JP2009186301A - Object detection device for vehicle - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an object detection device for a vehicle capable of detecting accurately a continuous object continuing to a remote place. <P>SOLUTION: An ECU extracts a point constituting an object on a road surface based on a vertical direction distance from the road surface from among each point constituting an image imaged by a camera, maps the extracted point on a plane specified by a vehicle body front-to-back direction and a vehicle width direction based on a front-to-back direction distance and a vehicle width direction distance of each corresponding point, extracts a point row arranged approximately in the vehicle front-to-back direction from among each point whose front-to-back direction distance is a prescribed front-to-back distance or shorter, derives an approximate curve for specifying the point row on the mapped plane, extracts each point whose vehicle width direction distance between itself and the approximate curve is shorter than a prescribed vehicle width distance from among each point whose front-to-back direction distance is the prescribed front-to-back distance or longer as a remote distance continuous point row continuing to a short distance continuous point row, and estimates the short distance point row and the remote distance continuous point row as a continuous object extending in the vehicle body front-to-back direction. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a vehicle object detection device that detects an object in front of a vehicle, and belongs to the field of vehicle safety technology.

  2. Description of the Related Art Conventionally, an object detection apparatus for a vehicle that detects a continuous object such as a road side wall or a guard rail existing in front of the host vehicle from an image acquired by an imaging unit is known.

  As such a vehicle object detection device, for example, Patent Document 1 discloses an imaging unit that images the front of the vehicle and a distance in the front-rear direction from the vehicle for each point that constitutes an image captured by the imaging unit. From the three-dimensional position information detecting means for detecting the vehicle width direction distance and the vertical distance from the road surface, and from the road surface detected by the three-dimensional position information detecting means from among the points acquired by the imaging means. Based on the vertical distance of the object, the object constituent point extracting means for extracting the points constituting the object on the road surface, and the points extracted by the object constituent point extracting means are detected by the three-dimensional position information detecting means Mapping means for mapping on a mapping plane defined by the vehicle longitudinal direction and the vehicle width direction based on the longitudinal distance and the vehicle width direction distance, and the vehicle width direction distance detected by the three-dimensional position information detecting means Having a means for deriving a linear formula corresponding to a continuous object (road side wall in Patent Document 1) based on a point within a predetermined range is disclosed.

JP-A-6-266828

  By the way, in the thing of the said patent document 1, although the road side wall is assumed to be a linear object, in the road which curves, a side wall will also become a curve. In addition, the size of the continuous object on the image is small at a distance from the host vehicle, and an obstacle such as a preceding vehicle may overlap on the image. As a result, a continuous object that continues to a distance can be detected with high accuracy. Has become difficult.

  Then, this invention makes it a subject to provide the vehicle object detection apparatus which can detect continuously objects, such as a road side wall which continues to a distant place, with sufficient precision.

  In order to solve the above-described problems, the present invention is configured as follows.

  First, the invention according to claim 1 of the present application is a vehicle object detection device that detects an object in front of a vehicle, and includes an imaging unit that images the front of the vehicle, and an image that is captured by the imaging unit. A three-dimensional position information detecting means for detecting a longitudinal distance from the vehicle, a vehicle width direction distance, and a vertical distance from the road surface, and the third order from the points acquired by the imaging means. Based on the vertical distance from the road surface detected by the original position information detection means, the object constituent point extracting means for extracting points constituting the object on the road surface, and the points extracted by the object constituent point extracting means, Mapping means for mapping on a mapping plane defined by the vehicle longitudinal direction and the vehicle width direction based on the longitudinal distance and the vehicle width direction distance from the vehicle detected by the three-dimensional position information detecting means, and the mapping Among the points mapped by the steps, a proximity line is extracted in which the front-rear direction distance from the vehicle detected by the three-dimensional position information detecting means is equal to or less than a predetermined front-rear distance and is continuously arranged substantially in the front-rear direction of the vehicle body. Based on the point sequence extracted by the distance point sequence extraction unit and the short-range point sequence extraction unit on the mapping plane, an approximate curve corresponding to the point sequence extends forward from the predetermined longitudinal distance position. Of the points mapped by the mapping means, the approximate curve deriving means for deriving, and the distance in the front-rear direction from the vehicle detected by the three-dimensional position information detecting means is greater than a predetermined front-rear distance, and the approximate curve deriving means A point whose distance in the vehicle width direction from the approximate curve derived in step 2 is equal to or less than a predetermined vehicle width distance is extracted as a long-distance point continuous to the short-distance point sequence extracted by the short-distance point sequence extraction means Long-distance point extracting means, a short-distance point sequence extracted by the short-distance point sequence extracting means, and a long-distance side point extracted by the long-distance point extracting means, And continuous object estimation means for estimation.

  According to a second aspect of the present invention, in the vehicle object detection device according to the first aspect, between the points mapped by the mapping unit and the approximate curve derived by the approximate curve deriving unit. The vehicle has an obstacle estimation means for estimating that the distance in the vehicle width direction is greater than the second predetermined vehicle width distance equal to or greater than the predetermined vehicle width distance as an obstacle existing in front of the vehicle. .

  According to a third aspect of the present invention, in the vehicle object detection device according to the first or second aspect, the approximate curve deriving unit is a short-distance point sequence extracted by a short-distance point sequence extraction unit. It is characterized in that an approximate curve is derived by performing multiple regression analysis.

  According to a fourth aspect of the present invention, in the vehicle object detection device according to any one of the first to third aspects, the short-distance point sequence extracting means includes a predetermined number in a substantially longitudinal direction of the vehicle body. A sequence of points arranged continuously is a short-range point sequence.

  According to a fifth aspect of the present invention, in the vehicle object detection device according to any one of the first to fourth aspects of the present invention, the three-dimensional position information detection means is a vehicle width direction distance. The distance from the center of the vehicle width is detected, and the short-range point sequence extracting means determines the distance from the center of the vehicle width detected by the three-dimensional position information detecting means among the points mapped by the mapping means. 3. A point of a predetermined vehicle width distance or more is extracted as a short distance point sequence.

  The invention according to claim 6 is the vehicle object detection device according to any one of claims 1 to 5, wherein the brightness detection means for detecting the brightness in front of the vehicle, the brightness And a predetermined front-rear distance changing unit that decreases the predetermined front-rear distance in the short-distance point sequence extracting unit and the long-distance point extracting unit as the brightness detected by the height detecting unit decreases. .

  The invention according to claim 7 is the vehicle object detection device according to any one of claims 1 to 6, wherein the short-distance point sequence extraction means is more than the predetermined front-rear distance position. It is configured to extract a sequence of points that are continuously arranged in the front-rear direction of the vehicle body in the front-rear direction in the forward direction, and the approximate curve deriving unit is configured to extract the short-range point ahead of the predetermined front-rear distance position. When there is a sequence of points that are continuously arranged in the longitudinal direction of the vehicle in the sequence, an approximated curve is derived including the sequence of points, and the long-distance point extraction means includes a point in front of the sequence of points. It is characterized in that a point on the far side is extracted from the inside.

  Next, the effect of the present invention will be described.

  First, according to the first aspect of the present invention, the front of the vehicle is imaged by the imaging means, and for each point constituting the image captured by the imaging means, the longitudinal distance from the vehicle, the vehicle width direction distance, And the vertical distance from the road surface is respectively detected by the three-dimensional position information detecting means, and based on the vertical distance detected by the three-dimensional position information detecting means from among the points acquired by the imaging means, A point constituting the object on the road surface is extracted by the object component point extracting unit, and the point extracted by the object component point extracting unit is the distance in the front-rear direction detected by the three-dimensional position information detecting unit by the mapping unit, Based on the vehicle width direction distance, mapping is performed on a mapping plane defined by the vehicle body longitudinal direction and the vehicle width direction.

  In the present invention, among the points mapped by the mapping means, the distance from the vehicle detected by the three-dimensional position information detecting means is not more than a predetermined longitudinal distance and is substantially continuous in the longitudinal direction of the vehicle body. A sequence of points arranged in a line is extracted by the short-range point sequence extraction unit, and based on the point sequence extracted by the short-range point sequence extraction unit, it extends to the front of the point sequence and corresponds to the point curve Is derived by the approximate curve deriving means. Of the points mapped by the mapping means, the front-rear direction distance from the vehicle detected by the three-dimensional position information detection means is larger than a predetermined front-rear distance, and the approximate curve derived by the approximate curve deriving means A point whose distance in the vehicle width direction is equal to or less than a predetermined vehicle width distance is a long-distance point continuous by the long-distance point extraction unit and the short-distance point sequence extracted by the short-distance point sequence extraction unit. The short-distance point sequence extracted by the short-distance point sequence extraction unit and the long-distance point extracted by the long-distance point extraction unit are estimated as continuous objects extending in the longitudinal direction of the vehicle body by the continuous object estimation unit. The That is, it is possible to accurately estimate a distant portion of a continuous object from a nearby portion that can be detected with high accuracy. In addition, as this continuous object, a road side wall, a guardrail, a large truck, etc. are contained, for example.

  According to a second aspect of the present invention, the distance in the vehicle width direction between the points mapped by the mapping means and the approximate curve derived by the approximate curve deriving means is the predetermined vehicle width distance. A point larger than the second predetermined vehicle width distance is estimated by the obstacle estimating means as an obstacle existing in front of the vehicle. That is, even when an obstacle or the like exists near the continuous object, the obstacle can be estimated (detected) with high accuracy.

  According to the invention described in claim 3, the approximate curve is derived by performing multiple regression analysis on the short distance point sequence extracted by the short distance point sequence extraction means. Therefore, the approximate curve can be derived with high accuracy.

  According to the fourth aspect of the present invention, a sequence of a predetermined number or more of points arranged in the longitudinal direction of the vehicle body is extracted as a short-range point sequence. In this case, if the predetermined number is a number that can exclude passenger cars, for example, it is possible to accurately detect continuous objects extending in the longitudinal direction of the vehicle body, such as large trucks, road side walls, and guard rails.

  According to the fifth aspect of the present invention, the short distance point sequence is extracted from the points whose distance from the vehicle width center of the vehicle is equal to or greater than the third predetermined vehicle width distance. Therefore, for example, even when a vehicle such as a passenger car exists near the center of the vehicle width, the vehicle is prevented from being detected as a continuous object. In other words, it is possible to accurately detect a short-distance point sequence corresponding to a continuous object extending forward and backward on the side of the vehicle body, such as a road side wall or a guard rail.

  By the way, in general, when the front of the vehicle becomes dark, noise components increase in the image acquired by the imaging unit, and the point sequence extraction accuracy by the short-distance side point sequence extraction unit decreases as the distance increases. That is, the accuracy of derivation of the approximate curve is reduced.

  However, according to the sixth aspect of the invention, the predetermined front-rear distance in the short-distance point sequence extraction unit and the long-distance point extraction unit is predetermined as the brightness in front of the vehicle is detected by the brightness detection unit and becomes darker. It is reduced by the front-rear distance changing means. Therefore, for example, even if the front of the vehicle becomes dark, it is possible to prevent the extraction accuracy of the short-range point sequence from greatly decreasing.

  According to the seventh aspect of the present invention, the short-distance point sequence extracting unit generates a sequence of points that are continuously arranged in the near-distance point sequence substantially in the vehicle longitudinal direction in front of the predetermined front-rear distance position. If it is extracted and exists, an approximate curve including this sequence of points is derived by the approximate curve deriving means. Therefore, the number of point sequences for deriving the approximate curve increases, and the accuracy of deriving the approximate curve is improved. In this case, the points on the long distance side may be extracted from the points ahead (far) from the line of points that are continuously arranged in the longitudinal direction in the vehicle front-rear direction.

  Hereinafter, a vehicle object detection device according to an embodiment of the present invention will be described.

  As shown in FIG. 1, a vehicle object detection device 10 according to an embodiment of the present invention is mounted on a vehicle 1, and includes a pair of left and right cameras (stereo cameras) 11L and 11R, an illuminance sensor 112, and an ECU 100. It has. The vehicle 1 is provided with an alarm device 13.

  Each camera 11L, 11R overlaps each other's imaging region at the front end of the roof portion of the vehicle 1 so that an image ahead of the vehicle 1 as shown in FIG. Are arranged in a predetermined direction. FIG. 2 shows an example of the image of the left camera 11L, but the image of the right camera 11L is obtained as an image in which a left-right parallax shift occurs with respect to the image of the left camera 11L. Each of the cameras 11L and 11R is configured by a digital camera having an image sensor such as a CCD, and converts the acquired image into a signal and transmits it to the ECU 100.

  The illuminance sensor 12 is disposed in the vicinity of the front window glass of the vehicle 1 so that the illuminance in front of the host vehicle 1 can be detected. The illuminance sensor 12 is constituted by a CdS (cadmium sulfide) sensor, for example, and transmits a signal corresponding to the illuminance outside the vehicle to the ECU 100.

  The alarm device 13 is arranged on the instrument panel in front of the driver's seat, receives a signal from the ECU 100, generates an alarm sound, or displays an alarm on the display together with the alarm sound.

  The ECU 100 includes a storage unit such as a ROM and a RAM, a CPU that performs an operation on data stored in the storage unit, and the image signal transmitted from each of the cameras 11L and 11R is transmitted at predetermined intervals. The received image signal is recorded in the storage unit as digital image data. This digital image data includes coordinate (xi, yi) information on the image plane of each pixel constituting each camera image GL, GR and luminance value information of each pixel. The ECU 100 converts the digital image data into the digital image data. Based on the detection result of the illuminance sensor 12, a predetermined process to be described later is performed. Then, ECU 100 outputs an alarm output signal to alarm device 13 based on the calculation result.

  Next, details of the object detection processing control executed by the ECU 100 will be described based on the flowchart of FIG. Note that the control according to this flowchart is repeatedly performed at predetermined intervals.

  First, in step S1, image signals are acquired from the cameras 11L and 11R and stored as image data in a storage unit. An illuminance signal is input from the illuminance sensor 12.

  In step S2, corresponding pixels are searched (detected) in the left camera image GL and the right camera image GR. Here, the corresponding pixels are pixels constituting the same part in various imaging targets in the image.

  Here, the detection of the corresponding pixel is performed based on an area-based matching method which is a known method. The area-based matching method is an attempt to search for a local luminance value (density) pattern around a pixel corresponding to a pixel in one image from the other image. is there. Specifically, as shown in FIGS. 4A and 4B, a 3 × 3 pixel window is set around one pixel of the left camera image GL, and an image surrounded by the window is set as a template image GT. As described above, matching is performed within the search range GS set on the epipolar line EL of the right camera image GR. In the present embodiment, this matching process is performed on all the pixels of the left camera image GL, and an SAD algorithm (Sum of Absolute Difference) is adopted as an algorithm used for this matching. That is, a total value (SAD value) of luminance differences between corresponding pixels between the template image GT set in the left camera image GL and an image in the right camera image GR to be matched is expressed by the following equation (1). When the total value is equal to or less than the predetermined value, the coordinates on the image plane are stored as the corresponding pixel (corresponding point) is detected. The algorithm used for the matching is not limited to the SAD algorithm. For example, an SSD (Sum of Squared Difference) algorithm or an NCC (Normalized Cross Correlation) algorithm may be used. Hereinafter, the corresponding pixel is referred to as a corresponding point.

M _L : Luminance value of the pixel of the left image M _R : Luminance value of the pixel of the right image

  In steps S3 to S5, the three-dimensional position information of each corresponding point detected in step S2 is calculated.

  That is, in step S3, for each corresponding point detected in step S2, the vehicle body longitudinal distance Lzi from the host vehicle 1 is used based on the positional deviation amount (parallax) of the corresponding point between the left and right camera images GL and GR. Calculated by the principle of triangulation. The vehicle body longitudinal distance Lzi from the host vehicle 1 (hereinafter referred to as the longitudinal distance Lzi) is a distance from the focal positions of the cameras 11L and 11R. FIG. 5 illustrates each corresponding point detected in step S2 at a pixel position (coordinate position) constituting the corresponding point of the left camera image GL with a density according to the front-rear direction distance Lz. The side walls, the vehicle in front of the host vehicle 1, a part of the road surface, and the like are shaded. Note that FIG. 4 is displayed as an example so that the detection results of corresponding points can be easily grasped visually, and these are not displayed during the process of this control.

  In step S4, for each corresponding point detected in step S2, a distance Lxi from the vehicle width center of the host vehicle 1 (hereinafter referred to as a vehicle width direction distance Lxi) is calculated. This distance Lxi is calculated based on the front-rear direction distance Lzi and the coordinates of the corresponding point.

  In step S5, the vertical distance Lyi (height from the road surface) of each corresponding point detected in step S2 from the road surface is calculated. The vertical distance Lyi is calculated based on the coordinates (x, y) of the corresponding point, the height of the cameras 11L and 11R from the road surface, the optical axis direction of the cameras 11L and 11R, and the like.

  In step S6, a corresponding point whose vertical distance Lyi detected in step S5 is a predetermined value or more (for example, 10 cm) is extracted. By doing so, the corresponding points constituting the road surface and relatively small objects on the road surface are excluded from the corresponding points, and the corresponding points for relatively tall objects such as road side walls and vehicles in FIG. Only will be extracted.

  In step S7, among the corresponding points extracted in step S6, corresponding points located at predetermined distances m1, m2, m3,... (Set in plural) from the vehicle 1 are extracted, and the extracted corresponding points. 6, the vehicle longitudinal direction (Z direction) and the vehicle width direction (X direction) based on the longitudinal distance Lzi and the vehicle width direction distance Lzi of the corresponding points calculated in steps S3 and S4, as shown in FIG. Are mapped on a plane defined by the following (hereinafter, this plane is referred to as a mapping plane). Here, the reason why only the corresponding points located at the predetermined distances m1, m2, m3,... Are extracted will be described together with the description of the extraction of the short distance point sequence described later.

  In step S8, a short distance point sequence extraction region is set on the mapping plane. This short distance point sequence extraction region is set to a region where the distance in the front-rear direction from the vehicle 1 is equal to or less than the predetermined front-rear distance Lzn and the distance in the vehicle width direction is equal to or greater than the predetermined vehicle width distance Lxn. In this case, the predetermined vehicle width distance Lxn is a constant value, but the predetermined longitudinal distance Lyn is set based on the illuminance detected by the illuminance sensor 12. Specifically, as shown in FIG. 7, the smaller the illuminance detected by the illuminance sensor 12, that is, the darker the predetermined longitudinal distance Lzn is, the darker the illuminance is.

  In step S9, among these mapped corresponding points, from the corresponding points belonging to the short-distance point sequence extraction region, a sequence of points lined up substantially in the longitudinal direction of the vehicle body is extracted as a short-distance point sequence.

  An example of this short distance point sequence extraction will be described with reference to FIG. 8. First, a line L1 connecting a point P1 having the smallest distance Lzi in the front-rear direction from the own vehicle 1 and a point P2 having the next smallest distance. It is determined whether or not the angle θ1 formed with respect to the longitudinal direction of the vehicle body (Z direction) is smaller than the predetermined angle θ0. If not, the point P1 is continuously arranged substantially in the longitudinal direction of the vehicle body with respect to P2. And stored in the storage unit. Next, it is determined whether or not the angle θ2 formed with respect to the line L1 of the line L2 connecting the point P2 and the point P3 having the next smallest distance from the host vehicle 1 is smaller than the predetermined angle θ0. In the case of being small, the point P3 is regarded as a point that is continuously arranged substantially in the vehicle longitudinal direction with respect to the point P2, and is stored in the storage unit. Next, it is determined whether or not an angle θ3 formed with respect to the line L2 of the line L3 connecting the point P3 and the point P4 having the next smallest distance from the host vehicle 1 is smaller than the predetermined angle θ0. . In the case of FIG. 8, since it is not small, the point P4 is not a point that is continuously arranged substantially in the vehicle longitudinal direction with respect to the point P3. That is, for the points belonging to the short-distance point sequence extraction region, the above-described processing is repeated in order from the point having the smallest distance from the host vehicle 1 until a point that is not substantially aligned in the longitudinal direction of the vehicle body is detected.

  Here, the reason why only the corresponding points located at the predetermined distances m1, m2, m3,. That is, if all the corresponding points in the short distance point sequence extraction region are considered for the short distance point sequence, when the pixels are arranged in a grid pattern as shown in FIG. 4, the angle θn is set between the adjacent corresponding points. The vehicle body longitudinal direction (Z direction) and the direction approximately 45 degrees with respect to this direction only exist, and it cannot be substantially determined whether or not they are aligned in the vehicle body longitudinal direction. Therefore, only corresponding points located at predetermined distances m1, m2, m3,... Are extracted so that the angle θn can be accurately determined. In addition, the amount of calculation can be reduced by thinning out in this way.

  Next, in step S10, it is determined whether or not the total number of points continuously arranged in the substantially longitudinal direction detected in step S8 is equal to or greater than a predetermined number α. Even if the points continuously arranged in the front-rear direction of the vehicle body are detected in step S9, when the total number of points is small, these points constitute a continuous object extending in the front-rear direction of the vehicle body, such as a side wall or a guard rail. However, since it is considered that it is a point constituting another object, it is determined which object is more likely.

  If the total number of points is equal to or greater than the predetermined number α (YES), the plurality of points are stored in the storage unit as a short-distance point sequence arranged substantially continuously in the longitudinal direction of the vehicle body, and the process proceeds to step S11. On the other hand, when the number is less than the predetermined number α (NO), it is not recognized as a short-distance point sequence, and proceeds to END.

In step S11, based on this short distance point sequence, as shown in FIG. 9, on the mapping plane, an approximate curve C1 extending forward from the predetermined front-rear distance Lzn position and corresponding to the short distance point sequence is obtained. Derived by multiple regression analysis. In the present embodiment, the approximate curve C1 is determined by a quadratic function shown in Equation 2. In that case, the coefficients a, b, and c are obtained by the least square method by substituting the coordinate values (xi, zi) of a plurality of points constituting the short-distance point sequence detected in step S8 into Equation 2. The approximate expression may be determined by a cubic function shown in Expression 3. In this case, as in the case of the quadratic function, the coefficients a, b, c, and d may be obtained by the least square method. In addition to the quadratic function and the cubic function, a function that can be appropriately approximated may be applied.

  In step S12, as shown in FIG. 10, among the points where the longitudinal distance Lzi is greater than the predetermined longitudinal distance Lzn, the distance in the vehicle width direction to the approximate curve C1 is within the predetermined vehicle width distance Lx0. Are extracted as long-distance points that are continuous with the short-distance point sequence. Specifically, for each point where the longitudinal distance Lzi is equal to or greater than the predetermined longitudinal distance Lzn, all the distances in the vehicle width direction from the approximate curve C1 are calculated, and whether each distance is within the predetermined vehicle width distance Lx0 or not is calculated. The point determined to be within the predetermined vehicle width distance Lx0 is extracted, and the extracted point is stored as a point on the far side. In this embodiment, in order to further improve the estimation accuracy (detection accuracy) of the continuous object and the obstacle, the approximate curve is again used by using the extracted far-side point in step S13 and the subsequent steps. However, it is also possible to estimate these short-distance point sequences and far-distance points extracted up to step S12 as continuous objects extending in the longitudinal direction of the vehicle body.

  In step S13, as shown in FIG. 11, on the mapping plane, on the basis of both the short-distance point sequence extracted as described above and the points on the long-distance point side, approximate curves corresponding to these points. C2 is derived by multiple regression analysis. This may be performed in the same manner as step S11.

  In step S14, as shown in FIG. 12, the distance in the vehicle width direction between the approximate curve C2 calculated again in step S13 among the points where the longitudinal distance Lzi from the host vehicle 1 is equal to or greater than the predetermined longitudinal distance Lzn. Are within the predetermined vehicle width distance Lx0, and are re-extracted as long-distance points continuous with the short-distance point sequence. This process may be performed in the same manner as step S12.

  In step S15, it is estimated that the short-distance point sequence and the points on the far-distance point side extracted as described above are continuous objects extending substantially in the longitudinal direction of the vehicle body. Here, the continuous object includes, for example, a road side wall, a guard rail, a large truck traveling on the side of the host vehicle 1, and the like.

  In step S16, a point in which the distance in the vehicle width direction between the approximate curve C2 calculated again in step S13 is a predetermined vehicle width distance Lx0 or more constitutes an obstacle existing in the traveling direction of the host vehicle 1 and presume. In this case, if points having a distance between adjacent points within the predetermined vehicle width distance Lx0 or more are grouped together, even if there are a plurality of obstacles, they are separated and recognized. can do. Here, the predetermined vehicle width distance in step S16 may be a value larger than Lx0.

  Then, the ECU 100 generates an alarm sound or the like from the alarm device 13 according to the positional relationship between the continuous object and the obstacle detected as described above.

  As described above, according to the vehicle object detection device 10 according to the present embodiment, the front of the host vehicle 1 is imaged by the cameras 11L and 11R. The ECU 100 detects the front-rear direction distance Lzi, the vehicle width direction distance Lxi, and the vertical distance Lyi from the road surface for each point constituting the captured image, and is acquired by the cameras 11L and 11R. The points constituting the object on the road surface are extracted from each of the points based on the vertical distance Lyi, and the extracted points are determined based on the longitudinal distance Lzi and the vehicle width direction distance Lxi. As shown in FIG. 6, mapping is performed on a mapping plane defined by the vehicle body longitudinal direction (z direction) and the vehicle width direction (x direction).

  In the present embodiment, among the mapped points, a short-distance point sequence in which the front-rear distance Lzi is equal to or smaller than the predetermined front-rear distance Lzn and is arranged substantially in the front-rear direction of the vehicle body is extracted. Based on the sequence, an approximate curve C1 extending forward from the short-range point sequence and corresponding to the point sequence is derived. Of the mapped points, a point where the front-rear direction distance Lzi is greater than the predetermined front-rear distance Lzn and the distance in the vehicle width direction to the approximate curve C1 is equal to or smaller than the predetermined vehicle width distance Lx0 is a short distance point sequence. Are extracted as long-distance points that are continuous with each other. These short-distance point sequences and long-distance points are estimated as continuous objects extending in the longitudinal direction of the vehicle body. That is, it is possible to accurately estimate a distant portion of a continuous object from a portion that exists in the vicinity that can be detected with high accuracy. Examples of the continuous object include a large truck, a road side wall, and a guardrail.

  Further, among the mapped points, a point whose distance in the vehicle width direction from the approximate curve C2 is larger than the predetermined vehicle width distance Lx0 is estimated (detected) as an obstacle existing in front of the host vehicle 1. The That is, even when an obstacle or the like exists in addition to the continuous object, the obstacle can be accurately estimated (detected).

  The approximate curves C1 and C2 are derived by performing multiple regression analysis on the short distance point sequence. Therefore, the approximate curves C1 and C2 can be derived with high accuracy.

  Further, when extracting a short-distance point sequence, it is extracted as a short-distance point sequence on the condition that a predetermined number α or more are arranged substantially in the longitudinal direction of the vehicle body. In this case, if the predetermined number α is set to a number that can exclude passenger cars, for example, it is possible to accurately detect continuous objects extending in the longitudinal direction of the vehicle body, such as large trucks, road side walls, and guard rails.

  Further, a short distance point sequence is extracted from points where the distance Lxi from the vehicle width center is equal to or greater than a predetermined vehicle width distance Lxn (predetermined value). Therefore, for example, even when a vehicle such as a passenger car exists near the center in the vehicle width direction, the vehicle is prevented from being detected as a continuous object. That is, it is possible to accurately detect a short-range point sequence corresponding to a continuous object extending in the front-rear direction on the side of the vehicle body, such as a road side wall, a guard rail, or a large truck.

  In general, the noise component increases in the images acquired by the cameras 11L and 11R as the front of the vehicle becomes darker. In the present embodiment, the brightness in front of the vehicle is detected by the illuminance sensor 12 and becomes darker. The predetermined longitudinal distance Lzn is reduced. Therefore, for example, even if the front of the vehicle becomes dark, it is possible to prevent the extraction accuracy of the short-range point sequence from greatly decreasing.

  Here, the correspondence between the processing of each step by the ECU 100 and the claims will be described. Steps S3 to S5 correspond to the three-dimensional position information detection means, and step S6 corresponds to the object composing point detection means. Step S7 corresponds to the mapping means, step S8 corresponds to the predetermined longitudinal distance changing means, step S9 corresponds to the short distance point sequence extracting means, step S10 corresponds to the short distance point sequence extracting means, and step S11 is performed. Corresponding to the approximate curve deriving means, Step S12 corresponds to the long distance point extracting means, Step S13 corresponds to the approximate curve deriving means, Step S14 corresponds to the long distance point extracting means, and Step S15 is the continuous object estimating means. Step S16 corresponds to the obstacle estimation means.

  Note that, as shown in FIG. 13, points that are continuous to the short-distance point sequence at a position where the distance in the front-rear direction from the vehicle 1 is larger than the predetermined front-rear distance Lzn, that is, at a position ahead of the short-range point sequence extraction region. There may be columns. Therefore, the sequence of these points may be detected as a sequence of points continuous to the short-range point sequence, and the approximate curve C1 ′ may be derived including this sequence of points. By doing this, the number of points for deriving the approximate curve increases, and the accuracy of deriving the approximate curve is improved. In addition, what is necessary is just to extract the point by the side of a long distance from the points located ahead rather than the row | line | column of the said point which continues in a short distance point sequence. This configuration corresponds to the configuration of claim 7 in the claims.

  INDUSTRIAL APPLICABILITY The present invention can provide a vehicle object detection device that can accurately detect a continuous object extending far away, and can be widely used in the automobile industry.

1 is a schematic diagram of an automobile equipped with a vehicle object detection device according to an embodiment of the present invention. It is an example of the image imaged with the camera. It is an example of the flowchart which shows the control by ECU of the object detection apparatus for vehicles. It is explanatory drawing of the Area-based matching technique used for a corresponding point search. It is the image which displayed and visualized the pixel detected as a corresponding point with the density | concentration according to the front distance of this corresponding point. It is explanatory drawing of the detection control of the continuous object by this ECU (the 1. It shows the state which mapped the corresponding point on the mapping plane prescribed | regulated by the vehicle body front-back direction and a vehicle width direction). It is a characteristic view which shows the relationship between the predetermined front-rear distance of a short distance point sequence extraction area | region, and illumination intensity. It is explanatory drawing of short distance point sequence extraction by this ECU. It is explanatory drawing of the detection control of the continuous object by this ECU (the 2). It is explanatory drawing of the same detection control (the 3). It is explanatory drawing of the same detection control (the 4). It is explanatory drawing of the same detection control (the 5). It is explanatory drawing of the other example of detection control of the continuous object by this ECU.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Vehicle 10 Vehicle object detection apparatus 11L, 11R Camera (imaging means)
12 Illuminance sensor (brightness detection means)
13 alarm device 100 ECU (three-dimensional position detection means, object constituent point extraction means, mapping means, short distance point sequence extraction means, approximate curve derivation means, long distance point extraction means, continuous object estimation means, obstacle estimation means, predetermined Front / rear distance change means)
Lzn predetermined longitudinal distance Lxn predetermined vehicle width distance (third predetermined vehicle width distance)
Lx0 predetermined vehicle width distance (predetermined vehicle width distance, second predetermined vehicle width distance)

Claims (7)

A vehicle object detection device for detecting an object in front of a vehicle,
Imaging means for imaging the front of the vehicle;
Three-dimensional position information detection means for detecting the distance in the front-rear direction, the width in the vehicle width direction, and the distance in the vertical direction from the road surface for each point constituting the image imaged by the imaging means;
Object component point extraction means for extracting points constituting the object on the road surface from the points acquired by the imaging means based on the vertical distance from the road surface detected by the three-dimensional position information detection means When,
The points extracted by the object constituent point extracting means are defined in the vehicle longitudinal direction and the vehicle width direction based on the longitudinal distance and the vehicle width direction distance from the vehicle detected by the three-dimensional position information detecting means. Mapping means for mapping on a mapping plane;
From the points mapped by the mapping means, a sequence of points in which the front-rear direction distance from the vehicle detected by the three-dimensional position information detection means is equal to or less than a predetermined front-rear distance and is continuously arranged substantially in the front-rear direction of the vehicle body is extracted. Short-distance point sequence extracting means for
An approximate curve deriving unit for deriving an approximate curve corresponding to the point sequence extending forward from the predetermined longitudinal distance position on the mapping plane based on the point sequence extracted by the short distance point sequence extracting unit; ,
Of the points mapped by the mapping means, the front-rear direction distance from the vehicle detected by the three-dimensional position information detection means is larger than a predetermined front-rear distance, and the approximate curve derived by the approximate curve deriving means A long-distance point extracting means for extracting a point whose distance in the vehicle width direction is a predetermined vehicle width distance or less as a long-side point continuous with the short-distance point sequence extracted by the short-distance point sequence extracting means;
Continuous object estimation means for estimating a short distance point sequence extracted by the short distance point sequence extraction means and a long distance side point extracted by the long distance point extraction means as a continuous object extending substantially in the longitudinal direction of the vehicle body; An object detection apparatus for a vehicle, comprising: In the vehicle object detection device according to claim 1,
Among the points mapped by the mapping means, a point in which the distance in the vehicle width direction between the approximate curve derived by the approximate curve deriving means is greater than a second predetermined vehicle width distance equal to or greater than the predetermined vehicle width distance. An object detection apparatus for a vehicle, comprising obstacle estimation means for estimating an obstacle present in front of the vehicle. In the vehicle object detection device according to claim 1 or 2,
The vehicle object detection device, wherein the approximate curve deriving unit derives an approximate curve by performing multiple regression analysis on the short distance point sequence extracted by the short distance point sequence extracting unit. In the vehicle object detection device according to any one of claims 1 to 3,
The short-distance point sequence extracting means uses a sequence of a predetermined number of points arranged in the longitudinal direction of the vehicle body as a short-distance point sequence. In the vehicle object detection device according to any one of claims 1 to 4,
The three-dimensional position information detecting means detects a distance from the vehicle width center of the vehicle as a vehicle width direction distance,
The short distance point sequence extracting unit is configured to select a point whose distance from the vehicle width center detected by the three-dimensional position information detecting unit is not less than a third predetermined vehicle width distance among the points mapped by the mapping unit. An object detection apparatus for a vehicle, characterized by being extracted as a point sequence. In the vehicle object detection device according to any one of claims 1 to 5,
Brightness detection means for detecting brightness in front of the vehicle;
A predetermined front-rear distance changing unit that reduces the predetermined front-rear distance in the short-distance point sequence extracting unit and the long-distance point extracting unit as the brightness detected by the brightness detecting unit is darker. A vehicle object detection device. In the vehicle object detection device according to any one of claims 1 to 6,
The short-distance point sequence extracting means is configured to extract a sequence of points that are continuously arranged in the front-rear direction of the vehicle body in the short-range point sequence in front of the predetermined front-rear distance position,
The approximate curve deriving means includes an approximate curve including the sequence of points when there is a sequence of points continuously arranged in the vehicle longitudinal direction in the short distance point sequence ahead of the predetermined longitudinal distance position. Derived,
The vehicle object detection device, wherein the long distance point extracting means extracts a point on a long distance side from points ahead of the row of points.