auxiliary line Are required to be identifiable. However, when the resolution of the camera mounted on the vehicle is insufficient, the lane markings and the auxiliary lines may not be distinguishable on the image. In particular, when the camera mounted on the vehicle is a wide-angle camera, the range on the image corresponding to the area away from the vehicle becomes smaller, so the lane markings and the auxiliary lines included in such an area are identified. It is difficult to do.

  Further, since the auxiliary line is not always provided, there are lanes provided with the auxiliary line and lanes provided with no auxiliary line. On the other hand, in the technique described in Patent Document 2, the appearance frequency of the lane width type is used to specify the main lane width, so it can not cope with the switching of the presence or absence of the auxiliary line.

  Furthermore, the actual auxiliary line may be blurred due to wear or the like, and in such a case, it is difficult to detect the auxiliary line on the image in which the road is photographed. In addition, since the blocks forming the auxiliary line are close to each other, the auxiliary line may appear as a solid line instead of a short broken line on the image. In such a case, in the technique described in Patent Document 3, the auxiliary line may not be detected as the inner composite line, and as a result, there is a possibility that the position of the lane marking may be misrecognized.

  In one aspect, the present invention aims to provide a lane detection device capable of accurately detecting the lane in which the vehicle is traveling from an image obtained by photographing the surroundings of the vehicle.

  According to one embodiment, a lane detection device is provided. The lane detection device includes a storage unit for storing a first lane shape parameter representing the shape of a lane in which the vehicle is traveling, and an image on the image of the surroundings of the vehicle generated by an imaging unit mounted on the vehicle. A lane line candidate detection unit that detects lane line candidates that are candidates for lane lane lines that represent lane boundaries and lane shape parameter calculation that calculates a second lane shape parameter that indicates the lane shape based on the lane line candidates And a model generation unit that generates a plurality of models representing the shape of the lane by changing the position of the lane markings represented by the markings and the lane markings in the direction of expanding the width of the lane; , A model identification unit for identifying a model most similar to the lane shape represented by the first lane shape parameter, and a second lane shape parameter corrected to represent the lane shape corresponding to the identified model And a correction unit to be stored in the storage unit the second lane shape parameters as the first lane shape parameter.

  The lane in which the vehicle is traveling can be accurately detected from an image obtained by photographing the surroundings of the vehicle.

(A) And (b) is a figure which shows an example of an image in which it is difficult to detect a lane line and an auxiliary line. FIG. 1 is a block diagram of a vehicle equipped with a driving support device that is an embodiment of a lane detection device. It is a schematic block diagram of the driving assistance device by one embodiment. It is a functional block diagram of a control part. It is a figure which shows typically the image obtained by a camera. (A) is a figure which shows an example of the edge pixel detected from the image shown by FIG. 5, (b) shows an example of the edge line to which the edge pixel shown by (a) was connected. FIG. (A) is a figure which shows an example of area | region RR and RL, and an example of the edge line extracted, (b) is a figure which shows an example of edge lines other than noise in area | region RR and RL. It is a figure which shows an example of the compound edge line calculated | required from the example of the edge line shown by FIG.7 (b). It is a figure showing an example of a transition of a multiline state. It is a figure which shows the relationship between the parameter showing a lane shape, and the model of the appearance of the lane in an image. It is a figure which shows an example of the model of a lane shape. It is an operation flowchart of division line candidate determination processing. 5 is an operation flowchart of driving support processing including lane detection processing. It is a block diagram of the computer which operate | moves as a lane detection apparatus, when the computer program which implement | achieves the function of each part of the control part of the lane detection apparatus by said each embodiment or its modification operate | moves.

  The lane detection device will be described below with reference to the drawings. First, an example of an image in which it is difficult to accurately detect a lane will be described.

  FIG. 1A is a view showing an example of an image in which a blurred lane line and an auxiliary line are shown. In the image 100, the lane line 101 and the auxiliary line 102 on the right side of the lane in which the vehicle is traveling are blurred in the vicinity of the vehicle, and part of the lane line 101 and the auxiliary line 102 in the vicinity of the vehicle on the image 100 It has become difficult to identify Furthermore, in the distance of the vehicle, each block appears to be connected on the image 100 because the distance between the blocks of the auxiliary line 102 is small. As a result, it is difficult to identify that the auxiliary line 102 is a short broken line specific to the auxiliary line.

  In FIG. 1B, the blocks of the auxiliary line 111 appear to be connected in the image 110 because the intervals between the blocks of the auxiliary line are narrow. Therefore, in the image 110, it is difficult to accurately identify whether the auxiliary line 111 is an auxiliary line or a lane line.

  Thus, it is difficult to accurately detect lane markings and auxiliary lines from one image, and as a result, it may be difficult to detect a lane. Therefore, the lane detection device detects candidates for lane markings from the latest image, and creates a plurality of models representing the shape of the lane based on the candidates. Then, the lane detection device identifies a model most similar to the lane shape detected at the time of image acquisition immediately before among a plurality of models, and acquires the lane shape represented by the identified model as the latest image acquisition time Lane shape seen from the camera mounted on the vehicle.

FIG. 2 is a block diagram of a vehicle equipped with a driving support device that is an embodiment of the lane detection device.
The driving assistance device 1 is mounted on a vehicle 10 and is connected to each other by a camera 2 and a display 3 and a network 4. In FIG. 2, for convenience of explanation, the shape and size of the vehicle 10 and the arrangement of the driving support device 1 and the like are different from actual ones.

  In the present embodiment, the driving support device 1 detects the lane in which the vehicle 10 is traveling from the image obtained by capturing the front of the vehicle 10 with the camera 2. Then, when the vehicle 10 runs out of the lane, the driving support device 1 issues a warning to the driver via the display 3. The details of the driving support device 1 will be described later.

  The camera 2 is an example of an imaging unit, and is disposed, for example, in the cabin of the vehicle 10 toward the front of the vehicle 10, and generates an image representing the front of the vehicle 10. To that end, the camera 2 has an image sensor having a plurality of solid-state imaging devices arranged in a two-dimensional manner, and an imaging optical system for forming an image in front of the vehicle 10 on the image sensor.

  The camera 2 captures an area in front of the vehicle 10, for example, at fixed time intervals (for example, 1/30 seconds), and generates an image in which the area in front of the vehicle 10 is represented. Then, each time the camera 2 generates an image, the camera 2 transmits the image to the driving support device 1 via the network 4.

  The display 3 is an example of a display device, and can be, for example, a liquid crystal display. The display 3 is disposed in the compartment of the vehicle 10 so that the display screen faces the driver. For example, the display 3 is disposed in an instrument panel. Alternatively, the display 3 may be arranged independently of the instrument panel. Then, the display 3 displays information and the like received from the driving support device 1 via the network 4.

  The network 4 is a network for communicating between various devices mounted on the vehicle 10, and can be, for example, a network based on a control area network (CAN).

  FIG. 3 is a schematic block diagram of the driving support device 1. The driving support device 1 includes a storage unit 11, a communication unit 12, and a control unit 13.

  The storage unit 11 includes, for example, semiconductor memories such as electrically rewritable non-volatile memory and volatile memory. The storage unit 11 stores various computer programs executed by the control unit 13, for example, computer programs of driving support processing including lane detection. Furthermore, the storage unit 11 stores various data used in the driving support process. Furthermore, the storage unit 11 may store the image generated by the camera 2 for a certain period.

  The communication unit 12 includes a communication interface that communicates with the camera 2, the display 3, a vehicle control unit (not shown), and the like through the network 4 and a control circuit thereof. Then, the communication unit 12 passes the image received from the camera 2 via the network 4 to the control unit 13. The communication unit 12 also outputs the warning message received from the control unit 13 to the display 3. Furthermore, the communication unit 12 may receive information from various sensors that detect the behavior of the vehicle 10 (speed, acceleration, rotation angle of steering, etc.) via the network 4 and send it to the control unit 13.

  The control unit 13 includes one or more processors and their peripheral circuits, and executes driving support processing including lane detection processing.

  FIG. 4 shows a functional block diagram of the control unit 13. As shown in FIG. 4, the control unit 13 includes a lane line candidate detection unit 21, a line type determination unit 22, a multiline state update unit 23, a lane shape parameter calculation unit 24, a model generation unit 25, and a section. A line candidate determination unit 26, a correction unit 27, and a deviation determination unit 28 are included. These units included in the control unit 13 are implemented, for example, as functional modules realized by a computer program executed on a processor included in the control unit 13. Alternatively, these units included in the control unit 13 may be implemented as firmware. Alternatively, these units included in the control unit 13 may be implemented as a dedicated circuit. Note that among the units included in the control unit 13, the units other than the departure determination unit 28 relate to the lane detection process.

  Each time the driving support apparatus 1 receives an image from the camera 2, the lane line candidate detection unit 21 detects a lane line candidate that is a candidate for a lane line appearing on the image.

  FIG. 5 is a view schematically showing an image obtained by the camera 2. As shown in FIG. 5, in the image 500, the lane markings 501 and the auxiliary lines 502 are along the vertical direction of the image 500. And since lane markings 501 and auxiliary lines 502 are generally drawn on the road in white or yellow, they are brighter than the surrounding road surface. Therefore, on the image 500, the outline of the lane marking 501 and the outline of the auxiliary line 502 become an edge having a luminance difference in the horizontal direction on the image.

  Therefore, the parting line candidate detection unit 21 detects a parting line candidate by executing the same processing as the candidate line determination processing disclosed in, for example, JP 2013-120458A. Specifically, the dividing line candidate detection unit 21 applies an edge detection filter such as a sobel filter or a prewitt filter to each pixel of the image to detect an edge in the horizontal direction to calculate edge strength. . Then, the parting line candidate detection unit 21 detects a pixel for which the absolute value of the edge intensity is equal to or more than a predetermined threshold as an edge pixel that may correspond to the lane line or the outline of the auxiliary line.

  When an edge pixel is detected, the parting line candidate detection unit 21 performs an labeling process on each edge pixel to obtain an edge line in which the edge pixels are continuous. The marking line candidate detection unit 21 may detect an edge line by applying a Hough transform to a set of detected edge pixels.

  6 (a) is a diagram showing an example of edge pixels detected from the image 500 shown in FIG. 5, and FIG. 6 (b) is a diagram in which the edge pixels shown in FIG. 6 (a) are connected. It is a figure which shows an example of the edge line. As shown in FIG. 6A, each edge pixel 611 is detected along the outline of the lane marking 501 or the auxiliary line 502. Therefore, as shown in FIG. 6B, each edge line 612 corresponds to the entire lane marking drawn by a solid line, or one side of one block of a lane marking or an auxiliary line drawn by a broken line. It becomes a line to In FIG. 6A, each edge pixel 611 is represented such that each edge pixel 611 is separated for convenience of explanation, but in actuality, a plurality of edge pixels which are the same edge line are Adjacent to each other.

  The lane line candidate detection unit 21 deletes an edge line unrelated to the lane line or the auxiliary line among the detected edge lines. For example, in the outline of the lane line or auxiliary line located on the right side of the vehicle 10, which is closer to the center of the lane in which the vehicle 10 is traveling (hereinafter referred to as the inside for convenience) Is brighter than the left side of the contour. On the other hand, in the outline of the lane marking or auxiliary line located on the left side of the vehicle 10, which is farther from the center of the lane in which the vehicle 10 is traveling, the left side of the outline is more than the right side of the outline Too bright. Further, in the present embodiment, since the camera 2 is attached to the front of the vehicle 10, an area corresponding to the right side of the vehicle 10 is shown in the right half of the image obtained by the camera 2. It is assumed that an area corresponding to the left side of the vehicle 10 is shown in the left half. Therefore, in the right half of the image, an area RR is set where it is assumed that a lane line or an auxiliary line on the right side of the lane in which the vehicle 10 is traveling is shown. Similarly, in the left half of the image, an area RL is set in which it is assumed that a lane line or an auxiliary line on the left side of the lane in which the vehicle 10 is traveling is shown. Then, the dividing line candidate detection unit 21 determines that the average value of the brightness of the pixels adjacent to the right of the edge line is included in the region RR among the detected edge lines and the average value of the brightness of the pixels adjacent to the left of the edge line. Extract edge lines higher than the mean value. Similarly, among the detected edge lines, the dividing line candidate detection unit 21 is included in the region RL, and the average value of the luminance of the pixels adjacent to the left side of the edge line is adjacent to the right side of the edge line. Extract edge lines higher than the average value of luminance.

  If the edge line is a contour of a lane marking, the edge line has a certain length. Similarly, even if the edge line is an outline of the auxiliary line, the edge line has a certain length. However, since the auxiliary line is represented by a short broken line, it is assumed that if the edge line is an outline of the auxiliary line, it is shorter than the case where the edge line is a lane marking represented by a broken line.

  Therefore, the dividing line candidate detection unit 21 examines the length along the road surface in real space for each of the extracted edge lines, and according to the length, the type of the edge line is a solid line, a broken line, or a short broken line And noise. Then, the dividing line candidate detection unit 21 excludes very short edge lines classified as noise.

Assuming that the position of the upper end point of the edge line on the image is (x1, y1) and the position of the lower end point on the image is (x2, y2), expressed.
Here, H represents the height from the road surface to the camera 2, and fy represents the focal length of the camera 2 in the vertical direction, taking into account the horizontal and vertical pixel sizes. Also, p represents the pitch angle of the camera 2 with respect to the road surface. Then, (cx, cy) represents the center of the image, that is, the coordinates of the pixel on the image corresponding to the optical axis direction of the camera 2.

For each extracted edge line, the dividing line candidate detection unit 21 classifies the edge line as follows, for example, based on the length LL of the edge line.
When the length LL is longer than the long solid line length ThL1: edge line type = long solid line When the length LL is less than the long solid line length ThL1 and longer than the solid line length ThL2: edge line type = solid line Length LL is the solid line length If less than ThL2 and the broken line length ThL3 or more: edge line type = broken line If the length LL is less than the broken line length ThL3 and the short broken line length ThL4 or more: edge line type = short broken line Length LL is the short broken line When the length is less than ThL4: edge line type = noise Note that ThL1 to ThL4 are predetermined threshold values. ThL1 and ThL2 are set to values longer than the length of one block when the lane markings are represented by broken lines. Further, ThL3 is set to a value shorter than the length of one block when the lane markings are represented by broken lines. For example, ThL1 = 20 m, ThL2 = 5.5 m, ThL3 = 3.6 m, and ThL4 = 1 m. The edge line type corresponds to the length of one block included in the lane markings or auxiliary lines represented by the edge.

  FIG. 7A is a diagram showing an example of the regions RR and RL and an example of the edge line to be extracted. FIG. 7B is a diagram showing an example of edge lines other than noise in the regions RR and RL. As shown in FIG. 7A, in the right half of the image 700, the region RR is set so as to include a region substantially parallel to the traveling direction of the vehicle 10 and corresponding to the right side of the vehicle 10. Similarly, in the left half of image 700, region RL is set to include a region substantially parallel to the traveling direction of vehicle 10 and corresponding to the left side of vehicle 10. Then, among the detected edge lines, an edge line 701 included in the area RR or the area RL is extracted.

  Further, as shown in FIG. 7B, in the image 700, among the edge lines included in the region RR or RL, the edge line type is not noise, and of the luminance outside the inside of the edge line. Edge lines 702 that are higher are extracted.

  The lane line candidate detection unit 21 aligns the extracted edge lines in one row along the traveling direction of the vehicle 10 so as to correspond to the same lane line or auxiliary line for each of the areas RR and RL. Group by edge line. Then, the parting line candidate detection unit 21 sets groups of edge lines as compound edge lines.

  In the present embodiment, since the parting line candidate detection unit 21 executes the same processing for the regions RR and RL, the processing to be performed for the region RR will be described below. The dividing line candidate detection unit 21 sets one of the edge lines included in the region RR as an edge line to which attention is first focused. At that time, it is preferable that the parting line candidate detection unit 21 sets an edge line whose lower end is located on the lowermost side, that is, an edge line closest to the vehicle 10 as an edge line to which attention is first focused. On the image, the larger the position closer to the vehicle 10, the larger the edge line closer to the lower end of the image, and the more likely it is that the lane line or the auxiliary line is accurately represented.

The dividing line candidate detection unit 21 determines the depth direction on the road surface between the lower end side end point of the edge line and the upper end side end point of the focused edge line for each of the other non-grouped edge lines in the region RR. And the distance difference DX in the direction orthogonal to the depth direction. For example, the dividing line candidate detection unit 21 calculates the distance differences DZ and DX according to the following equation.
Here, c represents the curvature of the lane. Also, fx represents the focal length of the camera 2 in the horizontal direction, taking into account the horizontal and vertical pixel sizes.

  The dividing line candidate detection unit 21 specifies an edge line for which the distance difference DZ is equal to or less than the threshold ThDZ and the distance difference DX is equal to or less than the threshold ThDX. If the edge line type of the identified edge line and the edge line type of the focused edge line are the same or if the edge line type of the focused edge line is a short broken line, the parting line candidate detection unit 21 identifies the edge line type Included edge line in the same group as the focused edge line. Then, the parting line candidate detection unit 21 sets the identified edge line as the next focused edge line. The threshold ThDZ is set to, for example, 6 m, and the threshold ThDX is set to, for example, 10 cm. On the other hand, if there is no edge line satisfying the above condition that includes another edge line in the same group as the focused edge line, the parting line candidate detection unit 21 sets the group including the focused edge line as one compound edge line. . Then, the parting line candidate detection unit 21 sets one of the other edge lines not included in any compound edge line as an edge line to be focused next. For example, among the other edge lines not classified into any group, the dividing line candidate detection unit 21 sets an edge line whose lower end is located on the lowermost side as an edge line to be focused next. Then, the parting line candidate detection unit 21 newly creates a group including the focused edge line, and executes the same grouping processing as described above.

  The marking line candidate detection unit 21 obtains one or more composite edge lines by repeating the above grouping processing until there are no edge lines not classified into any group. Further, for each compound edge line, the parting line candidate detection unit 21 sets the edge line type of the longest edge line among edge lines included in the compound edge line as the edge line type of the compound edge line. Alternatively, the parting line candidate detection unit 21 sets the edge line type of the edge line closest to the vehicle 10 among the edge lines included in the compound edge line as the edge line type of the compound edge line. Good.

  FIG. 8 is a view showing an example of a composite edge line obtained from the example of the edge line shown in FIG. 7 (b). In this example, two composite edge lines 801 and 802 are determined in the region RR. In addition, two combined edge lines 803 and 804 are also required in the region RL.

When the compound edge line is determined, the parting line candidate detection unit 21 sets the compound edge line located at the innermost position as the parting line candidate used for estimation of the lane shape for each of the regions RR and RL. For example, in the example shown in FIG. 8, the composite edge line 801 and the composite edge line 803 are division line candidates. For each of the areas RR and RL, the dividing line candidate detection unit 21 determines, for each composite edge line in the area, the traveling direction of the vehicle 10 with respect to the lower end of the lowermost edge line in the composite edge line. Coordinates in the real space in the orthogonal direction (hereinafter referred to as the horizontal direction) are calculated. Then, the dividing line candidate detection unit 21 sets a compound edge line whose coordinates are closest to the center of the vehicle 10 as dividing line candidates for each of the areas RR and RL. As a result, even when an auxiliary line of an adjacent lane is provided outside the lane line of the lane in which the vehicle 10 is traveling, setting of the auxiliary line as a line candidate is suppressed. Therefore, it is suppressed that the width | variety of a lane is misrecognized by the auxiliary line of the adjacent lane. If the coordinates of the lower end of the lowermost edge line in the compound edge line on the image are (xs, ys), the coordinate Xs in the horizontal real space of the lower end is calculated according to the following equation.
However, k = 1 for the region RR, ie, the compound edge line on the right side of the vehicle 10, and k = −1 for the region RL, ie, the compound edge line on the left side of the vehicle 10.

  As described above, for each of the right side and the left side of the vehicle 10, the innermost composite edge line of the composite edge lines is set as the division line candidate. On the other hand, when an auxiliary line such as a deceleration mark is provided on the road separately from the lane line, the auxiliary line exists inside the lane line. Therefore, it is assumed that the compound edge line outside the candidate for the lane line represents a true lane line if both the lane line and the auxiliary line are identifiably captured in the image.

Therefore, the dividing line candidate detection unit 21 searches the area corresponding to one lane dividing line or one auxiliary line in real space outside the dividing line candidates in the horizontal direction with respect to each of the areas RR and RL. Set The search range [Xmin, Xmax] is set, for example, according to the following equation.
Here, Mmin and Mmax are set to, for example, 10 cm and 50 cm.

The parting line candidate detection unit 21 is other than the parting line candidate in which the horizontal coordinate of the lower end of the lowermost edge line included in the compound edge line is included in the search range on the right side of the vehicle 10, that is, for the region RR. Extract complex edge lines of Then, among the extracted compound edge lines, the parting line candidate detection unit 21 identifies a compound edge line whose coordinate in the depth direction at the lower end is closest to the vehicle 10. The identified composite edge line is likely to represent the inside contour of the lane marking. Therefore, the parting line candidate detection unit 21 stores the identified composite edge line in the storage unit 11 as a main parting line. In addition, the parting line candidate detection unit 21 is configured such that the horizontal coordinate of the lower end of the lowermost edge line included in the parting line candidate and the horizontal coordinate of the lower end of the lowermost edge line included in the main division line stored in the storage unit 11 the difference as a correction amount DX _R. Furthermore, the marking line candidate detection unit 21 increments the number of times of detection of the main marking line C _SF ^R by one, and sets the number of times of main marking line loss C _SR ^R representing the number of continuous failures to detect the main marking line to 0. Reset. In the case where there is no compound edge line in which the horizontal coordinate of the lower end of the lowermost edge line included in the compound edge line falls within the search range, the parting line candidate detection unit 21 has no main parting line I will judge. Then, the parting line candidate detection unit 21 increments by one the main parting line lost frequency C _SR ^R representing the number of continuous failures in detection of the main parting line. The lane line candidate detection unit 21 stores the number of times of detection of the main lane line C _SF ^R and the number of times of the main lane line loss C _SR ^R in the storage unit 11.

The dividing line candidate detection unit 21 similarly specifies the main dividing line also on the left side of the vehicle 10, that is, the region RL. Then, the lane line candidate detection unit 21 updates the correction amount DX _L for the left side of the vehicle 10, the number of times of detection of the main lane line C _SF ^L and the number of times of the main lane line loss C _SR ^L , and stores them in the storage unit 11.

  The line type determination unit 22 determines, for each of the left and right dividing line candidates of the vehicle 10, the indication line type of the dividing line candidate. The indication line type indicates whether the division line candidate is an auxiliary line or not by determination based on the information from the image in which the division line candidate is detected, and the final judgment whether the division line candidate is a lane line or an auxiliary line Does not represent the result of The indicator type is used to update the multiline state. In the present embodiment, the line type determination unit 22 determines the section to which attention is paid when the line type of the target division line candidate is a short broken line and the number N of edge lines included in the division line candidate is two or more. The marking line type of the line candidate is determined to be an auxiliary line. Further, in other cases, the line type determination unit 22 determines that the indicator line type of the section candidate line of interest is a normal line. Then, the line type determination unit 22 stores, in the storage unit 11, a flag indicating the determination result as to whether the dividing line candidate is an auxiliary line or a normal line for each dividing line candidate.

Table 1 is a table showing the correspondence between the combination of the edge line type of the dividing line candidate and the number of edge lines included in the dividing line candidate and the indication line type.

  Every time an image is obtained from the camera 2, the multiple line state update unit 23 updates multiple line states representing the detected states of the lane markings and the auxiliary lines for each of the left and right of the vehicle 10.

  In general, the auxiliary line is provided over a section having a certain length, so when the vehicle 10 is traveling in such a section, it is assumed that the auxiliary line is projected over a plurality of continuously obtained images . In such a case, ideally, for each of the left and right sides of the vehicle 10, the lane markings and the auxiliary lines appear in the image captured by the camera 2. Therefore, in each of the regions RR and RL, a plurality of composite edge lines Is present, ie, multiple lines are detected from the image. However, in practice, the auxiliary wire may be blurred due to wear and the like. In such a case, even if the vehicle 10 travels in a section provided with the auxiliary line, an image in which the auxiliary line is not captured may be obtained among a plurality of images in which the auxiliary line is captured. Therefore, even if an auxiliary line is detected in the images up to immediately before and no auxiliary line is detected in the latest image, the section in which the auxiliary line is actually provided does not necessarily end. Therefore, the multiline state update unit 23 changes the detection state of the multiple line in time series according to the detection result of the auxiliary line in the latest one or more images and the detection result of the auxiliary line in the latest image. Update the multiline state so that you can record In the following, updating of the multiline state on the right side of the vehicle 10 will be described, but the multiline state updating unit 23 performs the same processing on the left side of the vehicle 10 to update the multiline state.

  FIG. 9 is a diagram illustrating an example of the transition of the multiline state. As shown in FIG. 9, in the present embodiment, four types of states 901 to 904 are defined as multiline states. A state 901 (without multiple lines) represents a state without multiple lines. State 902 (discovery) represents the state in which multiple lines have been detected. Further, the state 903 (stable) represents a state in which multiple lines are stably detected. A state 904 (lost) indicates a state in which the multiple line once detected is missed. The states that can be transitioned from each state and the transition conditions are predetermined. For example, from state 901 (without multiple lines), transition to state 902 (discovery) is possible. Also, transition from the state 902 (discovery) to the state 901 (without multiline) or the state 903 (stable) is possible. Similarly, from state 903 (stable), transition to state 902 (discovered) or state 904 (lost) is possible. Then, from state 904 (lost), transition to state 901 (without multiple lines) is possible.

  The condition for transition from the state 901 to the state 902 is that the indicator line type of the division line candidate detected in the latest image is an auxiliary line. On the other hand, if the marking line type of the dividing line candidate detected in the latest image is the normal line in the state 901, the state 901 is maintained.

Further, the condition for transitioning from the state 902 to the state 901 is that the marking line type of the division line candidate detected in the latest image is a normal line. On the other hand, the transition condition from state 902 to state 903 is that the indicator line type of the dividing line candidate detected in the latest image is an auxiliary line, and the number of times of detection of the main dividing line at the time of the image C _SF ^R Is equal to or greater than a predetermined number threshold ThN1. The number threshold ThN1 is set to, for example, 1 to 5. If these conditions are not satisfied, that is, the indicator line type of the dividing line candidate detected in the latest image is an auxiliary line, and the number of times of detection of the main dividing line at the time of the image C _SF ^R is the number threshold value If less than ThN1 or the main dividing line is not detected, the state 902 is maintained.

The condition for transitioning from state 903 to state 902 is that the indicator line type of the dividing line candidate detected in the latest image is an auxiliary line, and the main dividing line loss frequency C _SR ^R is a predetermined threshold value ThN2 or more. It is to become. The threshold ThN2 is set to, for example, 20 to 60. On the other hand, the condition for transitioning from the state 903 to the state 904 is that the indicator line type of the division line candidate detected in the latest image is a normal line. If these conditions are not satisfied, that is, if the marking line type of the dividing line candidate detected in the latest image is an auxiliary line and the main dividing line loss frequency C _SR ^R is less than the threshold ThN 2, the state 903 Is maintained.

Furthermore, the condition for transitioning from the state 904 to the state 903 is that the indicator line type of the division line candidate detected in the latest image is an auxiliary line. On the other hand, the condition for transition from the state 904 to the state 901 is that the number of times the state 904 is maintained continues for a predetermined number of times or more. The predetermined number of times is set to, for example, 5 to 15. If a transition from state 904 to state 901, multiplet state updating unit 23 resets the number of times of detection C _SF ^R 0. If these conditions are not satisfied, that is, if the marking line type of the dividing line candidate detected in the latest image is a normal line, and the number of times the state 904 is maintained is less than a predetermined number of times, the state 904 is Maintained. Then, the number of times the state 904 is maintained is incremented by one.

  Each time the control unit 13 acquires an image from the camera 2, the multiple line state update unit 23 is shown in FIG. 9 based on the marking line type of the division line candidate detected from the image and the preceding multiple line state. The multiline state is updated for each of the left and right of the vehicle 10 according to the transition diagram. Then, the multiple line state update unit 23 stores, in the storage unit 11, information representing the updated multiple line state for each of the left and right of the vehicle 10.

  The lane shape parameter calculation unit 24 is a lane shape parameter that represents the shape of the lane in which the vehicle 10 is traveling, based on the division line candidate obtained for the image each time the control unit 13 acquires an image from the camera 2 Calculate

  In the present embodiment, the lane shape parameter calculation unit 24 uses the lane width W, the road curvature c, the distance from the camera 2 to the center of the lane in which the vehicle 10 is traveling (hereinafter referred to as the center distance) E) The azimuth angle t of the camera 2 and the depression angle p of the camera 2 are calculated.

FIG. 10 is a diagram showing the relationship between parameters representing the lane shape and a model of how the lane looks in the image. As shown in FIG. 10, a lane 1000 having a radius of curvature R (= 1 / c) and a width W is provided at an origin O located at a height H from the road surface and a distance e from the center of the lane 1000 It is assumed that the image is captured in an image 1001 captured by the camera 2 facing the azimuth angle t and the depression angle p. In FIG. 10, the optical axis direction of the camera 2 is represented by Zc. In this case, a model equation representing the relationship between the line corresponding to the dividing line candidate on the road to be estimated and the position (x _i , y _i ) on the image is given by, for example, the following equation.
However, in the coordinate system of the image 1001, the right direction of the horizontal coordinate x is positive, and the vertical direction y is positive. Then, the upper left end pixel of the image 1001 is set as the origin (0, 0). Further, in the equation (5), the parameter k is a parameter indicating whether the vehicle 10 is on the right side or the left side, and k = −1 for the left side and k = 1 for the right side.

Since the equation (5) is non-linear, first, the lane shape parameter calculator 24 linearly approximates the equation (5) with the parameters (W, c, e, t, p) of the lane shape. This linear approximation is expressed by the following equation.
Where s _i is the residual for (x _i , y _i ), and (W 0, c 0, e 0, t 0, p 0) are the initial values of (W, c, e, t, p) respectively It is. This initial value can be, for example, a parameter value of the lane shape at the time of image acquisition immediately before.

The lane shape parameter calculation unit 24 calculates the coordinates (x _L ⁱ , y _L ⁱ ) and (x _R ⁱ , y _R ⁱ ) of the edge pixels on each edge line included in each lane line candidate according to the obtained linear approximation equation. Substituting and finding the parameter solution by the least squares method. Then, the lane shape parameter calculation unit 24 obtains deviations (ΔW, Δc, Δe, which are absolute values of differences between the parameter solutions for the respective parameters and the initial values so that the parameter solutions are similarly obtained as the next initial values. The process is repeated until Δt, Δp) becomes less than or equal to the threshold. Then, the lane shape parameter calculation unit 24 calculates the parameter values (Wr, cr, er, tr, pr) when the deviations (.DELTA.W, .DELTA.c, .DELTA.e, .DELTA.t, .DELTA.p) become equal to or less than the threshold as the lane shape parameter (W). , c, e, t, p).

  The lane shape parameter calculation unit 24 can execute this repetitive calculation by executing the same procedure as the procedure of the road state estimation processing disclosed in Japanese Patent Laid-Open No. 2013-120458, and the lane shape parameter (W, c, e, t, p) can be obtained. Alternatively, the lane shape parameter calculation unit 24 applies the lane shape parameter (W, c) by applying another optimization method such as another gradient method or simulated annealing to the equation (5) or (6). , e, t, p) may be obtained. The lane shape parameter calculation unit 24 stores lane shape parameters (W, c, e, t, p) in the storage unit 11.

  The model generation unit 25 determines the position of the lane marking represented by the marking candidate based on the lane shape parameters (W, c, e, t, p) obtained for the latest image and the correction amount. Multiple lane widths are generated depending on whether or not the width is changed in the expanding direction.

  FIG. 11 is a diagram illustrating an example of a lane shape model. In the present embodiment, four types of models 1101 to 1104 are assumed. The model 1101 corresponds to the case where both of the left and right dividing line candidates 1111 and 1112 represent a lane line. Therefore, in the model 1101, right and left division line candidates are respectively set as lane division lines.

  The model 1102 corresponds to the case where the right dividing line candidate 1112 of the vehicle 10 represents a lane dividing line, but the dividing line candidate 1111 of the left of the vehicle 10 represents an auxiliary line. Therefore, in the model 1102, the lane line candidate 1112 is used as the lane line for the right side of the vehicle 10, while the lane line candidate 1111 for the left side of the vehicle 10 is shifted outward by the correction amount for the left side of the vehicle 10. The moved line 1121 is taken as a lane line.

  The model 1103 corresponds to the case where the lane line candidate 1111 on the left side of the vehicle 10 represents a lane line, while the lane line candidate 1112 on the right side of the vehicle 10 represents an auxiliary line. Therefore, in the model 1103, the lane line candidate 1111 is used as the lane line for the left side of the vehicle 10, while the lane line candidate 1112 for the right side of the vehicle 10 is shifted outward by the correction amount for the right side of the vehicle 10. The moved line 1122 is taken as a lane line.

  The model 1104 corresponds to the case where the left and right dividing line candidates 1111 and 1112 both represent auxiliary lines. Therefore, in the model 1104, lines 1121 and 1122 which are moved outward by the correction amount for the left side or the right side of the vehicle 10 from the lane line candidates 1111 and 1112 for the left and right sides of the vehicle 10 are taken as lane line.

The model generation unit 25 determines the latest image of the lane shape parameters (W _p ^(k) , c _p ^(k) , e _p ^(k) , t _p ^(k) , p _p ^(k) ) of each model. The parameters (W, c, e, t, p) of the obtained lane shape are substituted. The index k is a parameter representing the type of model, and when k = 1, the parameter of the lane shape represents the model 1101 (no enlargement of the lane width on both left and right). Similarly, if k = 2, the lane shape parameter represents model 1102 (expanding the lane width for the left side). Also, if k = 3, the lane shape parameter represents model 1103 (expanding the lane width for the right side). Then, if k = 4, the lane shape parameter represents the model 1104 (both the left and the right expand the lane width).

The model generation unit 25 generates lane shape parameters (W _p ⁽²⁾ , c _p ⁽²⁾ , e _p ⁽²⁾ , t _p ⁽²⁾ , p _p ⁽²⁾ ) in order to create the model 1102. Among them, the lane width W _p ⁽²⁾ and the center distance e _p ⁽²⁾ are corrected according to the following equation.
Here, L _L is a correction amount for the left side of the vehicle 10.

Similarly, the model generation unit 25 generates lane shape parameters (W _p ⁽³⁾ , c _p ⁽³⁾ , e _p ⁽³⁾ , t _p ⁽³⁾ , p _p ^{(3 The} lane width W _p ⁽³⁾ and the center distance e _p ⁽³⁾ are corrected according to the following equation.
Here, L _R is a correction amount for the left side of the vehicle 10.

Further, the model generation unit 25 generates lane shape parameters (W _p ⁽⁴⁾ , c _p ⁽⁴⁾ , e _p ⁽⁴⁾ , t _p ⁽⁴⁾ , p _p ⁽⁴⁾ in order to create the model 1104 ^). Lane width W _p ⁽⁴⁾ and center distance e _p ⁽⁴⁾ are corrected according to the following equation.

Note that the model generation unit 25 stores the correction amount L _R for the right side of the vehicle 10 in the storage unit 11 at that time when the multiple line state for the right side of the vehicle 10 is “stable” or “lost”. the correction amount DX _R stored. On the other hand, when the multiple line state about the right side of the vehicle 10 is “discovered”, the model generation unit 25 sets the correction amount L _R as a preset offset value. Similarly, if the correction amount L _L for the left side of the vehicle 10 is “stable” or “lost” for the left side of the vehicle 10, the model generation unit 25 stores the correction amount L _L at that time. It is assumed that the correction amount DX _L stored in On the other hand, when the multiple line state of the left side of the vehicle 10 is “discovered”, the model generation unit 25 sets the correction amount L _L as a preset offset value. This offset value is set, for example, to a common value, for example 45 cm, of the difference in horizontal position between the auxiliary line and the lane markings.

  The model generation unit 25 stores lane shape parameters of a model of each lane shape in the storage unit 11.

  The dividing line candidate determination unit 26 determines whether the dividing line candidate is a lane dividing line or an auxiliary line based on the edge line type of the dividing line candidate and the multiline state determined for each of the left and right of the vehicle 10 based on the latest image. Determine if Further, if the parting line candidate determination unit 26 can not determine whether the parting line candidate is a lane line part or an auxiliary line from the edge line type of the parting line candidate and the multiline state, the parting line candidate determination unit 26 immediately before the lane shape model. Identify the model most similar to the lane shape at the time of image acquisition. Then, the lane line candidate determining unit 26 determines whether the lane line candidate is a lane line or an auxiliary line based on the identified model. The marking line candidate determination unit 26 is an example of a model specification unit.

  FIG. 12 is an operation flowchart of the parting line candidate determination process performed by the parting line candidate determination unit 26. Each time the control unit 13 acquires an image from the camera 2, the lane line candidate determination unit 26 determines the left and right lane line according to the operation flowchart. In this operation flowchart, determination of a lane line on the right side of the vehicle 10 will be described, but the lane line candidate determination unit 26 may determine a lane line according to the same operation flowchart on the left side of the vehicle 10. At that time, “right” in the following operation flowchart is read as “left”.

  The marking line candidate determination unit 26 determines whether the edge line type of the marking line candidate is a long solid line (step S101). If the edge line type of the dividing line candidate is a long solid line (step S101-Yes), there is no such long solid line in the auxiliary line, so it is assumed that the dividing line candidate represents a lane line. Therefore, the lane line candidate determination unit 26 determines that the lane line candidate is a lane lane line (step S102). As a result, the lane line candidate determination unit 2 can accurately estimate the position of the lane line.

  On the other hand, when the edge line type of the parting line candidate is not a long solid line (step S101-No), the parting line candidate determination unit 26 determines whether the edge line type of the parting line candidate is a short broken line (step S103). If the edge line type of the dividing line candidate is a short broken line (step S103-Yes), the dividing line candidate determination unit 26 determines whether or not the multiple line state on the right side of the vehicle 10 at this time is "stable" (step S104). When the multiline state is "stable" (step S104-Yes), since the multiple line is detected on the right side of the vehicle 10 continuously in time series, the division line candidate which is a short broken line is a lane line It is assumed to represent the inner auxiliary line of. Therefore, the parting line candidate determination unit 26 determines that the parting line candidate is an auxiliary line provided inside the lane line (step S105). As a result, the lane line candidate determination unit 2 can accurately estimate the position of the lane line.

  On the other hand, when the multiple line state is not "stable" (step S104-No), the fact that the division line candidate is a short broken line may be attributed to the fading of the lane line. Further, even if the edge line type of the dividing line candidate is not a short broken line at step S103 (No at step S103), the dividing line candidate determination unit 26 determines that the dividing line candidate is a lane section from the edge line type of the division line candidate. I can not identify the line or the auxiliary line. Therefore, the lane line candidate determination unit 26 compares the lane shape at the time of the previous image acquisition with the model of each lane shape. Then, the parting line candidate determination unit 26 specifies a model most similar to the lane shape at the time of the previous image acquisition among the models of each lane shape (step S106).

The lane line candidate determination unit 26 determines whether the model most similar to the lane shape at the time of the previous image acquisition is any of the models (the model 1103 or 1104 in FIG. 11) that enlarges the lane width on the right side (step S107). ). If the model most similar to the lane shape at the time of image acquisition at the previous time is any of the models for expanding the lane width on the right side (step S107-Yes), the parting line candidate determination unit 26 It is determined that the auxiliary line is provided inside the line (step S105).
On the other hand, if the model most similar to the lane shape at the time of image acquisition at the previous time is not a model for enlarging the lane width on the right side (step S107-No), the lane line candidate determination unit 26 determines the lane line candidate as a lane line It determines (step S102).
After step S102 or S105, the parting line candidate determination unit 26 ends the parting line candidate determination process.

In step S106, the dividing line candidate determination unit 26 sets, for example, the similarity between the lane shape and each model at the time of the previous image acquisition as the sum df of horizontal errors of the left and right lane dividing lines of the vehicle 10. Calculated according to the following equation. Then, the lane line candidate determination unit 26 sets a model that minimizes the sum df of errors as a model that is most similar to the lane shape at the time of the previous image acquisition.
Here, W _b and e _b are the lane width and the center distance at the time of the previous image acquisition, respectively.

  The dividing line candidate determining unit 26 notifies the correcting unit 27 of the determination result of dividing line candidates for each of the left and right of the vehicle 10.

  The correction unit 27 determines the lane width among the lane shape parameters (W, c, e, t, p) obtained for the latest image based on the determination result of the dividing line candidate for each of the left and right of the vehicle 10 By correcting W and the center distance e, the positions of the lane lines on the left and right of the vehicle 10 are obtained. Thus, the corrected lane shape parameters (W, c, e, t, p) can represent the shape of the lane in which the vehicle 10 is traveling as viewed from the camera 2 at the time of the latest image acquisition.

Correcting unit 27, when the lane line candidate of the left side of the vehicle 10 represents a lane line, a correction amount DX _L stored left correction amount K _L in the storage unit 11. On the other hand, when the dividing line candidate on the left side of the vehicle 10 represents an auxiliary line, the left correction amount K _{L is set} to zero. Similarly, the correction unit 27, if the right-hand lane line candidate of the vehicle 10 represents a lane line, a correction amount DX _R stored right correction amount K _R in the storage unit 11. On the other hand, when the dividing line candidate on the right side of the vehicle 10 represents an auxiliary line, the right correction amount K _{R is set} to zero.

The correction unit 27 corrects the lane width W and the center distance e according to the following equation.
The correction unit 27 stores the corrected lane width W and the center distance e in the storage unit 11.

  The departure determining unit 28 determines whether the vehicle 10 has deviated from the lane based on the calculated lane shape parameter. For example, the departure determining unit 28 determines that the vehicle 10 has deviated from the lane when (W / 2− | e |) becomes equal to or less than 1⁄2 of the width of the vehicle 10. Then, the departure determination unit 28 causes the display 3 to display a warning message to that effect when the vehicle 10 is out of the lane or when there is a possibility of going out of the lane. Alternatively, when a speaker (not shown) is disposed in the compartment of the vehicle 10, the departure determination unit 28 may cause the speaker to output an alarm sound.

  FIG. 13 is an operation flowchart of driving support processing including lane detection processing, which is executed by the control unit 13. Each time the driving support device 1 receives an image from the camera 2, the control unit 13 executes the driving support process according to the following operation flowchart. Note that steps S201 to S208 among the following steps S201 to S209 correspond to the lane detection process.

The dividing line candidate detection unit 21 detects the left and right dividing line candidates of the vehicle 10 on the image (step S201). Furthermore, the parting line candidate detection unit 21 determines whether or not the main parting line exists outside the parting line candidate, and according to the determination result, the number of times of detection of the main parting line (C _SF ^R , C _SF ^L And the number of times that the main dividing line is lost ( _CSR ^R , C _SR ^L ) and the correction amount (DX _R , DX _L ) are updated (step S202).

  The line type determination unit 22 determines the marking line type of the dividing line candidate for each of the dividing line candidates on the left and right of the vehicle 10 (step S203). Then, the multiple line state update unit 23 updates the multiple line state for each of the left and right sides of the vehicle 10 based on the indicator line type of each division line candidate, the number of times of detection of the main division line, the number of main division line losses, etc. (Step S204).

  Further, the lane shape parameter calculation unit 24 calculates lane shape parameters (W, c, e, t, p) representing the shape of the road on which the vehicle 10 is traveling, based on the left and right division line candidates of the vehicle 10 (Step S205). The model generation unit 25 generates a plurality of lane shape models based on the lane shape parameters (W, c, e, t, p), the left and right multiline states of the vehicle 10, and the correction amount. (Step S206).

  The lane line candidate determination unit 26 determines that the lane line candidates for the left and right sides of the vehicle 10 are lanes based on the edge line types of the lane line candidates for left and right of the vehicle 10, the multiline state, the model of each lane shape, and the last lane shape. It is determined whether the dividing line or the auxiliary line is represented (step S207). Then, the correction unit 27 corrects the lane width W and the center distance e among the lane shape parameters (W, c, e, t, p) based on the determination result (step S208).

  The departure determination unit 28 determines whether or not the vehicle 10 has deviated from the lane in which the vehicle 10 is traveling based on the corrected lane width W and the center distance e, and executes warning processing according to the determination result (step S209). Thereafter, the control unit 13 ends the driving support process.

  As described above, the driving support apparatus, which is an example of the lane detection apparatus, generates a plurality of models representing the lane shape based on the parting line candidates detected from the latest image. Then, the driving support device identifies a model that is most similar to the lane shape at the time of image acquisition immediately before, of each model, and corrects the position of the dividing line candidate as needed based on the model. Find the position of the lane markings. Thus, the driving support device can accurately detect the lane even if the lane markings or the auxiliary lines are blurred or difficult to distinguish on the image.

  The lane detection device according to the above-described embodiment or the variation thereof is not limited to the driving assistance device, and is used for various applications. For example, the lane detection device may be applied to applications such as analyzing a moving image recorded by a drive recorder and identifying a lane included in each image included in the moving image. In this case, the departure determination unit may be omitted. In addition, when applied to such an application, the camera mounted on the vehicle may be installed facing the rear of the vehicle. In addition, the lane detection device may be realized, for example, as a computer used for analysis of such a moving image.

  FIG. 14 is a block diagram of a computer that operates as a lane detection device by operating a computer program that implements the functions of the control unit of the lane detection device according to each of the above-described embodiments or the variation thereof.

  The computer 100 includes a user interface unit 101, a communication interface unit 102, a storage unit 103, a storage medium access device 104, and a processor 105. The processor 105 is connected to the user interface unit 101, the communication interface unit 102, the storage unit 103, and the storage medium access device 104 via, for example, a bus.

  The user interface unit 101 includes, for example, an input device such as a keyboard and a mouse, and a display device such as a liquid crystal display. Alternatively, the user interface unit 101 may have a device such as a touch panel display in which an input device and a display device are integrated. Then, the user interface unit 101 selects a moving image to be subjected to the lane detection process or outputs an operation signal for starting the lane detection process to the processor 105 according to the user's operation.

  The communication interface unit 102 may have a communication interface for connecting the computer 100 to a moving image input device (not shown) such as a drive recorder and a control circuit therefor. Such communication interface may be, for example, High-Definition Multimedia Interface (HDMI) (registered trademark) or Universal Serial Bus (Universal Serial Bus, USB).

  Furthermore, the communication interface unit 102 may have a communication interface for connecting to a communication network conforming to a communication standard such as Ethernet (registered trademark) and its control circuit. In this case, the communication interface unit 102 acquires a moving image to be subjected to the lane detection processing from another device connected to the communication network, and passes the moving image to the processor 105.

  The storage unit 103 includes, for example, a readable and writable semiconductor memory and a read-only semiconductor memory. Then, the storage unit 103 stores a computer program for executing the lane detection process, a moving image to be an object of the lane detection process, and the like that are executed on the processor 105.

  The storage medium access device 104 is a device that accesses the storage medium 106 such as, for example, a magnetic disk, a semiconductor memory card, and an optical storage medium. The storage medium access device 104 reads, for example, a computer program for lane detection processing, which is executed on the processor 105 stored in the storage medium 106, and passes it to the processor 105. The storage medium access device 104 may also read from the storage medium 106 a moving image to be subjected to lane detection processing.

  The processor 105 detects the lane for each image included in the moving image by executing the computer program for lane detection processing according to any one of the above-described embodiments or the modification. Then, the processor 105 stores the detection result in the storage unit 103 or outputs the detection result to another device via the communication interface unit 102.

  All examples and specific terms cited herein are intended for instructional purposes to help the reader understand the concepts contributed by the inventor to the present invention and the promotion of the art. It should be understood that the present invention is not to be limited to the construction of any of the examples herein, and to the specific listed examples and conditions relating to showing superiority and inferiority of the present invention. Although embodiments of the present invention have been described in detail, it should be understood that various changes, substitutions and modifications can be made thereto without departing from the spirit and scope of the present invention.

1 Driving support device (lane detection device)
Reference Signs List 2 camera 3 display 4 network 10 vehicle 11 storage unit 12 communication unit 13 control unit 21 division line candidate detection unit 22 line type determination unit 23 multiline state update unit 24 lane shape parameter calculation unit 25 model generation unit 26 division line candidate determination unit 27 correction unit 28 deviation judgment unit 100 computer 101 user interface unit 102 communication interface unit 103 storage unit 104 storage medium access device 105 processor 106 storage medium

Claims (5)

A storage unit storing a first lane shape parameter representing a shape of a lane in which the vehicle is traveling;
A parting line candidate detection unit for detecting a parting line candidate serving as a candidate for a lane parting line on the image captured by the imaging unit mounted on the vehicle and in which the surroundings of the vehicle are captured;
A lane shape parameter calculation unit that calculates a second lane shape parameter that represents the shape of the lane based on the lane line candidate;
A model generation unit that generates a plurality of models representing the shape of the lane according to whether or not the position of the lane line represented by the lane line candidate is changed in a direction to expand the width of the lane;
A model specification unit that specifies a model that is most similar to the shape of the lane represented by the first lane shape parameter among the plurality of models;
The second lane shape parameter is corrected to represent the shape of the lane corresponding to the specified model, and the corrected second lane shape parameter is stored in the storage unit as the first lane shape parameter. A correction unit to store;
Lane detection device having.   The parting line candidate detection unit detects a plurality of edge lines in which edge pixels are connected on the image, and classifies each of the plurality of edge lines according to groups aligned along the route of the vehicle. A group corresponding to a position closest to the vehicle among the groups within the range corresponding to the right side of the vehicle, and a group within the range on the image corresponding to the left side of the vehicle The lane detection device according to claim 1, wherein a group corresponding to a position closest to the vehicle is set as the division line candidate. The model specifying unit may set a predetermined value longer than one block of the lane markings, the length of which is represented by a broken line when projecting any of the edge lines included in the marking candidates into the real space. In some cases, the lane line candidate is determined to represent the position of the lane line regardless of the most similar model,
The lane detection device according to claim 2, wherein the correction unit corrects the second lane shape parameter in accordance with the position of the lane marking represented by the marking candidate. In the model specifying unit, the division line candidate includes at least two of the edge lines, and for each of the at least two edge lines, a length when the edge line is projected to a real space is indicated by a broken line. At a first position on the image which is equal to or less than a predetermined value shorter than one block of the lane markings and which corresponds to a position farther from the vehicle than the markings for a certain period in the past If the group is detected for a first predetermined number of times or more and the group does not fail to continuously detect for a second predetermined number of times or more at the first position, the section is detected regardless of the most similar model It is determined that the line candidate represents the position of an auxiliary line provided inside the lane with respect to the lane line,
The lane detection device according to claim 2, wherein the correction unit corrects the second lane shape parameter in accordance with a difference between the position of the auxiliary line represented by the division line candidate and the position of the lane division line. A lane line candidate which is a candidate for a lane lane line representing a lane boundary on which the vehicle is traveling is detected on the image, which is generated by an imaging unit mounted on the vehicle, and in which the surroundings of the vehicle are taken;
Calculating a lane shape parameter representing a shape of a lane on which the vehicle is traveling based on the division line candidate;
Generating a plurality of models representing the shape of the lane according to whether or not the position of the lane line represented by the line candidate is changed in a direction in which the width of the lane is expanded;
Among the plurality of models, a model that is most similar to the shape of the lane represented by the past lane shape parameter obtained from the image generated by the imaging unit in the past is specified,
Correcting the lane shape parameter to represent the shape of the lane corresponding to the identified model;
Lane detection method including.