JP3424334B2 - Roadway detection device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a traveling road detecting device for recognizing a traveling road region in an automobile or an automated guided vehicle by image processing.

[0002]

2. Description of the Related Art Various types of vehicles, such as automobiles and automatic guided vehicles, have various types for the purpose of "auxiliary alarm generation for danger avoidance", "assistance of human judgment operation", "fully automatic driving", etc. Roadway detection devices are being researched. An example of a traveling road detecting device is an image capturing input unit that collects an image of a traveling road surface and performs image processing to highlight the white line portion, and white line detection that describes the position of the white line on the road surface image using the coordinate system of the road surface image. And means for estimating the three-dimensional shape of the road based on the position information of the white line described in the coordinate system of the road surface image. In the traveling road detection device, it is necessary to determine the three-dimensional shape of the road surface in front of the vehicle every moment in a very short time of 1/10 to 1/100 seconds. However, enormous calculation processing is required to exactly determine the three-dimensional function of the road from the road surface image. Therefore, various approximations and simplifications are applied to the processing at each stage from image processing to estimation of a three-dimensional shape.

A traveling road detection device for a vehicle traveling on a general road was announced at "7th Image Sensing Symposium in Industry (July 7, 1992)", "Road structure and vehicle posture from continuous road images". "Real-time estimation". This paper describes a method for simultaneously estimating road curvature, gradient, and vehicle attitude (pitch angle, yaw angle, roll angle) from image data of white lines extracted from road surface images. FIG. 37 shows a conceptual diagram of the method. As the white line model, the road coordinate system (X,
A multi-dimensional curve described by Y, Z) is used. White line candidates are extracted from the input image, and from the image coordinates (x, y) to the camera coordinates (U, V, W), the camera coordinates (U, V,
After the two-step coordinate conversion from W) to road coordinates (X, Y, Z), the white line candidates are converted to the road coordinate system (X, Y, Z).
Z) projected onto the road model based on the white line candidate at the previous time. Based on the comparison result, each numerical value of the parameter of the multi-order curve equation is determined. While updating the multi-dimensional curve formula of the road model with these numerical values, the traveling path and the vehicle attitude described above are recognized.

[0004]

SUMMARY OF THE INVENTION The conventional roadway detecting apparatus according to the prior art (1) describes a road shape by a function of a white line and complements only the white line. (2) A limited number of white lines on the white line. Determine the coordinate value only for the candidate and use it for calculation,
(3) The function of the road model is kept constant, and only the parameter part included in the function is sequentially updated. The number of operations required is greatly reduced compared to the conventional processing. . As a result, although a somewhat special arithmetic device is required, it is possible to capture 30 screens per second and calculate and output the road shape, vehicle position, and vehicle attitude in real time.

However, from the viewpoint of practical implementation of actually mounting the traveling road detection device in a vehicle, the speed is further reduced and the capacity is reduced.
It is desired to be able to perform arithmetic output at a speed equal to or higher than that using an arithmetic device having a more general structure. In addition, it is desired that the arithmetic device mounted on the vehicle can interrupt not only the recognition of the three-dimensional shape of the road but also other arithmetic processing based on the recognition result. For example, it is a recognition of a preceding vehicle on a road surface image, a warning about a lane change, and the like. From this point of view, the conventional roadway detection device has to perform two-step coordinate conversion from the image coordinate system to the road coordinate system when comparing the road model and the white line candidate obtained from the image. There was a problem that the calculation became very complicated.

On the other hand, since a highway or the like which often uses the traveling road detecting device is composed of a gentle curve or gradient having a constant curvature of 300R or more, the road model is approximated by a multi-dimensional expression like the conventional example. Practical results may be obtained without doing so. Also, on highways, the roll angle of the vehicle can be ignored. After all, in the roadway detection device of the conventional example, when viewed from the error of the position coordinates obtained from the image processing,
There has been a problem that unnecessary approximations are performed on expressways.

The present invention further reduces the calculation load,
It is an object of the present invention to provide a traveling road detection device capable of outputting a three-dimensional shape of a road at high speed and with high accuracy even with a small computing capacity.

[0008]

According to another aspect of the present invention, there is provided an image pickup inputting means for collecting a road surface image in front of a vehicle and performing image processing to extract a white line on the road surface. In the running road detection device for detecting the white line and identifying the three-dimensional shape of the road surface and the relative positional relationship of the vehicle with respect to the road surface, in the parameter description expression obtained by converting the quadratic curve approximation expression of the white line into the coordinate system of the road surface image. , A vehicle position in the driving lane, a plane curvature of the road, a vehicle inclination with respect to the lane, a road inclination, a detection parameter holding means for holding a plurality of parameters respectively associated with the road width, and the white line on the road surface image. For a plurality of white line candidate points, white line candidate point detection means for obtaining position information using the coordinate system of the road surface image,
For virtual white line candidate points calculated from a plurality of past parameters held in the detection parameter holding means, the position information of the latest white line candidate points obtained from the road surface image is compared, and the plurality of parameters Parameter correction means for obtaining the latest value and updating the plurality of parameters of the detection parameter holding means.

According to a second aspect of the present invention, in the traveling road detecting apparatus according to the first aspect, the parameter correcting means calculates a deviation amount between the virtual white line candidate point and the latest white line candidate point. A deviation amount calculating means; and a parameter fluctuation amount calculating means for calculating fluctuation amounts of the plurality of parameters based on the deviation amount, and the parameter fluctuation amount calculating means uses the least squares method to calculate the parameters. The amount of fluctuation was calculated.

According to a third aspect of the present invention, there is provided the road route detecting apparatus according to the first or second aspect, wherein the parameter correcting means is based on the updated latest plurality of parameters. Road shape output means for calculating and outputting the inclination of the vehicle and the inclination of the road with respect to is included.

According to a fourth aspect of the present invention, there is provided a traveling road detecting device.
2 or 3, the parameter description formula includes a plurality of parameters associated with a vehicle position in a traveling lane, a plane curvature of a road, a vehicle inclination with respect to a lane, a road inclination, and a road width, respectively. a, b, c, d,
x, (y + d) + b / (y-d) + c, where e is the coordinate system of the road surface image, x is the coordinate system of the road image, and k is an integer.
It was supposed to be.

According to a fifth aspect of the present invention, there is provided an image pickup input means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line on the road surface, and at least a traveling lane based on the detection result of the white line. A road parameter estimating unit that calculates two road parameters associated with the width and the position of the own vehicle in the own lane with respect to each of the road surface images every moment, in a traveling road detecting device, along the estimated position of the white line. For each of the plurality of small areas set on the road surface image, white line candidate point detecting means for obtaining the passing position coordinates of the white line extracted by the image processing means, and the white line for setting the small area on the road surface image. And the number of small areas allocated to each of the white lines,
The white line candidate point detection area setting means set in the white line candidate point detection means, and own vehicle deviation determination means for identifying the own vehicle position in the own lane based on the road parameter, and the white line Candidate point detection area setting means, based on the own vehicle position, the number of white lines to set the small area,
The number of small areas assigned to each white line is changed so that when a vehicle approaches one white line, the small area is also assigned to another white line in the approach direction.

According to a sixth aspect of the present invention, there is provided the road route detecting apparatus according to the fifth aspect, wherein the white line candidate point detection area setting means determines a difference between a width of the traveling lane and a position of the own vehicle in the own lane. If the vehicle detects one of the white lines and the difference exceeds the predetermined judgment reference value range, the small area is set to the three white lines including the white line and the two white lines on both sides. I decided to allocate it.

According to a seventh aspect of the present invention, there is provided the road route detecting apparatus according to the fifth aspect, wherein the white line candidate point detection area setting means sets a difference between a width of the traveling lane and a position of the own vehicle in the own lane. If the difference is within the predetermined judgment reference value range after the vehicle has detected one of the white lines, the small area is assigned only to the two white lines on both sides of the new own lane. And

According to another aspect of the present invention, there is provided a traveling road detecting device.
6 or 7, the white line candidate point detection area setting means holds a constant set in advance in association with the width of the traveling lane, and the parameter estimation is performed for the width of the traveling lane. Regardless of the means, the constant value is used to detect the difference.

A traveling road detecting device according to a ninth aspect is the fifth aspect.
In the traveling road detection device described in 6, 7, or 8, the white line candidate point detection area setting means changes the total number of allocations of the small areas even if the number of white lines for setting the small areas changes. The small area is set without doing this.

A traveling road detecting device according to a tenth aspect is the traveling road detecting device according to any one of the fifth, sixth, seventh, eighth and ninth aspects, wherein the white line candidate point detection area setting means corresponds to three white lines. When allocating the small areas, more small areas are allocated to the approaching white line than to the other two white lines on both sides.

According to another aspect of the present invention, there is provided an image pickup inputting means for collecting a road surface image in front of a vehicle and performing image processing to extract a white line on the road surface. Road parameter estimating means for calculating, for each of the road surface images, two road parameters that are respectively associated with the width and the position of the own vehicle within the own lane;
In the traveling road detecting device, the white line candidate point detecting means for obtaining the passage position coordinates of the white line extracted by the image processing means for each of the plurality of small areas set on the road surface image along the estimated position of the white line. And a white line candidate point detection area to be set in the white line candidate point detection means by determining the number of white lines to set the small area on the road surface image and the number of small areas assigned to each of the white lines. Setting means, and own vehicle deviation determination means for identifying the own vehicle position in the own lane based on the road parameters, and the white line candidate point detection area setting means, based on the own vehicle position Then, the number of white lines to set the small area and the number of assigned small areas to each white line are changed. When the vehicle approaches one white line, another white line in the approaching direction is changed. Also assigns the small area and determines whether or not the turn indicator of the host vehicle is operated, and when the host vehicle approaches one of the white lines, the direction indicator determines the direction. An alarm generating means for outputting an alarm to the driver when the non-operation of the indicator is identified is provided.

According to a twelfth aspect of the present invention, there is provided an image pickup inputting means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line on the road surface. Road parameter estimating means for calculating, for each of the road surface images, two road parameters that are respectively associated with the width and the position of the own vehicle within the own lane;
In the traveling road detecting device, the white line candidate point detecting means for obtaining the passage position coordinates of the white line extracted by the image processing means for each of the plurality of small areas set on the road surface image along the estimated position of the white line. And a white line candidate point detection area to be set in the white line candidate point detection means by determining the number of white lines to set the small area on the road surface image and the number of small areas assigned to each of the white lines. Setting means, and own vehicle deviation determination means for identifying the own vehicle position in the own lane based on the road parameters, and the white line candidate point detection area setting means, based on the own vehicle position Then, the number of white lines to set the small area and the number of assigned small areas to each white line are changed. When the vehicle approaches one white line, another white line in the approaching direction is changed. Lane area recognizing means for allocating the small area and recognizing the own lane area on the road surface image, and obstacle detecting means for detecting an obstacle ahead of the lane area recognized by the lane area recognizing means. A collision possibility judgment means for judging the possibility of collision between the obstacle and the vehicle detected by the obstacle detection means, and an alarm for outputting a warning to the driver when the possibility of collision is affirmed And a generating means.

According to a thirteenth aspect of the present invention, there is provided an image pickup input means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line on the road surface. Road parameter estimating means for calculating, for each of the road surface images, two road parameters that are respectively associated with the width and the position of the own vehicle within the own lane;
In the traveling road detecting device, the white line candidate point detecting means for obtaining the passage position coordinates of the white line extracted by the image processing means for each of the plurality of small areas set on the road surface image along the estimated position of the white line. And a white line candidate point detection area to be set in the white line candidate point detection means by determining the number of white lines to set the small area on the road surface image and the number of small areas assigned to each of the white lines. Setting means, and own vehicle deviation determination means for identifying the own vehicle position in the own lane based on the road parameters, and the white line candidate point detection area setting means, based on the own vehicle position Then, the number of white lines to set the small area and the number of assigned small areas to each white line are changed. When the vehicle approaches one white line, another white line in the approaching direction is changed. Lane area recognizing means for allocating the small area and changing the area to be recognized by the number of white lines detected by the white line candidate point detection area setting means, and the small area on the outer white line when the vehicle approaches the white line. Adjacent lane obstacle detection means for identifying the presence or absence of an obstacle in the adjacent lane area when the area is assigned, and whether or not the lane can be changed without colliding with the obstacle when the obstacle is present. And a lane change propriety judging means for judging.

A traveling road detecting apparatus according to a fourteenth aspect of the present invention has image pickup input means for collecting a road surface image in front of the vehicle, performing image processing to extract a white line on the road surface, and detecting a white line on the road surface image, In a traveling road detecting device for identifying a three-dimensional shape of a road surface and a relative positional relationship of a vehicle to the road surface, a detection parameter holding that holds a plurality of parameters in a parameter description expression obtained by converting the approximate expression of the white line into a coordinate system of a road surface image Means, a plurality of white line candidate points along the white line on the road surface image, white line candidate point detecting means for obtaining position information using the coordinate system of the road surface image, and a plurality of past pasts held in the detection parameter holding means. For the virtual white line candidate points calculated from the parameters, the position information of the latest white line candidate points obtained from the road surface image is compared to obtain the latest values of the plurality of parameters. , A parameter correction means for updating the plurality of parameters of the detection parameter holding means, and a white line shielding range detection means for obtaining a range in which the white line is blocked by a preceding vehicle in its own lane, and the parameter description expression , A plurality of white line models that store a first parameter description formula that approximates a white line with a quadratic curve and a second parameter description formula that is simplified when the white line is shielded to simplify the first parameter description formula. Storage means, select a plurality of parameter description formulas in the plurality of white line model storage means according to the detection state of the white line occlusion range detection means, when the white line is shielded,
A white line model switching unit that causes the parameter correcting unit to execute the process according to the second parameter description expression is provided.

According to a fifteenth aspect of the present invention, there is provided a traveling road detecting apparatus according to the first aspect.
In the traveling road detecting device according to claim 4, the white line shielding range detecting means identifies the preceding vehicle on the own lane from the road surface image to obtain the position of the preceding vehicle, and the vehicle position detecting means based on the obtained vehicle position. And a shielding range estimating means for estimating a range in which the white line is shielded.

The traveling road detecting device according to claim 16 is the device according to claim 1.
In the traveling road detecting device described in 5, the vehicle position detecting means calculates the position of the preceding vehicle by obtaining the side edge coordinates of the preceding vehicle from the road surface image.

The traveling road detecting device according to claim 17 is the device according to claim 1.
In the traveling road detecting device described in 5, the vehicle position detecting means calculates the position of the preceding vehicle by obtaining the lower end coordinates of the preceding vehicle from the road surface image.

The traveling road detecting device according to claim 18 is the one according to claim 1.
In the traveling road detection device according to claim 4, the white line shielding range detection means is capable of detecting a preceding vehicle in a plurality of directivity directions and measuring a distance to an obstacle in each direction, Based on the parameters, the own vehicle traveling lane recognition means for identifying which of the plurality of directivities the preceding vehicle in the own lane corresponds to, and the inter-vehicle distance detection based on the discrimination result of the own vehicle traveling lane recognition means. And an inter-vehicle distance value selecting means for selecting the inter-vehicle distance of the preceding vehicle on the lane from the plurality of inter-vehicle distances obtained by the means.

A traveling road detecting device according to a nineteenth aspect is the one according to the first aspect.
4, 15, 16, 17, or 18, the first parameter descriptive expression is a vehicle position in a traveling lane, a plane curvature of a road, a vehicle inclination with respect to a lane, a road inclination, and a road width. The second parameter descriptive expression includes four parameters associated with each other, while the second parameter descriptive expression includes four parameters except one associated with the plane curvature of the road.

The traveling road detecting device according to claim 20 is the device according to claim 1.
In the traveling road detection device described in 4, 15, 16, 17, 18, or 19, the first parameter description formula is a quadratic curve formula, while the second parameter description formula is a linear linear formula.

According to a twenty-first aspect of the present invention, there is provided a traveling road detecting apparatus according to the first aspect.
4, 15, 16, 17, 18, 19, or 20, the traveling path detection device, wherein the parameter correction means is
A deviation amount calculating means for calculating an amount of deviation between a virtual white line candidate point by the selected parameter description expression and the latest white line candidate point obtained from the road surface image, and fluctuations of the plurality of parameters based on the deviation amount. And a parameter variation amount calculating means for calculating the amount, and the parameter variation amount calculating means calculates the parameter variation amount using a least square method.

A traveling road detecting device according to a twenty-second aspect is the one according to the first aspect.
4, 15, 16, 17, 18, 19, 20, or 21
In the traveling road detection apparatus described above, the parameter correction unit includes a road shape output unit that calculates a three-dimensional shape of the road based on the updated plurality of parameters and immediately outputs the calculated three-dimensional shape.

The traveling road detecting device of claim 23 is the same as that of claim 1.
4, 15, 16, 17, 18, 19, 20, 21, or 22, the vehicle position in the traveling lane, the plane curvature of the road, the inclination of the vehicle with respect to the lane, the inclination of the road, the width of the road. The parameters respectively associated with a, b, c, d, and e, and the coordinate system of the road surface image is x,
When y and i are integers, the first parameter description expression is x = (a + ie) (yd) + b / (yd) + c
On the other hand, the second parameter description expression is x = (a + i
e) (y−d) + c.

[0031]

In the traveling road detecting device according to the first aspect, the "white line candidate point at the previous time by the parameter description expression" and the "white line candidate point at the current time obtained from the road image" are compared in the coordinate system of the road image. The latest values of the plurality of parameters in the parameter description expression are obtained by performing only the calculation operation without performing any coordinate conversion calculation. The white line is approximated as a “plurality of parallel quadratic curves” in a plane and a slope with a constant inclination in the height direction. This approximate formula is converted into a coordinate system of the road surface image and used as a parameter description formula. There is. Therefore, a plurality of parameters can be directly obtained only by a calculation operation using the coordinate system of the road surface image, and it is not necessary to perform the coordinate conversion of the obtained numerical values of the white line candidate points. Details of the setting of the parameters and the parameter description formulas, the calculation method for determining the parameters, the calculation method for deriving the three-dimensional shape of the road from the parameters, and the like will be described in detail in Examples.

In the traveling road detecting device according to the second aspect, the least squares method is used when the variation amounts of the plurality of parameters are obtained from the deviation amounts of the "white line candidate point at the previous time" and the "white line candidate point at the current time". . According to another aspect of the present invention, in the traveling road detection device, the plane curvature of the road ahead, the inclination of the vehicle with respect to the lane, and the inclination of the road, which change every moment as the vehicle progresses, are calculated from the determined latest parameters and immediately output. It By using these numerical values, various functions can be assembled, for example, automatic steering following a curve of a road, risk evaluation combined with detection of a preceding vehicle, warning of drowsiness and carelessness of a driver. In the traveling road detecting device according to claim 4,
The parameter variation is calculated using the parameter description formula x = (a + ke) (yd) + b / (yd) + c.

In the traveling road detecting apparatus according to the fifth aspect, a small area (window) is set in the road surface image to obtain position information of the white line candidate points. A plurality of small areas are set for each white line to be detected, corresponding to the "white line candidate point at the previous time by the parameter description expression". Then, at the time of this setting, "an operation for determining how many small areas are allocated to which white line in the road surface image" is executed. When the vehicle approaches one of the white lanes of the driving lane, the quality of the position information of the white line candidate points obtained from the opposite white line deteriorates (the error of the obtained parameter increases). A small area is also set for the white line to reduce the rate at which the position information of the white line candidate point on the opposite white line is involved in parameter determination.

In the traveling road detecting device according to the sixth aspect, two thresholds are set for the width of the traveling lane obtained from the parameters. When the vehicle approaches one of the white lines exceeding this threshold value, a small area is reassigned from the two white lines sandwiching the traveling lane up to that point to the three white lines centering on the approaching white line. In the traveling road detecting device according to the seventh aspect, two thresholds are set for the width of the traveling lane obtained from the parameters. When the vehicle crosses the white lane and exceeds the threshold value of the adjacent lane, a small area is reassigned from the three white lanes until then to the two white lanes of the new lane. In the traveling road detecting device according to claim 8, the width of the traveling lane is held as a constant, and at least for the setting operation of the small area, even if the parameter related to the width of the traveling lane fluctuates, The arithmetic processing is performed with a constant value from beginning to end. In the traveling road detecting device according to the ninth aspect, the parameter is determined by using the same number of white line candidate points when detecting two white lines or when detecting three white lines. In the traveling road detection device according to the tenth aspect, when three white lines are detected, a larger number of small areas than the white lines on both sides are allocated to the central white line of the road surface image. This is because the white line in the center has higher linearity on the road surface image than the white lines on both sides and is closer from the imaging viewpoint, and therefore the quality of the position information of the white line candidate point is higher.

In the traveling road detecting apparatus according to the eleventh aspect, when the driver tries to cross the white line without taking a lane change direction instruction due to drowsiness or carelessness, the driver is warned accordingly.

According to another aspect of the present invention, the traveling road detecting device detects an obstacle such as a preceding vehicle in an adjacent lane crossing a white line which is about to be crossed, and when it is determined that there is a high possibility of collision, Warn the driver accordingly. The detection of the obstacle in the adjacent lane is started for the adjacent lane beyond the white lane when the vehicle approaches one white lane of the traveling lane. The roadway detection device according to claim 13 does not detect an obstacle in an adjacent lane when a small area is assigned to only two white lanes of the traveling lane, but the vehicle approaches one of the white lanes to detect three obstacles. When a small area is assigned to the white line, the obstacle in the adjacent lane is detected, and if there is an obstacle, the possibility of collision is evaluated.

According to the fourteenth aspect of the present invention, the first parameter description expression is replaced with the second parameter description expression when the white line in front of the vehicle is shielded by the preceding vehicle beyond the limit. This is because if only a short partial white line that has escaped the shielding near the vehicle can be supplemented on the road surface image, there is no benefit to collect the position information of the white line candidate points from the blocked portion.

In the traveling road detecting apparatus according to the fifteenth aspect, the preceding vehicle ahead is detected from the road surface image, the distance to the preceding vehicle is estimated, the white line shielding range is evaluated, and the parameter description formula to be used is determined. .. In the traveling road detecting device according to the sixteenth aspect, the image processing for emphasizing the vertical edge is performed on the road surface image so that the positions of both side ends of the preceding vehicle are highlighted. From this, the coordinate values of both ends of the preceding vehicle are obtained to calculate the position of the preceding vehicle. In the traveling road detecting apparatus according to the seventeenth aspect, image processing for emphasizing the lateral edge is performed on the road surface image so that the lower end (for example, a shaded portion) of the preceding vehicle is highlighted. From this, the lower end coordinates of the preceding vehicle are obtained to calculate the position of the preceding vehicle.

In the traveling road detecting apparatus according to the eighteenth aspect of the invention, instead of the means for obtaining the inter-vehicle distance of the preceding vehicle in the own lane from the road surface image, (or in coexistence with), a beam of radar radio waves, laser light, ultrasonic waves or the like is emitted from the vehicle. An inter-vehicle distance measuring unit that radiates forward and measures the inter-vehicle distance is provided. The plurality of directivity directions are managed by the own vehicle traveling lane recognition means in a format corresponding to the own lane or the adjacent lane recognized from the road surface image. The inter-vehicle distance numerical value selection means supplements the preceding vehicle on the own lane, which is not simply an obstacle in a fixed direction in front of the own vehicle, but is not influenced by the curved shape of the road.

According to the nineteenth aspect of the present invention, in the traveling road detecting device, the parameter description expression in the case where the white line can be sufficiently captured on the road surface image includes five parameters, while the parameter description expression in the case where the shielding is remarkable includes four parameters. It does not include the parameter related to the plane curvature of the road. In the traveling road detecting apparatus according to claim 20, the parameter description expression when the white line can be sufficiently captured on the road surface image is a quadratic curve expression, whereas the parameter description expression when the shielding is remarkable is a linear line. It is an expression. In the traveling road detecting device according to the twenty-first aspect, the least squares method is used when calculating the variation amount of the parameter. In the traveling road detecting device according to the twenty-second aspect, the three-dimensional shape of the road is calculated and output based on the updated latest parameter. In the traveling road detecting apparatus according to claim 23, the parameter description expression when the white line can be sufficiently captured on the road surface image is x =
(A + ie) (y-d) + b / (y-d) + c, whereas the parameter description expression when the shielding is significant is x =
(A + ie) (yd) + c.

[0041]

EXAMPLE A first example will be described with reference to FIGS.
1 is an explanatory diagram of a set coordinate system, FIG. 2 is an explanatory diagram of a white line model in a camera coordinate system, FIG. 3 is an explanatory diagram of corresponding points of new and old white lines in an image coordinate system, FIG. 4 is an explanatory diagram of a configuration, and FIG. 5 is processing. FIG. 6 is an overall flowchart, FIG. 6 is a flowchart for window updating, FIG. 7 is a flowchart for detecting white line candidate points, FIG. 8 is an explanatory diagram for a method for detecting white line candidate points, and FIG. The first embodiment is a traveling road detecting device having a basic function of determining the parameters of the white line model from the road surface image, and is commonly applied to the other embodiments below.

First, the white line model will be described. The white line model is an approximate expression expressing the three-dimensional shape of the white line.
The white line model is coordinate-converted into the coordinate system of the road surface image to create the parameter description formula of the first embodiment. The parameter description expression is a function expression of a white line on the road surface image. The relationship between the coordinate system (x, y) of the road surface image and the camera coordinate system (X, Y, Z) is shown in FIG. In FIG. 1, a white line 25 is drawn by the camera coordinate system 23 whose origin moves on the road 24 together with the vehicle.
Set the approximate expression of. Here, the camera is attached to the vehicle body so that the angle formed by the vehicle traveling direction and the camera optical axis and the angle formed by the camera optical axis and the road surface when the vehicle is stationary are zero. The white line model of the first embodiment is mainly intended for highways, and ignores rotation about the Z axis (roll angle), road slope, and bank angle. Instead of a fixed coordinate system on the ground, a road coordinate system with the momentary vehicle position (camera position) as the origin is used, so even if the road structure is approximated by a simple quadratic formula, it will correspond to the required measurement items. It was possible to secure practically sufficient detection accuracy for multiple road parameters.

The white line model uses road parameters that represent the three-dimensional shape of the road, the vehicle position, and the vehicle attitude, respectively.
The white line 25 is a quadratic equation in the horizontal plane (X-Z) and the vertical plane (X
-Y) is approximated by a linear expression. Road parameters are determined as shown in FIG. 2, and an approximate expression is set as in the following expression (1). In FIG. 2, (a) shows a horizontal plane and (b) shows a vertical plane.

[Equation 1] In FIG. 2A, the white lines are numbered 0, 1, 2, ... The white lines 0 to i are described by the common expression (1). Camera coordinate system (X, Y, Z)
The origin of moves every moment as the vehicle progresses,
The parameters A to E in the equation (1) are changed. Parameter A is the distance (hereinafter referred to as deviation) between the white line located on the left side of the vehicle and the center of the vehicle (attachment position of the imaging device),
Parameter B is the road curvature in front of the vehicle, parameter C is Z
= 0, the yaw angle of the vehicle with respect to the tangential direction of the white line, the parameter D is the pitch angle of the vehicle with respect to the road plane (the relative angle between the road and the Z axis), and the parameter E is the distance between the white lines (lane width for straight roads and Z = 0). Respectively correspond to. Perspective transformation from three-dimensional to two-dimensional is performed on this white line model to create a white line model on the road surface image described in the coordinate system (x, y) of the road surface image, that is, a parameter description expression of the first embodiment. .

As shown in FIG. 1, when an XYZ coordinate system, which is a three-dimensional space, is projected on a plane screen of an xy coordinate system through an optical lens having a focal length f, the XYZ of the three-dimensional space is projected.
The coordinates are converted into x and y coordinates through the perspective transformation of the following equation (2). The structure on the road described in the XYZ coordinate system is projected on the road surface image and becomes an image in the xy coordinate system converted by the equation (2).

[Equation 2] Based on this relationship, the equation (1) is expressed by the following equation (3),
Converted to a white line model on the road image. Here, the defined parameters a to e are parameters newly defined in association with the above-mentioned parameters A to E in order to simplify the description of the expression (3).

[Equation 3] In the first embodiment, after detecting the white line on the road surface image in the xy coordinate system, the parameters a to e of the equation (3) are directly estimated without conversion into the XYZ coordinate system. When the parameters a to e are determined, the road curvature ρ, the yaw angle tan β, the pitch angle t
Each road parameter of an α can be obtained by the following equation (4). If the road model includes parameters corresponding to the vehicle deviation and the lane width, the same effect as that of the first embodiment can be obtained by using a road model other than the equation (3).

[Equation 4]

Next, a method of estimating the parameters a to e will be described. It is assumed that the road structure on the road surface image changes smoothly with respect to the time axis. FIG. 3 shows the movement of the white line portion between the road surface image at the previous time and the road surface image at the current time. Here, the subscript new means the current frame and the subscript old means one frame before. In the first embodiment, frames of the road surface image are captured at intervals of 1/30 second, a plurality of white line portions are extracted from the white line of the road surface image, xy coordinate values are calculated, and parameters are estimated in real time. Parameter estimation is 1
By the method of comparing the white line position of the current frame with the previous white line position obtained from the road surface image before the frame.
As shown in FIG. 3, assuming that the variation amounts from the previously obtained parameters a to e are Δa to Δe, the road surface image (x, y)
Minute variation Δx of the j-th point xij of the i-th white line in
ij is expressed by the following equations (5) and (6) according to Taylor's theorem, if terms of the second or higher degree are ignored.

[Equation 5]

The least squares method is used to estimate the fluctuation amounts Δa to Δe. As an evaluation error function therefor, the following (7)-
Equation (9) is defined.

[Equation 6] Here, the expression (7) is an evaluation error function defined by the difference between the previous detection result xij-1 and the newly detected xij, and pij in the expression (7) is the probability of the white line candidate point. Represent Expression (8) is an evaluation error function expressing the assumption that the parameter smoothly moves in the time axis direction, and S is a weighting coefficient. The sum of the evaluation error functions shown above is represented by the sum etotal of the equation (9), and the parameters are updated as in the following equation (10) by obtaining Δa to Δe that minimizes the sum etotal.

[Equation 7]

Then, the variations Δa to Δe are obtained by solving the following linear simultaneous equations (11). In addition,
In the matrix, each element in the form of XijYijZij is described by omitting the double summation symbol. That is, ΣΣXijYijZ
ij (the first Σ is i and the second Σ is the summation symbol for j).

[Equation 8]

The first embodiment uses the parameter description formula (3) to detect (11) from the detection result of the white line portion at every moment.
The equations are solved to estimate the instantaneous parameters a to e. Figure 4
The functional configuration of the first embodiment is shown in FIG. The image capturing unit 11 captures an input image by capturing a road surface in front of the vehicle. Pre-processing edge detection is performed on the input image to highlight white line features. The white line candidate point detection unit 12 extracts position information of a plurality of white line candidate points described in the coordinate system of the road surface image. Here, as described later, a plurality of small areas (windows) are set at positions where a white line on the road surface image is assumed, and the white line portion in the window is detected. The "positional coordinates in the image coordinate system of" may be obtained by another method. The calculation unit 17 that calculates the point sequence on the previous white line model (curve formula) calculates the virtual position of each white line candidate point on the previous road surface image based on the parameters obtained from the previous input image. The deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence compares the white line candidate point position at the current time by the white line candidate point detection unit 12 with the previous white line candidate point position by the calculation unit 17 to advance the vehicle. The amount of movement of the white line candidate point from the previous time to the current time is calculated.

The parameter fluctuation amount calculation unit 14 solves the above equation (11) from the comparison result of the deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence, and then (7) to (9).
The parameter fluctuation amounts Δa to Δe are estimated by the least squares method of the equation. The correction unit 15 that corrects the previous parameter based on the variation amount substitutes the parameter variation amounts Δa to Δe obtained by the parameter variation amount calculation unit 14 into the equation (10),
The latest parameters a to e (new) are calculated, and the previous parameters a to e (old) of the detection parameter holding unit 16 are replaced. It also calculates and outputs road parameters (constants of road structure such as curvature) from the latest parameters. The detection parameter holding unit 16 holds the latest parameters, and the deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence.
To process the next screen in. With the above operation, the processing for one screen is completed.

A flowchart of the overall processing of the first embodiment is shown in FIG. A detailed flowchart of step 122 in the flowchart of FIG. 5 is shown in FIG.
A detailed flowchart of 16 is shown in FIG. Also,
A detailed flowchart of step 152 in the flowchart of FIG. 7 is shown in FIG. 8 (b). The processing of one screen in the configuration of FIG. 4 is executed in sequence through a loop of steps 116 to 123 in FIG.

In steps 111 to 115 in FIG. 5, initial values for detecting the positional information of the white line candidate points are fetched. These initial values are set by the driver's keyboard operation, but, for example, ROM is stored as constant data corresponding to a state where the vehicle is traveling straight in the center of the lane of a straight road.
You may keep it in. In the first embodiment, when the parameters are calculated based on the detection result of the white line candidate points every moment,
When the white line candidate points are detected (window setting) on the road surface image, the parameter description expression in which the parameters have been substituted is indispensable. Therefore, before the loop of steps 116 to 123 is started, an initial value is given to describe the parameters. Complete the formula. Also, the initial window position and layout are set in correspondence with the white line on the road surface image. That is, step 1
11, parameters a to e, step 112, the number m of white lines to be detected on the road image, step 113, the window initial setting position on the road image, step 114, the weighting constant of the equation (8), step 115. The window setting number n for each white line is set as an initial value.

Step 116 in FIG. 5 corresponds to the white line candidate point detection unit 12. In step 116, a plurality of small areas (windows) are set on the road surface image based on the initial value for the first time and the processing result of the previous screen from the second time. Then, the position information of the white line candidate points is detected for each of the plurality of windows. This position information is composed of the position coordinates x1ij and x2ij of each white line candidate point and the probability Pij. Details of the processing in step 116
This will be described with reference to FIGS. 7 and 9. FIG. 7 is a flowchart of the window setting process, and FIG. 9 shows a white line imaged on the screen. In FIG. 9, (a) shows a road surface image, and (b) shows detection of a white line portion in one window.

In FIG. 9A, a white line 25G on the road 24G is captured in the road surface image 26. The white lines 25G at the left end of the road 24G are numbered as i = 0, 1, 2, ... And four windows 21 are installed for each of the two white lines i = 0, 1. Window 21
Is determined from the parameter description formula, and a plurality of positions are set along the white line on the road surface image. Initial values are assigned to the parameter description formulas for the first time, and parameters obtained from the previous input image are substituted for the second time. The window settings are
The number of pixels (brightness data) involved in detecting the white line candidate point is reduced, and the white line portion can be detected only by clearly determining the contrast between the road surface and the white line. A white line portion is taken out in a window defined at a predetermined height position (y1, y2) on the road surface image. The x coordinate (x1, x2) where the white line crosses the upper side and the lower side of the window is obtained, and the position coordinates (x1, y1) and (x2, y2) of the two white line candidate points are determined.

In FIG. 9B, the position and size of one window 21 are determined by defining the center point (x1Wij, y1ij) at the upper end, the center point (x2Wij, y2ij) at the lower end, and the widths W1ij, W2ij. Determine the size. The height position (y1, y2) of the window is given as a predetermined value, and the horizontal position (x1Wij, x2Wij) is calculated by the parameter description formula. Here, the window height is set to dy = constant. The probability p1ij and p2ij of two white line candidate points obtained from one window 21 are calculated by the following equation (1
Defined by 2).

[Equation 9] Here, the numerical value pi-max is the maximum value of the sum of the density values detected by the window group on the i-th white line.

In step 116 of FIG. 5, white line candidate points are detected by the flow shown in FIG. In FIG. 7, the processing is initialized in step 151, and steps 152 to 15 are executed.
5, through n windows corresponding to m white lines at the same height position y on the road surface image, the x-coordinates x1ij and x2ij of the white line candidate points and the “probability pij of the white line candidate points in the window” are sequentially displayed. To detect. When n windows have been processed, step 157 is passed to shift to detection of a white line candidate point in the window corresponding to the white line located on the right side. In this way, step 15
The processing of 2 to 159 is repeated until m white lines are completed. In the window setting example shown in FIG. 9A, j = 0, 1, 2,
3 in order, followed by 4 windows 2 with i = 1
1 is processed in the order of j = 0, 1, 2, 3.

Details of the process in step 152 of FIG. 7 will be described below with reference to FIG. FIG. 8 is an explanatory diagram of a white line candidate point detection process in one window.
In FIG. 8, (a) is one window, and (b) is a flowchart of white line candidate point detection processing. In the white line candidate point detection processing, data corresponding to the range of the window is called from the memory data of the road surface image captured by the calculation device to perform identification and calculation. At this time, as shown in FIG. 9A, the range of the window 21 is displayed overlaid on the road surface image 26 displayed on the monitor screen. In FIG. 8A, the window 21 has an upper bottom [x =
occupies the coordinate values of x1i (i = 0 to n), y = y1], and the bottom bottom thereof is [x = x2j (j = 0 to m), y = y2 = y1 + d
It is a trapezoidal window that occupies the coordinate values of y]. Both ends of the white line portion cut out by the window 21 are white line candidate points.
Since the height position (y1, y2) of the window is given as a predetermined value, only the horizontal position (x1ij, x2ij) is obtained.

In FIG. 8B, after initializing each value in step 101, a pixel group of line segments connecting one pixel on the upper side and one pixel on the lower side is sequentially called to obtain the density value of the pixel. The sum p is calculated. In the loop 105, “a combination of one pixel on the upper side and m pixels on the lower side is tried (j = 0 to
m), the n pixels on the upper side are covered by the loop 106 (i = 0 to n). When the condition of step 107 is reached, processing of one window is completed. That is, step 1
In 02, the sum p of the density values of the pixels on the straight line connecting the point (x1i, y1) and the point (x2j, y2) is calculated, and in step 103, it is compared with the past maximum value pmax of the density sum p. Step 10
If the condition of 3 is satisfied, pmax, x1 in step 104
, X2 values are updated. In step 109, (x1i, x2j) when the density sum pmax becomes maximum is output as the x coordinate value of the white line candidate point. The density sum pmax obtained at this time is the certainty of the white line candidate points.

Steps 117 to 119 in FIG. 5 correspond to the shift amount calculation unit 13 between the white line candidate points and the white line model point sequence in FIG. In Steps 117 to 119, Step 116
The calculation of Expression (11) is executed from the detection result of the white line candidate points in. Step 120 corresponds to the parameter variation calculation unit 14. In step 120, the latest parameters ae are determined by the equation (10). Step 1
Reference numeral 21 corresponds to the correction unit 15 that corrects the previous parameter based on the variation amount in FIG. In step 121, the old parameters a to e are replaced with the latest parameters. Step 122 is included in the white line candidate point detection unit 12 in FIG. In step 122, the position where the window should be set in the next road surface image is obtained based on the latest parameters. Step 123 is included in the correction unit 15 that corrects the previous parameter based on the variation amount in FIG. In step 123, each value of the road parameter of the expression (6) is obtained from the latest parameter and output.

The window update processing in step 122 is executed according to the flowchart of FIG. 6, in step 131, i = 0 of (a) of FIG. 9,
The height position y at which the window 21 should be set is set for the window 21 of j = 0 for detecting the white line candidate point at the highest position (farthest from the own vehicle) on the road image 26 on the white line 1 of FIG. In step 132, the process is initialized, and through steps 133 to 136, the height position y1ij of the window 21 in FIG. The coordinates x1wij and x2wij are calculated. The window occupies a predetermined width (twice as large as w1ij and W2IJ in FIG. 9B) centering on the x-coordinate x1wij and x2wij on the road surface image. Steps 133 to 137 are repeated to determine the setting position of the window at the same height position y with respect to the m white lines to be detected, and then the window at the next lower position on the road surface image is set. Through steps 133 to 139, n windows are set for each of the m white lines.

In the flowchart of FIG. 5, the detection parameter holding means of the invention corresponds to step 121, the white line candidate point detection means of the invention corresponds to step 116, and the parameter correction means of the invention corresponds to steps 117 to 120.

In the traveling road detecting apparatus according to the first embodiment described above, the window is updated with the latest parameter description formula, and the white line is continuously followed to continuously output the required road parameters. At this time, the road parameters can be directly obtained only by the linear calculation processing in the coordinate system of the road surface image, and the calculation related to the coordinate conversion of the data is not required. It Also, since the approximate expression of the three-dimensional shape of the white line is described in the coordinate system in which the origin moves with the own vehicle, even a simple quadratic curve approximation can secure sufficient practical accuracy for the road parameter estimation result. . Then, according to this approximate expression, road parameters can be obtained with high accuracy for the “portion near the own vehicle” that requires high accuracy for automatic driving and various warnings. Further, since the least squares method is adopted, the practical accuracy of the road parameters can be secured even with a relatively small number of white line candidate points. Therefore, although the least squares matrix operation is included, it is possible to reduce the number of data and not to increase the required number of operations. Further, since the window is allocated to the white line without waste by estimating the position of the white line, the road parameter will not be lost due to lack of data even with a small number of windows.

In the traveling road detecting apparatus of the first embodiment,
Since the parameter is determined by considering the certainty of the white line candidate points, if the white line deviates from the window, the detection result of that window is ignored and the error data by that window does not adversely affect the road parameter estimation result. .. Further, due to the certainty of the white line candidate points, highly accurate data obtained from the clear white line portion on the road surface image close to the own vehicle is heavily used, and the accuracy of the obtained road parameter is increased. Also, even if the white line deviates from the window near the vehicle due to a sudden lane change, etc., the far white line portion will not fall out of the window, so the road model will be gradually corrected based on the detection result of the far white line portion. , The white line part at a distant position is automatically returned to the window, and the white line can be captured again in the window near the host vehicle. Then, the accuracy of the output road parameters continues to improve during the return.

A second embodiment will be described with reference to FIGS. 10 is an explanatory diagram of the configuration, FIG. 11 is a flowchart for switching the number of white line detection lines, FIG. 12 is an explanatory diagram showing an example of road white line detection, and FIG. 13 is an explanatory diagram showing window settings. Here, the window setting is changed according to the position of the own vehicle within the lane width. Note that the setting of the road model, the window processing, the method of estimating the road parameters, etc. are common to the first embodiment, and duplicate explanations are omitted.

In FIG. 10, the image pickup section 41 photographs the scenery in front of the vehicle and collects a road surface image. Image processing unit 42
Extracts the white line representing the driving lane and the characteristics of the preceding vehicle ahead in the captured road surface image. White line candidate point detection unit 43
Sets some small areas (windows) for straight line detection in the screen of the road surface image processed by the image processing unit 42, detects the line segment most likely to be a white line on the road surface in each area, The coordinates of the end points and the certainty of the white line candidate points are obtained. The road parameter estimation unit 44 uses the position information of the white line candidate points detected by the white line candidate point detection unit 43,
Calculate the parameters of the road model. It also calculates and outputs road parameters that describe the three-dimensional shape of the road.
The host vehicle deviation determination unit 45 determines which position in the travel lane the host vehicle is traveling from the parameter calculated by the road parameter estimation unit 44 and a known constant. The white line detection number setting unit 46 sets the number of white lines to be detected in the white line detection on the next screen from the traveling position of the own vehicle detected by the own vehicle deviation determination unit 45. The white line candidate point detection area setting unit 47 includes the road parameters calculated by the road parameter estimation unit 44 and the white line detection number setting unit 4.
From the number of white line detection lines set by 6, the number and position of small areas (windows) for detecting white line candidate points on the next screen are set.

FIG. 11 shows a flowchart for switching the number of white line detection lines. Similar to the case of the first embodiment, each function of FIG. 10 is sequentially executed by time-sharing one arithmetic unit. Hereinafter, description will be given according to the flowchart of FIG.

FIG. 12 (a) shows two white lines 25A, 25 representing the traveling lane of the own vehicle from the road surface image of the forward traveling scenery.
This is an example in which B is detected. Step 401 is shown in FIG.
Corresponding to the image pickup unit 41. In step 401, the CCD
An input image taken by an imaging device such as a camera is taken into the memory of the arithmetic device. Step 402 is shown in FIG.
Of the image processing unit 42. In step 402, for example, "sobel operator" is applied to the input image data.
Image processing for vertical / horizontal edge detection is performed by such as, and processing for emphasizing the features of the white line and the preceding vehicle is performed. Step 403 is included in the white line candidate point detection area setting unit 47 in FIG. In step 403, based on the processing result on the previous screen, as shown in (a) or (b) of FIG.
A small area for white line detection (hereinafter referred to as windows 21A and 21B) is set. Steps 404 and 405 correspond to the white line candidate point detection unit 43 in FIG. In step 404, straight line detection is executed for each of the windows 21A and 21B, and in step 405, the coordinates of the end points of the line segment that are most likely to be a white line (x1ij and x2ij in FIG. 2) and their certainty as a white line (Pmax in FIG. 8). Detect and.

The straight line detection in each window in step 404 is executed by the same procedure as in the case of the first embodiment described with reference to FIG. As shown in (a) of FIG.
A trapezoid is set as the shape of the window 21, and for all line segments formed by connecting one point on the upper bottom to one point on the lower bottom,
By the procedure shown in FIG. 8B, the sum of the density values in the edge image of the pixels on the line segment is calculated. Then, the line segment having the largest sum is set as the detection straight line in the window. Here, the position information of the white line candidate points can be obtained even if the line segment having the smallest sum of the density values is used as the detection straight line. Then, in the next step 405, the sum of the coordinates of the end points of this line segment and the density value (Pmax in FIG. 8) is held as the detection result. The sum of the density values serves as a parameter of certainty as a white line.

Steps 406 and 407 correspond to the road parameter estimating section 44 of FIG. When the detection results have been obtained for all the windows 21A and 21B in FIG. 12A, in step 406, the fluctuation amounts Δa to Δe of the parameters a to e of the road model are calculated by the least square method based on these. . Then, in step 407, the parameter a is calculated according to the calculated fluctuation amounts Δa to Δe.
~ E are updated as in the above equation (10). In this way, by updating the road parameter, the road white line can be detected following. Further, as shown in FIG. 2, since the parameters to be updated are common to all the detected white lines, the processing speed does not change unless the number of white line candidate points changes even if the white line detection number m increases. .

At this time, pay attention to the values of the parameters a and e. As described above, the parameter a is associated with the distance from the left end of the running lane to the host vehicle, and the parameter e is associated with the lane width. Indicates whether you are driving. If the vehicle is traveling in the center of the lane, (a) in FIG.
The two white lines 25A, 25B are detected as described above, but if the host vehicle is traveling at the edge of the lane, the three white lines 25A, 25B, 25C are detected as shown in FIG. 12B. Specifically, in step 408, the current white line detection number m is determined. In the case of two lines, in step 409, thresholds th1 and th2 as in the following equation (13) are set, and this threshold value is set. When it exceeds, in step 410, the white line detection number m is changed to 3 as shown in FIG. Step 40
9, 410 are included in the own vehicle deviation determination unit 45 and the white line detection number setting unit 46 of FIG.

[Equation 10] Here, the thresholds th1 and th2 are limited as follows.
If this value is exceeded, the imaging position will cross the lane, and a part of the vehicle will switch to three-line detection after entering the adjacent lane. As described above, by using the threshold value larger than 0 and smaller than e, the detection of three vehicles is switched before a part of the own vehicle enters the adjacent lane,
It is possible to recognize two regions, the own vehicle traveling lane and the traveling lane after the lane change.

Steps 411 and 412 are included in the own vehicle deviation determining section 45 and the white line detection number setting section 46 of FIG. When the number of white line detections m is 3 as shown in FIG. 12B, thresholds th3 and th4 as shown in the following equation (14) are set in step 411, and when the thresholds are exceeded, In step 412, the white line detection number m is set to 2 as shown in FIG.

[Equation 11] Here, the reason that the limits are set for th3 and th4 is that the own vehicle has completed the lane change at least when this value is exceeded. With the above settings, when the own vehicle tries to change the lane, the lane area of the change destination will be recognized when the own vehicle approaches one side of the driving lane.
For example, when a front obstacle is detected, there is no delay in determining the risk level. Step 413 is
This corresponds to the white line candidate point detection area setting unit 47 in FIG. Finally, in step 413, the number, position, and size of windows for the next screen are determined based on the detection result of this time, and the window settings are updated.

Next, a method of setting the size and position of the window will be described. 13A shows the number of white lines detected m
It is explanatory drawing of the window setting method when there are two. Since the quantization error due to pixels becomes large near the vanishing point 50 on the screen of the road surface image, the window 51 is set from a region several pixels (for example, 6 pixels) below the y coordinate of the vanishing point obtained from the road model. Further, as the window 51 is set at a position away from the vanishing point, the length of the window 51 in the y direction is set longer. This is because the lower the screen is, the closer the image is to the own vehicle, and the clearer the image is. As described above, in the straight line detection in the window 51, the sum of the densities of the pixels of the edge image on the line segment is used as the certainty of the white line candidate point at the end point. Therefore, if the window 51 is made longer, the line segment also becomes longer. The probability of white line candidate points is greater when detected in the long window 51 than when detected in the short window 51. Therefore, in order to increase the certainty of the white line candidate points in the clearly reflected portion, such window setting is performed.

Similarly, FIG. 13B is an explanatory diagram of the window setting when the number of white line detection lines m is three. For the same reason as in the case of FIG. 13, a window is not set near the vanishing point 50 on the screen of the road surface image, but is set from several pixels below. When three white lines are to be detected, the two white lines on both sides are close to horizontal, and the white line at a point far from the own vehicle is detected. Therefore, the length of the window 52 in the y direction is long. Shorten. On the contrary, since the center line is the white line near the host vehicle, the y-direction of the window 53 is set to be long and the weight of the white line candidate point of the center white line is increased. By setting the sizes and positions of the windows 52 and 53 as described above, a more reliable road shape estimation can be performed.

Next, the set number of windows will be described. As described above, the road model parameters used are those that are common to all detected white lines, such as road curvature, lane width, and the posture of the host vehicle. , White line depends on the number of candidate points. Therefore, even if the number of white line detections increases, the processing speed is not affected unless the number of windows set in one screen changes significantly. Comparing (a) and (b) of FIG. 13, the number of windows is 6 per white line on the screen of (a), whereas it is 4 per white line on the screen of (b). . Therefore, the number of white lines to be detected is 3/2 times, but the total number of windows per screen has not changed, so that the processing times of both are equal.

In the flowchart of FIG. 11, the image capturing input means of the invention is steps 401 to 402, the road parameter estimating means is steps 406 to 407, the white line candidate point detecting means of the invention is steps 403 to 405, and the white line candidates of the invention. The point detection area setting means corresponds to step 403, the vehicle deviation determination means of the invention corresponds to steps 409 and 411, and the white line candidate point detection area setting means of the invention corresponds to step 413. In the traveling road detection device of the second embodiment described above, the number of white line detection lines is switched according to the position of the vehicle within the lane width, and in the case of three, the window setting with emphasis on the central white line is performed. The “inaccurate detection data” obtained from the window set on the white line on the far side does not adversely affect the road parameter estimation result. As a result, the accuracy of the output road parameter is maintained high. Further, by setting the window for each white line so that the number of windows in the screen does not change even if the number of white lines detected increases, the processing speed of the arithmetic unit is not affected.

A third embodiment will be described with reference to FIGS. FIG. 14 is a block diagram showing the configuration of the third embodiment, FIG. 15 is an operation of the third embodiment, and FIG. 16 is an explanatory diagram of a lane departure warning. Here, two alarm functions are added to the function equivalent to that of the second embodiment. One is a warning for the driver's carelessness when changing lanes, and the other is a warning for the possibility of a collision with an obstacle ahead of an adjacent lane to change lanes.

In FIG. 14, the image input unit 61 takes in a road surface image taken by a CCD camera (not shown). The edge extraction unit 62 converts the road surface image into an edge screen to highlight the white line and the characteristics of the preceding vehicle. The straight line detection area setting unit 63 sets a plurality of windows on the edge screen based on the road parameters and the number of white line detection lines set based on the processing result of the previous screen. The white line candidate point detection unit 64 executes straight line detection for each window, and holds the coordinate values of a plurality of white line candidate points and the parameter of the certainty. The road parameter calculation unit 65 updates the parameters of the road model from the data of the white line candidate points, and the updated parameters are immediately sent to the straight line detection region setting unit 63, the lane change determination unit 66, and the lane region recognition unit 69. It

The lane change judging section 66 judges whether or not the own vehicle is in the lane change operation by using the road parameters corresponding to the lane width and those corresponding to the deviation of the own vehicle. To do. At this time, the direction indicator determining unit 73 determines whether or not the driver is operating the direction indicator, and when the host vehicle is trying to cross the white line without displaying the direction indicator, the alarm generator 75 It activates and outputs an alarm to the driver.

Further, according to the judgment result of the lane change judging section 66, the white line detection number setting section 67 sets the number of white lines detected on the next screen and detects the set value by a straight line, as in the case of the second embodiment. It is sent to the area setting unit 63. Further, according to the set value of the number of white lines detected in the white line detection line number setting unit 67, the recognized lane number determination unit 68 determines the number of lane regions that the lane region recognition unit 69 should recognize in the lane recognition on the next screen. .
When the number of white lines detected is two, the lane to be recognized is only one of the own lanes, but when the number of white lines detected is three, the lanes to be recognized are the own lane and the adjacent lane. There are two. This determination result is sent to the lane area recognition unit 69.

The lane area recognizing section 69 recognizes the own lane on the screen and the area of the adjacent lane set when the number of recognized lanes is two on the edge screen. For each area, the obstacle detection unit 70 on the own lane
However, for adjacent lanes, the obstacle detection unit 71 on the adjacent lane detects an obstacle such as a preceding vehicle. When an obstacle is detected by the on-lane obstacle detection unit 70, the collision possibility determination unit 74 determines whether the own vehicle collides with the detected obstacle and determines that there is a high possibility of collision. In this case, the alarm generator 75 outputs an alarm to the driver.

On the other hand, when the host vehicle approaches a white line on the adjacent lane side, the number of white lines detected becomes three, and the obstacle detection unit 71 on the adjacent lane detects an obstacle ahead of the adjacent lane to be changed lanes. It will be detected. When an obstacle such as a preceding vehicle is detected in an adjacent lane, the lane change permission / inhibition determination unit 72 considers the inter-vehicle distance, the relative speed of the own vehicle and the preceding vehicle, and the like, and the own vehicle collides with the detected obstacle. Decide whether to do it. Then, when it is determined that the possibility of collision is high, the alarm generation unit 75 is caused to output an alarm. In the alarm generation unit 75, there are three states to be alarmed: (1) lane change without direction indication, (2) increased possibility of collision with preceding vehicle in own lane, and (3) lane change impossible. The content of the alarm is changed depending on which one corresponds to, and the driver can instantly determine in which state the alarm is generated.

The imaging input means of the invention corresponds to the image input section 61 and the edge extraction section 62, and the parameter estimation means of the invention corresponds to the road parameter calculation section 65. The white line candidate point detection means of the invention is a white line candidate point detection section 64, and the white line candidate point detection area setting means of the invention is a white line detection number setting section 6.
7, the straight line detection area setting unit, and the own vehicle deviation determination means of the invention correspond to the lane change determination unit 66. The direction instruction determining unit of the invention corresponds to the direction instruction determining unit 73, and the alarm generating unit of the invention corresponds to the alarm generating unit 75. The lane area recognition unit of the invention corresponds to the lane area recognition unit 69, the own lane obstacle detection unit of the invention corresponds to the on-lane obstacle detection unit 70, and the collision possibility judgment unit of the invention corresponds to the collision possibility judgment unit 74. The adjacent lane obstacle detection unit of the invention corresponds to the adjacent lane obstacle detection unit 71, and the lane change availability determination unit of the invention corresponds to the lane change availability determination unit 72.

With the above configuration, in the third embodiment, it is possible to realize a lane departure warning that does not depend on the driver's intention and a front obstacle warning that also corresponds to a lane change. FIG. 15 (a) is a front view when the host vehicle is traveling in substantially the center of the travel lane L1, and two white lines representing the travel lane L1 of the host vehicle are detection targets. Here, when only the traveling lane L1 region is recognized, it can be recognized that the vehicle 76 is a vehicle traveling in front of the traveling lane L1. However, although it is understood that the other vehicle 77 is not in the traveling lane L1, it is traveling in the adjacent lane L2 as shown in FIG. 15B, or as shown in FIG. 15C. Further, it cannot be determined whether the vehicle is traveling in the outer lane L3. However, in the third embodiment, when an attempt is made to change the lane to the adjacent traveling lane L2, the adjacent lane L2 can be recognized when the own vehicle approaches the right white line of the traveling lane L1 to some extent. Vehicle 77 is next to lane L before part of it enters the next lane L2.
It is possible to judge whether or not it exists in 2. Therefore, the risk of lane change or lane departure can be determined early.

FIG. 16 is an explanatory diagram of a lane departure warning.
Here, the state of the road surface image when the host vehicle is about to depart from the lane is shown. When a driver's abnormality (for example, dozing etc.) is detected and the driver is about to depart from the lane, an alarm is issued. 16A shows a state where the vehicle is traveling in the center of the traveling lane, and FIG. 16B shows a state where the vehicle is protruding from the left end of the traveling vehicle. If it is determined that the host vehicle has moved from the state shown in (a) as indicated by the arrow and has moved to the state shown in (b), and the traveling position of the host vehicle has approached the white line beyond a predetermined limit, then automatically. An alarm is output to. Regarding the white line on the left side, depending on the parameter a of the white line model, the lane width W, and the camera mounting position (height H0, center of vehicle body), (W / 2) + aH0 in the following equation (15)>
When the relationship of 0 is established, it is judged that the vehicle has deviated from the lane.
As a result, it is possible to prevent an accident from occurring due to the driver's carelessness or dozing.

[Equation 12]

FIG. 17 is an explanatory diagram of a modification of the third embodiment. Here, the image processing system and the multi-beam type distance sensor are combined to configure the on-lane obstacle detection unit 70 and the adjacent-lane obstacle detection unit 71 in FIG. A plurality of inter-vehicle distances measured by the multi-beam are made to correspond to a plurality of vehicle positions on the road surface image detected by the image processing system.

In a vehicle having a rear-end collision warning device equipped with an image processing system and a multi-beam type distance sensor such as a laser radar, when the inter-vehicle distance to the preceding vehicle is measured when the lane of the own vehicle is changed, the lane of the own vehicle is measured. Since the yaw angle with respect to the vehicle changes, the preceding vehicle in the own lane may not exist in front of the front of the own vehicle, and the distance measured by either the left or right beam may be the inter-vehicle distance to the preceding vehicle. At this time, the vehicle 76 also appears in the lanes L1 and L3 adjacent to the distance measuring range (area) A1 measured as shown in FIG.
When 78 is present, it is impossible to know which vehicle is detected by the distance measurement data. In such a case, if the light receiving unit of the distance measuring sensor and the image pickup device are fixed to the vehicle, as shown in FIG. 17B, the distance measuring range A of the distance measuring sensor is displayed on the screen.
1, A2 and A3 can be limited respectively. Then, based on this limited relationship, when the own vehicle tries to change the lane from the own lane to the adjacent lane, the own lane on the input screen and the adjacent lane to which the lane is changed are recognized at the same time and the input screen is displayed. In which distance measuring range the preceding vehicle in the adjacent lane belongs is identified. Then, the required distance measurement data of the preceding vehicle in the adjacent lane is automatically selected from the plurality of distance measurement data obtained by the multi-beam type distance sensor.

In the case of FIG. 17, when the lane is changed, the distance measuring range A2
The distance measurement data of the preceding vehicle 77 is selected, and the same lane change determination and warning processing as in the third embodiment are executed. According to the modification of the third embodiment, the inter-vehicle distance can be measured much more accurately as compared with the case where the inter-vehicle distance is calculated from the road surface image, and the reliability of the alarm is improved. Further, since the program for calculating the inter-vehicle distance from the road image can be omitted, the calculation load of the calculation device can be reduced.

By the way, FIG. 38 is an explanatory diagram of a white line approximation state according to the first embodiment. (A) is a road surface image, (b)
In the equation (3) of the white line model, is a diagram illustrating a difference δx at each y coordinate when the parameter b that defines a curve or a straight line is about 300R and when the parameter is a straight line, that is, when b = 0. . In FIG. 38 (a), in the first embodiment, the parameter variation amount that minimizes the difference between the white line model (curve formula) described by a plurality of parameters on the image coordinate system and the input white line image. By estimating
Continue to estimate the road structure while updating the model (following the white line). However, if the distance between the preceding lane and the preceding vehicle in the own lane is shortened and the visible range of the white line is narrowed, it is not possible to capture the white line in the distant portion shielded by the preceding vehicle. Then, with only the data of the white line candidate points within the visible range of the white line, even if the parameter estimation is executed using the same parameter description formula of the above-mentioned formula (3), whether the road currently in progress is a straight line or a curved line. It becomes difficult to identify if there is any. Then, as shown in (b) of FIG. 38, the curvature estimation result becomes extremely unreliable, and since the estimation result cannot be adopted, the time spent for curvature estimation is wasted.

Therefore, in the following fourth embodiment, the white line model is switched in response to the above. 18 to 2
The fourth embodiment will be described using 0. FIG. 18 shows the overall construction of the fourth embodiment, FIG. 19 is a flowchart of the overall processing, and FIG. 20 is a flowchart of window updating.
Here, when a part for switching and using a plurality of white line models is added to the configuration of the first embodiment shown in FIG. 1 and the white line is largely shielded by the preceding vehicle, the white line model of the first embodiment is used. Road parameters are estimated using a simplified white line model. The same parts as those of the first embodiment are designated by the same reference numerals and detailed description thereof is omitted.

In FIG. 18, the image pickup section 11 picks up an input image by picking up an image of the road surface in front of the vehicle. Vertical edge detection is executed as pre-processing on the input image to highlight the white line and the characteristics of the preceding vehicle. White line candidate point detection unit 1
2 detects the white line candidate point data described in the coordinate system of the road surface image for each of the plurality of windows set at the assumed white line positions on the road surface image. The calculation unit 17 that calculates the point sequence on the previous white line model (curve formula) determines the assumed white line position based on the previous parameters. The deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence calculates the latest position information of the white line candidate points by the white line candidate point detection unit 12 and the point sequence on the previous white line model (curve formula). Part 17
The calculated amount of deviation of the white line candidate points is calculated by comparing with the assumed white line position. The parameter movement amount calculation unit 14 estimates the parameter fluctuation amount of the white line model by the least square method from the result of comparison by the deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence. The correction unit 15 that corrects the previous parameter based on the variation amount calculates the latest parameter using the parameter variation amount calculated by the parameter variation amount calculation unit 14 and updates the previous parameter. The detection parameter holding unit 16 holds the parameter updated by the correction unit 15 that corrects the previous parameter based on the variation amount until the processing of the next screen.

The white line shielding position detecting section 18 extracts the characteristic of the preceding vehicle from the image-processed input image and measures the "range in which the preceding vehicle shields the white line on the road surface image". The plurality of white line model storage units 19 hold “a plurality of parameter description expressions” selected according to the range in which the white line is shielded by the preceding vehicle. The white line model switching unit 20 determines, according to the white line occlusion range obtained by the white line occlusion position detection unit 18,
An appropriate one is selected from the “plurality of parameter description expressions” stored in the plurality of white line model storage units 19 and sent to the deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence. The deviation amount calculation unit 13 between the white line candidate points and the white line model point sequence executes the above calculation using the parameter description formula selected by the white line model switching unit 20.

In FIG. 19, the initial values of the parameters a to e are set in step 111, and the number m of white lines to be detected on the road surface image is set in step 112. Through steps 113 to 115, the parameters a to e are substituted into the parameter description expressions to estimate the white line positions on the road surface image, and white line detection windows for the first time are set for m white lines on the road surface image. In step 116, the position information of each white line candidate point is detected according to the white line detection program shown in FIG. In step 117B, from the detection result of the white line candidate points, expression (11) or (1) will be described later.
The calculation is performed by the matrix of the equation (8), and the latest parameters a to e of the equation (10) are calculated from the calculation results of Δa to Δe.
To confirm. In step 121, the old parameter a ~
Replace e with the latest parameters.

The step 127 inserted between the step 121 and the step 122 corresponds to the white line occlusion position detection section 18, the plurality of white line model storage sections 19, and the white line model switching section 20 of FIG. In step 127,
“Curve / straight line model selection processing of detecting a white line shielding range and selecting one from two types of parameter description expressions” is executed. In this processing, the position of the preceding vehicle is obtained from the road surface image and the two types of edge-enhanced images of the road surface image, and when the white line is not almost shielded by the preceding vehicle, a quadratic curve similar to that of the first embodiment The parameter description formula of the formula is selected, but when the white line is largely shielded by the preceding vehicle, the parameter description formula of the linear linear formula is selected instead of the parameter description formula of the quadratic curve formula. The parameter description expression selected here is used for calculation of window setting and parameter estimation until the next change. This processing will be described later in detail with reference to FIGS. 23 and 24. In step 122, the position where the window should be set on the next road surface image is obtained. In step 123,
Values of road parameters describing the three-dimensional shape of the road and the vehicle attitude are obtained from the latest parameters and output. In step 124, the next road surface image and its two types of edge enhancement screens are captured. In step 125, the detection of the preceding vehicle is executed by the vertical and horizontal edge detection.

FIG. 20 shows details of the processing in step 122. In step 131B of FIG. 20, the white line candidate point at the highest position (the position farthest from the own vehicle) on the road surface image in the white line of i = 0, 1 that sandwiches the own lane is detected j
For a window of = 0, the height position y at which the window should be set is set. When the white line is shielded by the preceding vehicle, a height yie lower than the shield position on the screen is selected, but when it is not shielded, the same as in the first embodiment. After the process is initialized in step 132,
In steps 133 to 136B, two height positions y on the road surface image are selected using one selected parameter description expression.
X-coordinates x1wij, x2 at which the white line should pass at 1ij, y2ij
wij is calculated. Steps 133 to 137 are repeated to determine the setting position of the window at the same height position y with respect to the m white lines to be detected, and then, passing through step 137, the setting of the window at the next lower position on the road surface image is performed. To be executed. Through steps 133-139,
n windows are set for each of the m white lines.

In the flow chart of FIG. 19, the imaging input means of the invention is steps 124 to 125, the detection parameter holding means of the invention is step 121, the white line candidate point detection means of the invention is step 116, and the parameter correction means of the invention is Each corresponds to step 117B. The white line shielding position detection means, the plurality of white line model storage means, and the white line model switching means of the invention are included in the processing of step 127. According to the fourth embodiment described above, when the white line to be detected is blocked by the preceding vehicle,
Since the window is set while avoiding the shielded area, there is no need to set the window on the image of the preceding vehicle to detect erroneous data. There is no fear of adversely affecting the parameter estimation result by executing straight line detection unrelated to the white line in the window. Further, since the data of white line candidate points is stored only in the proximity portion where the white line is apparent on the road surface image, the parameters of the road shape and the vehicle attitude can be accurately estimated for the proximity portion. Then, when the white line is shielded, the calculation load such as the least squares method is reduced by adopting a simple parameter description expression, so that the processing such as the position detection of the preceding vehicle is added. A high processing speed comparable to that of the first embodiment can be realized.

A fifth embodiment will be described with reference to FIGS. 21 to 29. FIG. 21 is an explanatory diagram of a basic configuration for selecting a white line model, FIG. 22 is an explanatory diagram of a selection model, FIGS. 23 and 24 are flowcharts for model selection, and FIGS. 25 and 26 are vehicle candidate detection (lower shadow detection). ) FIG. 27 is an explanatory diagram of a vehicle candidate detection (vehicle width detection) method, FIG. 28 is an explanatory diagram of a model selection threshold value determination method, and FIG. 29 is a selection model (only in the case of straight line detection). FIG. Here, the procedure of estimating the position of the preceding vehicle from the original image of the road surface image and the two types of edge enhancement screens will be described in detail.

When the white line in front of the host vehicle is blocked on a motorway such as an expressway, it is almost always caused by the preceding vehicle. Therefore, in the fifth embodiment, (1) after the own vehicle traveling lane is recognized by the method of the first embodiment, the vehicle detection is performed so that the white line blocking start coordinates (yie, i = 0, based on the vehicle position). Estimate 1). (2) The white line model is switched according to the estimated coordinates. The idea is adopted. In FIG. 21, the components denoted by the numbers 11 to 20 are the same as in the case of the fourth embodiment of FIG. 18, so detailed description will be omitted. The vehicle detection unit 10 determines the preceding vehicle position from the road surface image, and adjusts the window setting state in the white line blocking position detection unit.

23 and 24 are flowcharts of the fifth embodiment. Step 127 of the flowchart in FIG.
23 shows the details of the process in FIG. 23, and FIG. 24 shows the process between B1 and B2 in the flowchart of FIG. After updating the model parameters in step 121 of the flowchart of FIG. 19, step 60 of the flowchart of FIG.
Move to 1. After the initial value y = y0 is set in step 601, in step 602, using the horizontal edge image memory data in which the horizontal edge image shown in FIG. 26 is stored, the line segment x1 (y ) -X0
The edge intensity distribution of (y) is created. This process is performed in the loop of steps 601 to 603, 609, and 610, and the initial value y
= Y0, the sequence is sequentially moved downward and repeated. That is, in the edge intensity distribution created in step 602, it is checked in step 603 whether the number nh of pixels exceeding the threshold value Ih-t exceeds the real constant multiple nk (<1) of the edge intensity distribution creation range x1 -x0. To judge. If it exceeds, the y coordinate value is set to the vehicle lower shadow candidate coordinate ys in step 604.
And then proceed to the next. If not, step 6
The above operations are repeated until y = ymin through 09-610. As a result, the lower shadow candidate coordinates are searched (vehicle determination condition 1). In the range up to the height ymin on the road surface image, which corresponds to a position sufficiently far from the host vehicle, the lowest lateral edge that is longer than a certain proportion of the lane width is regarded as the lower shadow. The memory device attached to the arithmetic unit temporarily stores pixel data of the original image of the frame, the vertical edge-enhanced image of the original image, and the horizontal edge-enhanced image of the original image. The original image stored in the memory element,
The pixel data groups of the vertical edge-emphasized image and the horizontal edge-emphasized image are referred to as an original image memory, a vertical edge-emphasized image memory, and a lateral edge-emphasized image memory, respectively.

Next, in steps 605 to 606, FIG.
Based on the original image memory storing the original image shown in 5, y =
Line segment x1 (ys-δh) -x0 (ys-δ at ys-Δh
h) Obtain the minimum luminance value dmin above. When the dmin is a pixel darker than the threshold value d-t, ys is recognized as the y coordinate of the shadow of the preceding vehicle on the vehicle traveling lane (vehicle determination condition 2). Since the lower shadow has a thickness in the y direction, the line segment of the lower boundary of the lower shadow in the lateral edge image is identified by this processing.

Finally, in steps 607 to 608, as shown in FIG.
Using the vertical edge image memory storing the vertical edge image shown in FIG. 7, the edge intensity distribution on the line segment x1 (y) -x0 (y) at the height y = ys-δv on the vertical edge image is created. In the intensity distribution, the threshold value Iv-t is applied from both left and right sides.
When the edge strength is compared with, the x coordinate exceeding the threshold value is first detected as xie (i = 0, 1). A line segment x at the difference ΔXc between the two obtained x coordinates and the height ys
From the length ΔX of 1 (ys) -x0 (ys) and the road width E, the vehicle width equivalent amount W is estimated by the following equation (16).

[Equation 13] Since the width of the car can be specified within a certain range, step 6
08, the condition W-t0 <W <Wt1 is set to the vehicle determination condition 3
As a final decision, whether or not there is a vehicle ahead is performed. If it is determined that a vehicle exists, step 62
Go to 1.

At step 621, the process is initialized,
Starting from y = ys in step 622, the height y0k that satisfies the condition of step 625 is obtained for the white line with i = 0, and this value is held as y0e in step 626. Next, the same processing is performed for the white line of i = 1 and the result is held as y1e. When the processing of the two white lines with i = 0 and 1 is completed, the process exits from step 629 and proceeds to step 631.
Determine whether to switch the white line model with. That is, in step 623, xi on the white line is obtained from the parameter description expression, and in step 624, the difference d between the already detected vehicle edge position xie (i = 0, 1) and xi (y).
Find xi. Then, in step 625, the difference d
When xi becomes smaller than the threshold value dx-t, step 6
At 26, the y coordinate is stored as yie (i = 0, 1). This processing is schematically shown in FIG. Also,
The determination of model switching is made using the numerical value yie and the map shown in FIG. In FIG. 28A, y
Starting from = ys, the distance dx is obtained at successively higher positions y. The distance dx is calculated from xik (i = 0, 1) obtained by substituting y into the parameter description expression and vehicle edge xie (i = 0, 1).
Is the distance to. The white line visible distance corresponds to the height position yie (i = 0, 1) where the distance dx falls below the threshold value δx-t.

As shown in FIG. 38 (a), the curve shown in FIG. 28 (b) has a parameter b for defining a curve or a straight line in formula (3) of the white line model of about 300.
The difference δx is plotted at each y coordinate when R and when a straight line, that is, when b = 0. From this map, for example, y-t at which the difference starts to become smaller than δx-t may be used as the criterion for model switching. I.e. yie
> Y-t is used as the criterion for switching the curve-linear model.
When the following equation (17) is used as the straight line model, the equation (10) of the above curve model is changed to the following equation (18) to reduce the calculation load.

[Equation 14]

That is, in the fifth embodiment, as shown in FIG. 22 (a), the preceding vehicle is not present or is distant,
When the numerical value ys is smaller than the threshold value y-t (at a higher position), the same quadratic curve parameter description expression as that in the first embodiment is selected, but shown in FIG. As described above, when the preceding vehicle is close and the numerical value ys is larger than the threshold value y-t (at the lower position), the linear equation (17) is selected.

Further, in the case where the vehicle-to-vehicle distance becomes shorter, the white line candidate points need not be detected by a plurality of windows. In this case, as shown in FIG. 29, the two straight lines detected in the left and right windows are directly recognized as white lines. The linear equation of the white line with i = 0 is x0
= A1 y + b1, the linear linear expression of the white line of i = 1 is x1 = a
2 y + b2, if the coefficients a1, b1, a2, b2 are directly obtained by the window calculation processing, the yaw angle and pitch angle indicating the vehicle attitude are the values of c and d calculated from the following equation (19). Therefore, it can be obtained by the equation (4).

[Equation 15]

A sixth embodiment will be described with reference to FIG. FIG. 30 is an explanatory diagram of a basic configuration for selecting a white line model. Here, the white line shielding range is measured using a multi-beam type radar inter-vehicle distance sensor.

The lower shadow of the vehicle is clearly projected on the road surface in fine weather, but it may not appear in cloudy weather or at night. In such a case, the white line model switching judgment as in the fourth and fifth embodiments can be performed by using a means such as a radar that recognizes the inter-vehicle distance without depending on the external light. Of the components shown in the configuration of FIG. 30, components having the same functions as those of the components of the fourth embodiment are designated by the same reference numerals as those in FIG. 18, and detailed description thereof will be omitted. .

The running lane recognition unit 81, which is newly added to the configuration of the fourth embodiment, distinguishes and recognizes the running lane on the road surface image based on the calculated parameters. The inter-vehicle distance detecting unit 82 uses a radar beam emitted in front of the vehicle to measure the distance to an obstacle (leading vehicle) in front with a plurality of directivities. The radar distance value selection unit 84 selects one of the distance values measured by the inter-vehicle distance detection unit 82 with a plurality of directivities, based on the recognition result by the traveling lane recognition unit 81. The vehicle y coordinate value estimation unit 83 in the image coordinate system converts the distance value L selected by the radar distance value selection unit 84 into the y coordinate (ys) of the image coordinate system using the following equation (20). The white line model switching unit 20B
Unlike the case of the fourth embodiment, this coordinate-converted coordinate value ys is directly used for the model switching determination.

[Equation 16] The white line candidate point detection unit 12B extracts the position information of the white line candidate points through white line detection in the window and calculation processing of the detected data. The reason why the output of the white line model switching unit 20B is shown between the white line candidate point detection unit 12 and the imaging unit 11 is as follows.
This is because the white line candidate point detection unit 12 performs window setting using the switched white line model. According to the seventh embodiment, the white line shielding range can be accurately estimated and the white line model can be switched by recognizing the inter-vehicle distance even in the case where the shadow of the vehicle does not appear clearly in cloudy weather or at night.

The seventh embodiment will be described with reference to FIGS. 31 is an explanatory diagram of the vehicle detection result of the seventh embodiment, FIG. 32 is an explanatory diagram of the laser beam arrival position vector, and FIG.
Is a flowchart for selecting a radar distance value. In a conventional rear-end collision warning device using a multi-beam type laser radar device, when the distance value of the reflector existing on the outside roadside zone on a curved road is erroneously recognized as the distance value of the preceding vehicle and a false alarm is issued. was there. Therefore, in the seventh embodiment, the reliability of the warning of the rear-end collision warning device is improved by separately detecting the preceding vehicles on the plurality of lanes on the road surface image.

As shown in FIG. 32B, the own vehicle 86
Three laser rangefinders are installed at the front end of the so as to give a predetermined directivity indicated by an arrow. A seventh embodiment is constructed by combining a vehicle having a rear-end collision warning device using three laser rangefinders with the traveling road detection device of the sixth embodiment. Figure 3
In (b) of FIG. 2, the inter-vehicle distance dR of the preceding vehicle 87 is output from the laser range finder in charge of the directivity on the right side. On the other hand, the distances dL and d from the laser rangefinders in charge of the directivity on the center side and the left side to the reflector 88 on the roadside belt.
C is output. At this time, the road surface image is in the state shown in (a) of FIG. 32, and the range indicated by an ellipse corresponds to each of the distance measuring area cross sections of the three laser distance meters.

In FIG. 31, the vehicle existence range x0e-X1e is obtained by performing vehicle detection using the road surface image similar to that of the fifth embodiment. Here, since the positional relationship between the camera for capturing the road surface image and the three laser rangefinders is fixed, the vector in the distance measuring direction of the three laser rangefinders in the road surface image is constant. The relationship between the distances (dL, dC, dR) of FIG. 32 obtained from the three lasers and the laser center coordinate positions (G0, G1, G2: vector) on the image is stored in advance as memory data. Preceding vehicle 8
If the center x-coordinate of the existence range of 7 is xc, the distance value dX (X = X) that minimizes the distance on the image coordinate system between the vehicle candidate coordinate obtained from the measurement value of the laser rangefinder and the coordinate xc.
By recognizing L, C, and R) as the distance to the vehicle on the own lane, the distance to the obstacle on the own lane can be measured with higher accuracy. This series of processing is shown in the flowchart of FIG.

In step 701, the road surface image is changed to step 7
At 02, the laser distance values are loaded into the memory.
In step 703, switching of the white line model according to the white line shielding range is executed, the road parameter is obtained in step 704 using this white line model, and the white line model is updated. In step 705, the presence / absence and presence position of the preceding vehicle are obtained from the road surface image, and if there is no preceding vehicle, the radar distance value is ignored in step 710. On the other hand, when there is a preceding vehicle, the laser range finder capturing the preceding vehicle is specified through steps 707 to 709, and the output value thereof is output as the inter-vehicle distance. According to the seventh embodiment, there is no concern that the distance value of the reflector existing on the roadside belt on the curved road is mistaken for the distance value of the preceding vehicle, and an erroneous warning is not issued. Therefore, the accuracy and reliability of the rear-end collision warning device are improved.

A modification of the seventh embodiment will be described with reference to FIGS. 34 to 36. 34 is an explanatory diagram of the detection of the preceding vehicle by the image processing, FIG. 35 is a flowchart of the entire process, and FIG. 36 is a flowchart of the selection of the radar distance measurement value. Here, similarly to the seventh embodiment, one corresponding to the preceding vehicle on the own lane is selected from the distances L0, L1 and L2 obtained from the three laser rangefinders. However, the laser rangefinder is specified using a procedure different from that of the seventh embodiment. Also,
In the fifth embodiment, when the ratio of the lateral edge length to the lane width is equal to or greater than a certain value, the position of the vehicle candidate point is obtained by determining that the vehicle is a preceding vehicle. Figure out candidate point positions.

In FIG. 34 (a), since the preceding vehicle 87 on the own lane has a height, a shadow is projected on the own lane. From the detected white line model, the center line (x
m, ym) and the luminance data of the pixels on the center line are extracted from the original image memory to obtain the height y = ys to y.
A luminance distribution in the range of e is created as shown in FIG. The darkest brightness pc corresponding to the shadow is detected, and its position (xc, yc) is output as vehicle candidate point coordinates. Figure 3
The radar distance value that minimizes the “distance from the laser center coordinate position on the road image (marked by X in the figure) to the vehicle candidate point coordinates (xc, yc)” shown in (a) of 2 is It is recognized as the distance to the preceding vehicle 87 on the lane.

As in the case of the seventh embodiment, the distance L0 obtained from the three laser rangefinders is stored in the memory of the arithmetic unit.
, L1, L2 and the laser center coordinate position (G0,
The relationship with G1 and G2: vector) is stored in advance. In FIG. 35, in step 721, the data of one frame of the original image or the edge-enhanced image is loaded into the arithmetic unit. In step 722, the white line model correction amount is calculated based on the detection data of the white line candidate points obtained through the window processing. In step 723, the white line model obtained on the previous screen is updated using this white line model correction amount. In step 724, the center line model Xm (y) is created based on the latest white line model. In step 725, a luminance distribution on xm is created within the range of height y = ys to ye as described above. In step 726, the minimum brightness value pc and the vehicle candidate point height yc are extracted and output. In step 727, the candidate vehicle height yc is set to the centerline model X.
Substituting into m (y), the vehicle candidate point position xc is calculated. At step 728, the minimum brightness value pc is equal to the predetermined threshold value p.
-It is determined whether it is less than or equal to t. If it is true, the coordinate position (xc, yc) is recognized as the vehicle candidate point coordinate in the coordinate system of the road surface image, and the process proceeds to step 731 in FIG.

In FIG. 36, in step 731, 3
Capture the latest distance measurements from a laser rangefinder in your book. In step 732, the laser center coordinate vector (G0, G1, G2) is read from the memory table stored in advance.
Read out. In step 733, the center coordinate vector (G0, G1, G2) is multiplied by each distance measurement value to obtain the barycentric position of the reference range on the road surface image, and the vehicle candidate point coordinates (xc, yc) are calculated from each barycentric position. To calculate the distance Lgi (i = 0 to 2). Step 734-
At 738, the value of j that minimizes the distance Lgi is obtained. In step 739, the laser radar distance value Lgi corresponding to j
(J) is determined as the distance value to the preceding vehicle 87 on the own lane and is output. According to the modified example of the seventh embodiment, it is possible to measure the distance to the obstacle on the own lane with higher accuracy.

[0115]

According to the traveling road detecting apparatus of the present invention, it is not necessary to perform arithmetic processing for coordinate conversion of data through the processing of detecting a white line from a road surface image and estimating a road parameter. The amount of calculation is greatly reduced as compared with the traveling road detection device described above, and even when a general arithmetic device is used, a road surface image of 10 to 100 screens per second can be processed with a margin. As a result, the three-dimensional shape of the road and the posture of the host vehicle can be output in real time even if the white line status is supplemented at fine time intervals. Therefore, the accuracy, reliability, and practicability of the traveling road detection device are improved, such as the application to various alarms and automatic driving.

When the setting of the small area on the road surface image is switched, the small area is set at a portion that can be detected with high accuracy on the road surface image. Therefore, the road parameter with high accuracy is output regardless of the position of the own vehicle in the lane. I can continue. At this time, unless the total set number of small areas is changed, the processing time does not increase. Further, when an alarm is output when an attempt is made to change lanes without indicating a direction, it is possible to prevent an accident from occurring due to the driver's carelessness or drowsiness. Furthermore, when a preceding vehicle in an adjacent lane is detected when an attempt is made to change lanes to an adjacent lane, a possibility of a collision is determined and an alarm is output, an accident accompanying the lane change can be prevented.

That is, the road model for estimating the road shape is represented by an amount equivalent to a parameter common to all the lane markers such as the road curvature and the position of the vehicle, and the calculation results of these parameters and The driving position of the vehicle in the driving lane is calculated based on a known constant, and the number of road white lines detected can be switched according to the result. It is possible to obtain the effect that the traveling lane area can be recognized, and the distance measurement target of the data of the distance measurement sensor such as another radar can be determined.

When the parameter description formula is switched according to the white line shielding range, the road parameter can be obtained with the maximum accuracy from the white line portion that can be captured on the road surface image. Also,
By simplifying the parameter description formula, the calculation load can be further reduced, and the surplus calculation reserve can be assigned to other processing, for example, recognition of the preceding vehicle. That is, it is possible to estimate road parameters efficiently.

[Brief description of drawings]

FIG. 1 is an explanatory diagram of a set coordinate system.

FIG. 2 is an explanatory diagram of a white line model in a camera coordinate system.

FIG. 3 is an explanatory diagram of new and old white line corresponding points in the image coordinate system.

FIG. 4 is an explanatory diagram of a configuration.

FIG. 5 is a flowchart of the entire processing.

FIG. 6 is a flowchart for updating a window.

FIG. 7 is a flowchart of detecting white line candidate points.

FIG. 8 is an explanatory diagram of a method of detecting white line candidate points.

FIG. 9 is an explanatory diagram of window setting positions.

FIG. 10 is an explanatory diagram of a configuration.

FIG. 11 is a flowchart for switching the number of white line detection lines.

FIG. 12 is an explanatory diagram showing an example of road white line detection.

FIG. 13 is an explanatory diagram showing window settings.

FIG. 14 is a block diagram showing a configuration of a third embodiment. ,

FIG. 15 is an explanatory diagram of the operation of the third embodiment.

FIG. 16 is an explanatory diagram of a lane departure warning.

FIG. 17 is an explanatory diagram of a modified example of the third embodiment.

FIG. 18 is an explanatory diagram of the overall configuration of the fourth embodiment.

FIG. 19 is a flowchart of overall processing.

FIG. 20 is a flowchart of window update.

FIG. 21 is an explanatory diagram of a basic configuration for selecting a white line model.

FIG. 22 is an explanatory diagram of a selection model.

FIG. 23 is a flowchart for model selection.

FIG. 24 is a flowchart for model selection.

FIG. 25 is an explanatory diagram of a vehicle candidate detection (lower shadow detection) method.

FIG. 26 is an explanatory diagram of a vehicle candidate detection (lower shadow detection) method.

FIG. 27 is an explanatory diagram of a vehicle candidate detection (vehicle width detection) method.

FIG. 28 is an explanatory diagram of a threshold value selection method for model selection.

FIG. 29 is an explanatory diagram of a selection model (in the case of only straight line detection).

FIG. 30 is an explanatory diagram of a basic configuration for selecting a white line model.

FIG. 31 is an explanatory diagram of vehicle detection results of the seventh embodiment.

FIG. 32 is an explanatory diagram of a radar beam arrival position vector.

FIG. 33 is a flowchart for selecting a radar distance value.

FIG. 34 is an explanatory diagram of detection of a preceding vehicle by image processing.

FIG. 35 is a flowchart of overall processing.

FIG. 36 is a flowchart of selecting a radar distance measurement value.

FIG. 37 is an explanatory diagram of a traveling road detection device of a conventional example.

FIG. 38 is an explanatory diagram of the accuracy of quadratic curve approximation.

[Explanation of symbols]

11 Image capturing unit 12 White line candidate point detection unit 13 Deviation amount calculation unit between white line candidate points and white line model point sequence 14 Parameter variation amount calculation unit 15 Correction unit 16 that corrects the previous parameter based on the variation amount Detection parameter holding unit 17 Calculating unit 21 for calculating the point sequence on the previous white line model (curve formula) 21 Window 22, 23 Coordinate system 24, 24G Road 25, 25G White line 26 Road surface image 41 Imaging unit 42 Image processing unit 43 White line detecting unit 44 Road parameter estimation Part 45 Vehicle deviation judgment part 46 White line detection number setting part 47 White line candidate point detection area setting part 50 Vanishing points 51, 52, 53 Window 61 Image input part 62 Edge extraction part 63 Straight line detection area setting part 64 White line candidate point detection Section 65 Road parameter calculation section 66 Lane change determination section 67 White line detection number setting section 68 Recognized lane number determination section 69 Lane area recognition section 0 Obstacle detecting part 71 on own lane Obstacle detecting part 72 on adjacent lane Change lane judging part 73 Direction instruction judging part 74 Collision possibility judging part 75 Warning generating parts 76, 77, 78 Preceding vehicle 81 Driving lane recognizing part 82 Inter-vehicle distance detection unit 83 Vehicle y-coordinate value estimation unit 84 in the image coordinate system Radar distance value selection unit 87 Preceding vehicle A1, A2, A3 Detection area L1, L2, L3 Lane

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI // G05D 1/02 G05D 1/02 K (56) References JP-A 5-151341 (JP, A) JP-A 4- 311211 (JP, A) JP 4-293109 (JP, A) JP 4-291405 (JP, A) JP 6-124398 (JP, A) JP 4-36878 (JP, A) (58) Fields investigated (Int.Cl. ⁷ , DB name) G08G 1/00-1/16 B60R 21/00 G06T 1/00 G06T 7/60 G05D 1/02

Claims (23)

(57) [Claims] 1. A road surface image in front of the vehicle is sampled, and image pickup processing is performed to extract a white line on the road surface by image processing. The white line on the road surface image is detected to detect the three-dimensional shape of the road surface and the road surface. In a traveling road detecting device for identifying the relative positional relationship of the vehicle, in the parameter description formula in which the quadratic curve approximation formula of the white line is converted into the coordinate system of the road surface image, the vehicle position in the traveling lane,
A detection parameter holding unit that holds a plurality of parameters respectively associated with the plane curvature of the road, the inclination of the vehicle with respect to the lane, the inclination of the road, and the width of the road, and the road surface with respect to the plurality of white line candidate points along the white line on the road surface image. The white line candidate point detecting means for obtaining position information using the coordinate system of the image, and the virtual white line candidate points calculated from a plurality of past parameters held in the detection parameter holding means, the latest obtained from the road surface image The position information of the white line candidate points is compared to obtain the latest values of the plurality of parameters, and the parameter correction unit that updates the plurality of parameters of the detection parameter holding unit is included. Road detection device. 2. The parameter correction means calculates a deviation amount between the virtual white line candidate point and the latest white line candidate point, and a deviation amount of the plurality of parameters based on the deviation amount. And a parameter variation calculating means for calculating the parameter variation calculating means, wherein the parameter variation calculating means calculates the parameter variation using a least square method. . 3. The parameter correction means includes road shape output means for calculating and outputting a plane curvature of a road, a vehicle inclination with respect to a lane, and a road inclination based on the updated latest plurality of parameters. The traveling road detection device according to claim 1 or 2. 4. The parameter description expression is a vehicle position in a traveling lane, a plane curvature of a road, a vehicle inclination with respect to a lane,
Let a, b, c, d, and e be a plurality of parameters associated with the road slope and the road width, respectively, and let x and y be the coordinate system of the road surface image, and k be an integer: x = (a + ke) (yd ) + B / (y-d) + c, The traveling road detection device according to claim 1, 2, or 3. 5. An image pickup input means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line on the road surface, and at least the width of the traveling lane and the own vehicle within the own lane based on the detection result of the white line. A road parameter estimating unit that calculates two road parameters associated with each position with respect to the road image every moment, and a plurality of road parameters set on the road image along the estimated position of the white line. For each of the small areas, the white line candidate point detecting means for obtaining the passing position coordinates of the white line extracted by the image processing means, the number of the white lines for setting the small areas on the road surface image, and the respective white lines The number of small areas to be allocated to the white line candidate point detection area is determined and the white line candidate point detection area setting means is set in the white line candidate point detection means. Own vehicle deviation determination means for identifying the own vehicle position in the lane, and the white line candidate point detection area setting means, based on the own vehicle position, of the white line to set the small area The number of lines and the number of small regions assigned to each white line are changed, and when the vehicle approaches one white line, the small region is also assigned to another white line in the approach direction. Road detection device. 6. The white line candidate point detection area setting means detects a difference between the width of the traveling lane and the position of the own vehicle in the own lane, the vehicle approaches one of the white lines, and the difference is a predetermined criterion. The traveling path detection device according to claim 5, wherein when the value range is exceeded, the small area is assigned to three white lines including the white line and two white lines on both sides. 7. The white line candidate point detection area setting means detects a difference between the width of the traveling lane and the position of the vehicle in the own lane, and the difference is a predetermined criterion after the vehicle crosses one of the white lines. 6. The traveling road detecting device according to claim 5, wherein when the vehicle enters the value range, only the two white lines on both sides of the new own lane are assigned the small areas. 8. The white line candidate point detection area setting means holds a constant set in advance in association with the width of the traveling lane, and the width of the traveling lane is constant regardless of the parameter estimating means. 8. The traveling road detection device according to claim 5, wherein the difference is detected using a value. 9. The white line candidate point detection area setting means sets a small area without changing the total number of allocations of the small areas even if the number of white lines for setting the small area changes. The traveling road detecting device according to claim 5, 6, 7, or 8. 10. The white line candidate point detection area setting means comprises:
When allocating the said small area with respect to three white lines, the said small area is allocated with respect to the white line which approached more than any of the other two white lines on both sides. The traveling road detection device according to 5, 6, 7, 8 or 9. 11. An image capturing and inputting means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line on the road surface, and at least the width of the traveling lane and the own vehicle within the own lane based on the detection result of the white line. A road parameter estimating unit that calculates two road parameters associated with each position with respect to the road image every moment, and a plurality of road parameters set on the road image along the estimated position of the white line. For each of the small areas, the white line candidate point detecting means for obtaining the passing position coordinates of the white line extracted by the image processing means, the number of the white lines for setting the small areas on the road surface image, and the respective white lines The number of small areas allocated to the white line candidate point detection area setting means to set the white line candidate point detection means, based on the road parameters Own vehicle deviation determination means for identifying the own vehicle position in the own lane, and the white line candidate point detection area setting means, based on the own vehicle position, the white line to set the small area And the number of small areas allocated to each white line are changed, and when the vehicle approaches one white line, the small area is also allocated to another white line in the approach direction, and When the vehicle approaches one of the white lines, the direction indicator discriminating means discriminates whether the direction indicator of the own vehicle is operated or not. In some cases, an alarm generation unit that outputs an alarm to the driver is provided, and the traveling road detection device. 12. An image pickup input means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line on the road surface, and at least the width of the traveling lane and the own vehicle within the own lane based on the detection result of the white line. A road parameter estimating unit that calculates two road parameters associated with each position with respect to the road image every moment, and a plurality of road parameters set on the road image along the estimated position of the white line. For each of the small areas, the white line candidate point detecting means for obtaining the passing position coordinates of the white line extracted by the image processing means, the number of the white lines for setting the small areas on the road surface image, and the respective white lines The number of small areas allocated to the white line candidate point detection area setting means to set the white line candidate point detection means, based on the road parameters Own vehicle deviation determination means for identifying the own vehicle position in the own lane, and the white line candidate point detection area setting means, based on the own vehicle position, the white line to set the small area And the number of small areas allocated to each white line are changed, and when the vehicle approaches one white line, the small area is also allocated to another white line in the approach direction, and Lane area recognition means for recognizing the own lane area on the road surface image, obstacle detection means for detecting an obstacle ahead of the lane area recognized by the lane area recognition means, and obstacle detection means for detecting the lane area. A collision possibility determining means for determining the possibility of a collision between the obstacle and the vehicle, and an alarm generating means for outputting an alarm to the driver when the possibility of the collision is affirmed. And a road detection device. 13. An image pickup input means for collecting a road surface image in front of the vehicle and performing image processing to extract a white line of the road surface, and at least the width of the traveling lane and the own vehicle within the own lane based on the detection result of the white line. A road parameter estimating unit that calculates two road parameters associated with each position with respect to the road image every moment, and a plurality of road parameters set on the road image along the estimated position of the white line. For each of the small areas, the white line candidate point detecting means for obtaining the passing position coordinates of the white line extracted by the image processing means, the number of the white lines for setting the small areas on the road surface image, and the respective white lines The number of small areas allocated to the white line candidate point detection area setting means to set the white line candidate point detection means, based on the road parameters Own vehicle deviation determination means for identifying the own vehicle position in the own lane, and the white line candidate point detection area setting means, based on the own vehicle position, the white line to set the small area And the number of small areas allocated to each white line are changed, and when the vehicle approaches one white line, the small area is also allocated to another white line in the approach direction, and Lane area recognition means for changing the area to be recognized depending on the number of white lines detected by the white line candidate point detection area setting means, and an adjacent lane when the vehicle approaches the white line and the small area is assigned to the outer white line An adjacent lane obstacle detecting means for identifying the presence or absence of an obstacle in the area, and a lane change possibility determining means for determining whether or not the lane can be changed without colliding with the obstacle when the obstacle exists. Characterized by having Road detection device. 14. A road surface image in front of the vehicle is sampled, and image pickup means is provided for image processing to extract a white line on the road surface. The white line on the road surface image is detected to detect the three-dimensional shape of the road surface and the road surface. In a traveling road detection device for identifying a relative positional relationship between vehicles, a detection parameter holding unit that holds a plurality of parameters in a parameter description expression obtained by converting the approximate expression of the white line into a coordinate system of a road surface image, and the above-mentioned on the road surface image For a plurality of white line candidate points along the white line, a white line candidate point detection means for obtaining position information using the coordinate system of the road surface image, and a virtual white line calculated from a plurality of past parameters held in the detection parameter holding means. For the candidate point, the position information of the latest white line candidate point obtained from the road surface image is compared, the latest value of the plurality of parameters is obtained, and the detected parameter is held. A parameter correction means for updating the plurality of parameters of the means, a white line shielding range detection means for obtaining a range in which the white line is blocked by a preceding vehicle in the own lane, and a white line as a secondary expression for the parameter description expression. A plurality of white line model storage means for storing a first parameter description expression approximated by a curve and a second parameter description expression adapted to simplify the first parameter description expression when the white line is shielded; A plurality of parameter description expressions in the plurality of white line model storage means are selected according to the detection state of the white line occlusion range detection means, and when the white line is shielded, the parameter correction means performs processing by the second parameter description expression. And a white line model switching means for executing. 15. The white line shielding range detecting means detects a preceding vehicle on the own lane from the road surface image to obtain a position of the preceding vehicle, and a white line is shielded based on the obtained vehicle position. 15. The traveling road detecting device according to claim 14, further comprising: a shielding range estimating unit that estimates a range to be covered. 16. The traveling road detecting apparatus according to claim 15, wherein the vehicle position detecting means calculates the position of the preceding vehicle by obtaining the side edge coordinates of the preceding vehicle from the road surface image. 17. The traveling road detecting apparatus according to claim 15, wherein the vehicle position detecting means calculates the position of the preceding vehicle by obtaining the lower end coordinates of the preceding vehicle from the road surface image. 18. The inter-vehicle distance detection means capable of detecting a preceding vehicle in a plurality of directivity directions and measuring the distance to an obstacle in each direction, based on the parameter. The inter-vehicle distance detection means obtains based on the own vehicle traveling lane recognition means for identifying which of the plurality of directivities the preceding vehicle in the own lane corresponds to, and the identification result of the own vehicle traveling lane recognition means. An inter-vehicle distance value selecting means for selecting an inter-vehicle distance of a preceding vehicle on the own lane from a plurality of inter-vehicle distances,
The traveling road detection device according to claim 14, further comprising: 19. The first parameter descriptive expression includes five parameters respectively associated with a vehicle position in a driving lane, a plane curvature of a road, a vehicle inclination with respect to a lane, a road inclination, and a road width, while a second parameter is included. 18. The parameter description formula includes four parameters other than one associated with the plane curvature of the road, 14, 15, 16, 17,
Or the traveling road detection device according to item 18. 20. The first parameter description formula is a quadratic curve formula,
20. The second parameter description expression is a linear linear expression, 14, 15, 16, 17, 18, or 19.
The traveling road detection device described. 21. The shift amount calculating unit calculates the shift amount between a virtual white line candidate point based on a selected parameter description expression and a latest white line candidate point obtained from the road surface image, A parameter fluctuation amount calculating means for calculating a fluctuation amount of the plurality of parameters based on a deviation amount; and, the parameter fluctuation amount calculating means calculates the parameter fluctuation amount using a least square method. 21. The traveling road detecting device according to claim 14, 15, 16, 17, 18, 19, or 20. 22. The parameter correction means includes a road shape output means for calculating a three-dimensional shape of a road based on the updated plurality of parameters and immediately outputting it. , 17, 18, 19,
20. The traveling road detection device according to 20 or 21. 23. Parameters a, b, c, d, and e respectively associated with the vehicle position in the traveling lane, the plane curvature of the road, the inclination of the vehicle with respect to the lane, the inclination of the road, and the width of the road are defined as a, b, c, d, e. When the coordinate system is x, y and i is an integer, the first parameter description expression is x = (a + ie) (yd) + b / (yd) + c, while the second parameter description expression is Is x = (a + ie) (y−d) + c.
7. The traveling road detection device according to 7, 18, 19, 20, 21, or 22.