JP5293131B2 - Compound eye distance measuring device for vehicle and compound eye distance measuring method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a compound eye distance measuring device for a vehicle capable of reducing a distance measurement error following position deviation of a compound eye camera by a simple operation processing. <P>SOLUTION: The compound eye distance measuring device includes: a compound eye distance calculation means for calculating a distance between own vehicle and an object in imaging by a triangulation from imaging by two or more cameras 1, 2 loaded on own vehicle separately at a prescribed interval; a reference distance calculation means for calculating a distance between one's own vehicle and another vehicle as a reference distance based on imaging of a number plate (called as a number plate image) mounted on another vehicle near the own vehicle, which is imaged by the camera loaded on the own vehicle; and a measured distance calibration means for calibrating by using the reference distance, the distance between the own vehicle and the object calculated by the compound eye distance calculation means. <P>COPYRIGHT: (C)2010,JPO&amp;INPIT

Description

  The present invention relates to a compound eye distance measuring device for a vehicle and a compound eye distance measuring method.

  Development of a compound eye distance measuring device for vehicles that measures the distance between the vehicle and the surrounding objects based on the principle of so-called triangulation using the parallax of images from multiple cameras mounted on the vehicle. Yes. The vehicular compound eye distance measurement device uses the principle of so-called triangulation, which measures distance based on the parallax between images acquired by each camera, assuming that a plurality of cameras are in a predetermined positional relationship. Yes. For this reason, if a deviation occurs in the positional relationship of the presupposed camera, an error occurs in the measured distance. In the case of a camera mounted on a vehicle, the positional relationship between the cameras is likely to be shifted due to the vibration of the vehicle or the expansion / contraction of the camera housing due to a temperature difference. Citations 1 and 2 propose methods for reducing the distance measurement error associated with such positional deviation.

  In Cited Document 1, the height of the vehicle ahead is detected as the number of pixels in the captured image, and the vehicle ahead is determined from the relationship between the size (number of pixels in the image) and the inter-vehicle distance registered in advance. Proposed a front inter-vehicle distance measurement device that calculates the inter-vehicle distance from the vehicle ahead by calibrating the compound eye distance measurement value based on the principle of triangulation with the inter-vehicle distance calculated in this way Has been. This front inter-vehicle distance measuring device is capable of reducing a distance measurement error due to a positional deviation of a compound eye camera caused by vibration or the like.

Cited Document 2 discloses a three-dimensional coordinate acquisition apparatus that obtains a positional shift and a rotational shift between cameras based on a three-dimensional coordinate by a monocular stereo (motion stereo) method and a three-dimensional coordinate by a compound eye stereo method. . This three-dimensional coordinate acquisition apparatus is capable of calibrating a stereo camera without requiring a special apparatus if there are a plurality of cameras and an arithmetic unit.
Japanese Patent No. 3827368 JP 2007-263669 A

  In the front inter-vehicle distance measuring device described in Patent Document 1, an error due to triangulation is corrected based on the height of the photographed vehicle. However, due to the presence of a wide variety of vehicle types, the vehicle height varies, and the relationship between the type and height of the vehicle to be registered is very large and complicated, and there are many registered vehicles and vehicles that have been photographed. Matching becomes difficult, making it difficult to identify the vehicle type. In addition, depending on the vehicle, there is a possibility that the vehicle height is changed, or a load or a loading part is attached to the upper part, and there is a possibility that the original vehicle height is mistaken. Furthermore, if the vehicle type is classified and registered, the vehicle type search process becomes complicated and a search time is required. However, if the types of vehicles to be registered are limited, the registered vehicle type and the type of photographed vehicle may not match, and an error may occur in the height of the reference vehicle, which may increase the distance measurement error. is there.

  On the other hand, in the three-dimensional coordinate acquisition apparatus of Patent Document 2, it is necessary to perform a complicated and heavy calculation for the basic matrix estimation, and there are problems in terms of calculation cost and calculation time.

  In view of the above problems, an object of the present invention is to provide a compound eye distance measuring device for a vehicle and a compound eye distance measuring method capable of calibrating a distance measurement error caused by a positional deviation of a compound eye camera by simple arithmetic processing.

  The present invention uses a parallax of images by two or more cameras mounted on a vehicle apart from each other, and based on the principle of triangulation, a compound eye distance for measuring a distance from an object around the traveling vehicle It has a calculation means. The compound eye distance calculation means assumes that a plurality of cameras are in a predetermined positional relationship, and based on the parallax between images (captured) captured by each camera, the distance from the reference camera to the object being imaged Measure. Since the plurality of cameras are respectively mounted at predetermined positions of the vehicle, the distance from the vehicle to the object is different from the distance from the reference camera to the object. Even if it is assumed that the position of the camera is based on the reference position (usually the tip), the position of the reference camera is the position of the vehicle for convenience. When accurately calculating the distance from the reference position of the vehicle to the object, the reference camera mounting position to the vehicle is known in advance, so the reference camera mounting position and the reference position of the vehicle Based on the positional relationship, the distance from the reference camera mounting position to the object may be corrected to the distance from the vehicle reference position to the object.

  The compound eye distance measuring device of the present invention is based on the imaging of a license plate (referred to as a license plate image) attached to another vehicle in the vicinity of the host vehicle in an image taken by a camera mounted on the host vehicle. And a reference distance calculation means for calculating a distance between the vehicle and the other vehicle as a reference distance. The reference distance calculation means searches for a license plate image attached to another vehicle in the vicinity of the host vehicle in an image taken by a camera mounted on the host vehicle, and measures the size of the license plate image in the image. The distance from the camera (own vehicle) to the license plate (other vehicle) from the relationship between this size and the size of the license plate image in the image stored in advance and the distance from the own vehicle to the license plate Is calculated as a reference distance. When there are only one or several types of license plates as in Japan, the relationship between the size of the license plate image and the distance between the host vehicle and other vehicles can be easily calculated from the camera performance. This may be stored as a calculation formula.

The compound eye distance measuring device of the present invention includes measurement distance calibration means for calibrating the distance between the vehicle and the object calculated by the compound eye distance calculation means using the reference distance. The measurement distance calibration means calculates the distance between the own vehicle and the license plate attached to the other vehicle that is the object of distance measurement by the reference distance calculation means, calculated by the compound eye distance calculation means, by the reference distance calculation means. A calibration coefficient for adjusting to (the distance between the host vehicle and the number plate attached to the other vehicle) is calculated. Then, the measurement distance calibration unit calibrates the distance from the host vehicle to the object calculated by the compound eye distance calculation unit using the calculated calibration coefficient.
Specifically, the measurement distance calibration means includes the reference distance and the number of the license plate photographed by the two or more cameras at the same time when the license plate image is captured by the reference distance calculation means by the compound eye distance calculation means. A calibration coefficient is calculated from a ratio with the distance from the host vehicle to the license plate calculated from the imaging, and the distance from the host vehicle to the target calculated by the compound eye distance calculation means is calibrated using the calibration coefficient.

  Normally, in the distance measurement using only the compound eye distance calculation means, if the positional relationship between the cameras, which is a premise, deviates, an error is included in the measured distance. In the case of a camera mounted on a vehicle, the positional relationship between the cameras easily shifts due to the vibration of the vehicle or the expansion / contraction of the camera housing due to environmental temperature changes. It tends to be a big error in measurement. Therefore, with respect to the distance from the host vehicle to the object measured by the compound eye distance calculating means, the error due to the positional deviation of the plurality of cameras, which is a problem of the compound eye distance calculating means at the time of distance measurement, was calculated by the reference distance calculating means. By calibrating using the reference distance, the accuracy of distance measurement by the compound eye distance calculating means can be improved.

  In addition, the license plate has a constant size, size, shape, and color that are easy to detect, and is placed at an easy-to-see position in all traveling vehicles. In most cases, it is easy to explore. Therefore, if the distance is detected based on the size of this license plate image taken at the same time as the distance measurement in the compound eye distance calculation means, the reference distance is calculated, and used for the calibration of the distance in the compound eye distance calculation means, there is a problem. The measurement error of the compound eye distance calculation means due to the positional deviation between the cameras can be eliminated.

  Preferably, the reference distance calculation means includes a license plate size detection unit, a license plate size storage unit, and a reference distance calculation unit. The license plate size detection unit detects the size of a license plate image in an image taken by one or more cameras. The license plate size storage unit stores in advance the relationship between the size of the license plate image and the distance between the host vehicle and the other vehicle. The reference distance calculation unit is based on the size of the license plate image detected by the license plate image detection unit, the size of the license plate image stored in the license plate size storage unit, and the relationship between the distance between the host vehicle and the other vehicle. A reference distance is calculated.

  One camera is sufficient to capture the license plate image, but two or more cameras may be used in order to increase the field of view from the vehicle. Further, if the camera used in the license plate size detection unit is also used as one of two or more cameras used in the compound eye distance calculation means, there is an advantage that the processing operation in detecting the license plate image in the photographed image can be shortened.

  The measurement distance calibration means preferably includes an other vehicle distance detection unit, a calibration coefficient calculation unit, and a compound eye distance calibration unit. The other vehicle distance detection unit detects a compound eye calculation distance that is an inter-vehicle distance between the host vehicle calculated by the compound eye distance calculation means and the other vehicle whose license plate image has been taken by the reference distance calculation means. The calibration coefficient calculation unit compares the reference distance calculated by the reference distance calculation means with the compound eye calculation distance of the other vehicle calculated by the compound eye distance calculation means, and the object from the own vehicle calculated by the compound eye distance calculation means. Calculate the calibration coefficient that calibrates the distance to. The compound eye distance calibration unit calibrates the distance between the vehicle and the object calculated by the compound eye distance calculation means using the calibration coefficient calculated by the calibration coefficient calculation unit, and the vehicle and the object with a small measurement error. And the distance is calculated.

  The compound eye distance calculation means includes a parallax calculation unit that calculates parallaxes of two or more cameras, and a distance measurement object such as a license plate of the host vehicle and another vehicle based on the parallax calculated by the parallax calculation unit. It is preferable that a distance calculation unit for calculating the distance is provided.

  In the reference distance calculation means, the license plate image attached to the other vehicle in the photographed image is preferably the largest license plate image in the image photographed by the camera mounted on the host vehicle. A large license plate image is an image of a license plate existing in the vicinity of the host vehicle. If a reference distance is calculated based on this, a more accurate reference distance can be calculated.

  In addition, when a license plate image of a predetermined size or larger cannot be detected in an image captured by the reference distance calculation means, a fixed threshold is provided for the size of the license plate image. The compound eye calculation distance may be calibrated by the same calibration operation as the distance measurement, or if the previous distance measurement does not exist, the calibration coefficient is set to 1, that is, the substantial calibration may not be performed.

  If the license plate image in the photographed image is too small, the reference distance calculation error by the reference distance calculation means may increase. Therefore, the calibration operation in the previous distance measurement, for example, the compound eye using the same calibration coefficient as the previous one. In some cases, the distance measurement error is smaller when the calculated distance is calibrated or the compound eye calculated distance is calibrated with a calibration coefficient of 1.

In the compound eye distance measuring method of the present invention, the object around the host vehicle may be measured as described above using the vehicle compound eye distance measuring device described above. That is, the compound eye distance measuring method of the present invention detects the size of a license plate (referred to as a license plate image) captured by a camera mounted on the host vehicle and attached to another vehicle in the vicinity of the host vehicle. A license plate image detection step is used, and the relationship between the distance between the host vehicle and the other vehicle registered in advance and the size of the license plate image detected in the license plate image detection step are used to A reference distance calculating step of calculating a distance to be a reference distance; Further, the compound eye distance measuring method of the present invention is based on the principle of triangulation from imaging of an object in the vicinity of the own vehicle by two or more cameras mounted on the own vehicle at a predetermined interval. A compound eye distance calculating step for calculating a distance to an object, and the two or more cameras photographed at the same time as the reference plate and the license plate image in the reference distance calculating step in the compound eye distance calculating step A calibration coefficient is calculated from the ratio of the vehicle and the license plate distance calculated from the imaging of the license plate, and the distance between the vehicle and the object calculated in the compound eye distance calculation step is calibrated using the calibration coefficient. A measurement distance calibration step to perform.

  In a compound eye distance measurement method that is mounted on a vehicle and uses the parallax of images from multiple cameras to measure the distance between the vehicle and an object in the vicinity based on the principle of triangulation, the multiple cameras are located at predetermined positions. Assuming that there is a relationship, the distance is measured based on the parallax between images acquired by each camera. Therefore, if a deviation occurs in the positional relationship of the presupposed camera, an error is included in the measured distance. In the case of a camera mounted on a vehicle, the positional relationship between a plurality of cameras is likely to be shifted due to expansion and contraction of the camera housing due to vehicle vibrations and temperature differences, and this gap tends to increase distance measurement errors. This measurement error changes with changes in the driving state and driving environment of the vehicle. In the compound eye distance measuring method of the present invention, the distance to the vehicle is detected as the reference distance based on the size of the license plate attached to the vehicle in the vicinity of the host vehicle in the reference distance calculating step, and the compound eye distance measuring step The distance measurement value is calibrated using the reference distance in the measurement distance calibration step. Thereby, it is possible to provide distance data of objects around the vehicle in which measurement errors associated with camera position shifts are reduced in accordance with vehicle running conditions and environmental changes by simple calculations.

  According to the present invention, it is possible to provide a compound eye distance measuring device for a vehicle and a compound eye distance measuring method capable of calibrating a distance measurement error caused by a positional deviation of a compound eye camera by simple arithmetic processing.

  A specific embodiment of the present invention will be described with reference to FIG. FIG. 1 is a schematic block diagram of a compound eye distance measuring apparatus for a vehicle according to the present invention. In FIG. 1, reference numeral 1 denotes an integral left camera including a lens block 1a, an area sensor 1b, and a lens barrel 1c. An image of the scenery in front of the host vehicle in the upper part of the figure (in the direction of the arrow) is formed on the imaging surface of the area sensor 1b by the lens block 1a and is taken by the area sensor 1b. The area sensor 1b is composed of a solid-state image sensor such as a CCD or a CMOS.

  Similarly, reference numeral 2 denotes a right camera including a lens block 2a, an area sensor 2b, and a lens barrel 2c, and a predetermined distance from the left camera 1 on a plane substantially perpendicular to the lens optical axis of the lens block 1a and the lens block 2a. Are placed apart by. The left camera 1 and the right camera 2 have substantially the same lens configuration and camera characteristics, and the right camera 2 also captures the front scenery of the host vehicle by the area sensor 2b. The left camera 1 and the right camera 2 pick up substantially the same front landscape images that differ only in parallax based on the difference in their arrangement.

  Reference numeral 3 denotes a housing that holds the left camera 1 and the right camera 2 and makes an appearance of a stereo camera. This stereo camera is mounted at a position convenient for capturing a front landscape, for example, at both ends of the front window of the host vehicle or next to two headlights. In the compound eye distance measuring device of the present invention, the left camera 1 and the right camera 2 may be directed toward the rear or side of the host vehicle so as to capture a rear landscape or a side landscape, even if it is not a front landscape. Good. In this description, a compound eye distance measuring device using two cameras is used, but three or more cameras may be used. Further, although the left camera 1 is described as a reference camera, the left and right can be reversed, and two cameras can be arranged vertically.

  The front landscape of the host vehicle is imaged by the left camera 1 and the right camera 2 at the same timing by a trigger signal of an image input command from the trigger unit in FIG. If there is a difference in the imaging timing between the left camera 1 and the right camera 2, conditions other than the difference based on the parallax between the left camera 1 and the right camera 2, such as the influence of the traveling speed, occur between the acquired images. Therefore, in order to cause a distance measurement error, it is necessary to eliminate as much as possible the difference in imaging timing between the left camera 1 and the right camera 2.

  Image signals captured by the left camera 1 and the right camera 2 are transferred to the frame memory L and the frame memory R, respectively, and then subjected to predetermined arithmetic processing. First, the image signal L stored in the frame memory L is used for calculation of the reference distance by the reference distance calculation means. The reference distance calculation means detects the size of the number plate of another vehicle in the captured image (the number plate may be abbreviated as NP, and the number plate image may be abbreviated as number plate image or NP image). A license plate size detector (NP size detector), and a license plate size storage unit (NP) for storing the relationship between the actual distance between the host vehicle and the other vehicle with respect to the size of the license plate image in the image Size storage unit), a reference distance calculation using a distance calculated from the relationship between the size of the NP image detected by the NP size detection unit and the size of the NP image stored in the NP size storage unit and the inter-vehicle distance as a reference distance Part.

  The specific operation of the reference distance calculation means first reads the image signal L stored in the frame memory L to the NP size detection unit. The NP size detection unit searches an image area corresponding to the license plate in the image and detects an image area constituting the license plate image. The search for a license plate image uses, for example, a rectangular template model that assumes a license plate shape, finds a correlation value while changing the template size, position, and rotation as parameters, and obtains the largest correlation value (correlation It can be performed by a model matching process for obtaining a (high) image area. In this case, when a plurality of license plate images are detected in the entire image area of the image signal L, it is preferable to select the largest license plate image among them. This is because the largest license plate image is often the image of the license plate at the closest distance to the host vehicle, and the calculation error of the reference distance is reduced.

  Next, the number of pixels in the longitudinal direction in the rectangular image area constituting the license plate is detected to obtain the detection size of the license plate. In the NP size storage unit (number plate size registration DB), officially standardized relationship data between the license plate image size and the inter-vehicle distance is registered. In the reference distance calculation unit, the distance between the host vehicle and the other vehicle can be calculated from a comparison between the detected number image size and the registered image size of the license plate. In addition, the actual standardized license plate size is registered in the NP size storage unit (the license plate size registration DB), and the image size of the detected license plate is registered in the reference distance calculation unit. From the license plate size, the distance from the left camera 1 to the license plate can be detected by the following equation (1).

式（１）において、Ｅは自車両からナンバープレートまでの距離、Ｒは記憶してあるナンバープレートの長手方向のサイズ（長さ）、ｆはレンズブロック１aの焦点距離、Ｄは検出したナンバープレートの長手方向の画素数（ピクセル）、ｐはエリアセンサ１ｂの１画素のサイズ（長さ／ピクセル）である。

In Equation (1), E is the distance from the host vehicle to the license plate, R is the size (length) of the stored license plate in the longitudinal direction, f is the focal length of the lens block 1a, and D is the detected license plate. Is the size (length / pixel) of one pixel of the area sensor 1b.

  In addition, when there are a plurality of license plate sizes, the number plate type may be specified by detecting the difference in the shape, color, or pattern of the license plate by the above-described model matching process.

  On the other hand, the compound eye distance calculating means takes in the image signal L and the image signal R stored in the frame memory L and the frame memory R, and calculates the distance of the object in the image using the principle of triangulation. The compound eye distance calculation means calculates a distance for each object in the image based on the parallax calculation unit for detecting the parallax between the image signal L and the image signal R, and the parallax between the detected image signal L and the image signal R. A distance calculation unit for calculating is provided.

  The operation of the compound eye distance calculating means will be described in order with reference to the image signal L and the image signal R in FIG. An image signal L and an image signal R in FIG. 2 represent forward landscape images by the left camera 1 and the right camera 2, respectively. If the positions and orientations of the left camera 1 and the right camera 2 are adjusted precisely, parallax based on the distance between the two cameras occurs in both images, and there are no other error factors. The parallax detection unit accesses the image signal L stored in the memory, and first pays attention to the upper left minute region L (0, 0) in the image signal L as shown in FIG. Then, an image correlation value of a micro area having the same size as the micro area L (0, 0) in the image signal R with respect to the micro area L (0, 0) is obtained. While sequentially shifting the position of the micro area of interest in the image signal R, the position of the micro area R (0, 0) in the image signal R having the largest correlation value is obtained. In this embodiment, since the left camera 1 and the right camera 2 are arranged horizontally, the parallax occurs only in the horizontal direction, and the direction in which the minute area is shifted may be only in the horizontal direction. When parallax occurs also in the vertical direction, it is necessary to search by shifting a minute region in the vertical direction.

  The amount of deviation between the minute region L (0,0) in the image signal L and the minute region R (0,0) in the image signal R is the parallax δ (0,0) in the minute region L (0,0). In a similar manner, the micro area L (1, 0), (2, 0), (3, 0),... L (m, n) (m = 1) of the image signal L corresponding to the entire left image. ˜M, n = 0 to N) parallax δ (1, 0), (2, 0), (3, 0),... Δ (m, n) (m = 1 to M, n = 0) ~ N). Here, m and n represent arbitrary positions of a minute area in the horizontal direction x and the vertical direction y of the left image L in FIG. 2, respectively, and M and N are the x direction (horizontal direction) and the y direction of the left image L, respectively. This is a value obtained by dividing the number of pixels in the (vertical direction) by the number of pixels in each direction of the minute area, and represents the number of minute areas in the x and y directions. The value of δ is obtained in pixel units (pixels) of the area sensors 1b and 2b. The detected δ (m, n) (pixel) is substituted into the following equation (2) to obtain the distance A (m, n) for each minute region L (m, n).

ここで、Ｂは左カメラ１と右カメラ２の間隔、ｆはレンズブロック１ａ、２ａの焦点距離（ピクセル）、ｐはエリアセンサ１ｂ、２ｂの１画素のサイズ（長さ／ピクセル）である。

Here, B is the distance between the left camera 1 and the right camera 2, f is the focal length (pixels) of the lens blocks 1a and 2a, and p is the size (length / pixel) of one pixel of the area sensors 1b and 2b.

  The vehicular compound eye distance measuring device of the present invention includes a measuring distance calibration means for calibrating the distance A (m, n) for each minute region of the left image L (see FIG. 1). The measurement distance calibration means recognizes the image of another vehicle existing in the left image L from the distance data for each minute area calculated by the distance calculation unit, and detects the inter-vehicle distance between the own vehicle and the recognized other vehicle. The distance calculation unit calculates the distance between the other vehicle distance detection unit, the distance between the host vehicle and the other vehicle detected by the other vehicle distance detection unit, and the reference distance between the host vehicle and the other vehicle calculated by the reference distance calculation unit. The distance calibration coefficient calculation unit for calculating the calibration coefficient for calibrating the distance data for each minute area and the distance data for each minute area calculated by the distance calculation unit using the calibration coefficient calculated by the distance calibration coefficient calculation part are calibrated. And a compound eye distance calibration unit.

  The other vehicle distance detection unit uses the distance A (m, n) data for each minute region L (m, n) in the left image L to obtain an image region corresponding to the license plate image detected by the NP size detection unit. Explore. Specifically, similar to the NP size detection unit already described, the search may be performed by model matching processing. Or you may obtain | require the license plate image in the left image L with reference to the position of the license plate image detected by the NP size detection part. By referring to the position of the license plate image detected by the NP size detector, the image area of the license plate image in the left image L can be easily searched. In particular, if the camera for the reference distance measurement means and the camera 1 are shared, and the image data used in the NP size detection unit and the image data of the left image L are the same, the license plate image in the NP size detection unit is As it is, it becomes the license plate image in the left image L as it is.

  Next, using the data of the distance A (m, n) for each micro area L (m, n) of the image area corresponding to the searched license plate image, the distance of the image area corresponding to the searched license plate image is calculated. An average value and a representative value are calculated, and this is used as a compound eye calculation distance from the left camera 1 to the license plate. Normally, the distance from the left camera 1 to the license plate is the inter-vehicle distance from the host vehicle to other vehicles, but the inter-vehicle distance is calculated by taking into account the error in the inter-vehicle distance due to the left camera 1 and the license plate arrangement. May be.

  The distance calibration coefficient calculation unit is a distance calibration coefficient based on a ratio of the compound distance calculated by the other vehicle distance detection unit based on the reference distance that is the inter-vehicle distance calculated from the license plate size in the reference distance calculation unit. Is calculated.

  The compound eye distance calibration unit uses the distance calibration coefficient calculated by the distance calibration coefficient calculation unit, and the own vehicle (more precisely, the left camera 1) for each minute region L (m, n) in the left image L calculated by the distance calculation unit. ) From A) is calculated. In addition, when grasping the image of each object in the left image L, an object recognition method (model matching method) similar to the license plate recognition described in the NP size detection unit and the other vehicle distance detection unit is used. That's fine. Or you may recognize the whole area | region where the adjacent micro area | region L (m, n) exists in the substantially same distance A (m, n) as one object. In this case, the shape and distance of the object in the left image L can be recognized, but the object name cannot be recognized, but the distance from the host vehicle can be recognized.

  In this embodiment, the reference distance, which is the distance obtained from the license plate size by the reference distance calculation means, has a smaller measurement error due to camera vibration and temperature expansion / contraction than the distance measurement by the compound eye distance calculation means. However, in the reference distance calculation means, if the vehicle for detecting the license plate is far away, the size of the license plate image is recognized to be small, and the distance detection accuracy decreases as the number of pixels constituting the image decreases. . On the other hand, the error factor in the compound eye distance calculation means is mainly due to the fluctuation of the relative position between the cameras due to camera vibration and temperature expansion / contraction, and is relatively less affected by the distance of the object in the captured image. Therefore, in the compound eye distance measuring device for a vehicle according to the present embodiment, the reference distance is calculated by the reference distance calculation means from the license plate image, preferably the license plate image of another vehicle in the vicinity of the host vehicle, and the compound eye is calculated based on the reference distance. An accurate distance is calculated by calibrating the distance of an object such as another vehicle measured by the distance calculation means.

  It should be noted that the camera in the reference distance calculation means and the camera in the compound eye distance calculation means (any one may be used) can be shared, and the license plate image recognition process can also be used in common.

  In the reference distance calculation means, if the vehicle for detecting the license plate is far away, the size of the license plate image is recognized to be small, and the distance detection accuracy decreases as the number of pixels constituting the image decreases. easy. For this reason, it is preferable to perform a distance detection based on the license plate size when a vehicle is as close as possible by providing a threshold value such that the number of constituent pixels of the license plate image becomes a certain value or more.

  In this way, the compound eye distance measurement device for a vehicle according to the present embodiment is complicated and does not require a large-scale device or image processing, and eliminates the disadvantages of the above two distance measurement methods and is in all positions. The distance of an object can be measured accurately and quickly.

Block diagram of a compound eye distance measuring device for vehicles Example of imaging with two left and right cameras

Explanation of symbols

1: Left camera 1a: Lens block 1b: Area sensor 1c: Lens barrel 2: Right camera 2a: Lens block 2b: Area sensor 2c: Lens barrel 3: Housing

Claims (8)

Compound eye distance calculating means for calculating a distance between the own vehicle and the object being imaged according to the principle of triangulation from imaging by two or more cameras mounted on the own vehicle at a predetermined interval,
Based on the image of a license plate attached to another vehicle in the vicinity of the host vehicle (referred to as a license plate image) in an image taken by a camera mounted on the host vehicle, the distance between the host vehicle and the other vehicle is determined. A reference distance calculating means for calculating a reference distance;
The license plate from the vehicle calculated from the reference distance and the imaging of the license plate taken by the two or more cameras simultaneously with the imaging of the license plate image by the reference distance calculation means by the compound eye distance calculation means. A calibration coefficient is calculated from the ratio to the distance to, a measurement distance calibration unit that calibrates the distance between the vehicle and the object calculated by the compound eye distance calculation unit using the calibration coefficient ;
A compound eye distance measuring device for vehicles. The reference distance calculation means includes a license plate size detection unit that detects the size of the license plate image in an image taken by one or more cameras.
A license plate size storage unit for storing in advance the relationship between the size of the license plate image and the distance between the host vehicle and another vehicle;
From the size of the license plate image detected by the license plate image detector, the size of the license plate image stored by the license plate size storage unit, and the relationship between the distance between the host vehicle and the other vehicle, A reference distance calculation unit for calculating a reference distance;
The vehicular compound eye distance measuring device according to claim 1, comprising: The measurement distance calibration means detects the compound eye distance calculated by the compound eye distance calculation means, which is a distance between the host vehicle and another vehicle whose license plate image has been taken by the reference distance calculation means. A detection unit;
And the reference distance, from the comparison between the compound eye calculated distance calculated by the compound-eye distance calculating means, the calibration coefficient calculation unit for calculating the calibration coefficients,
A compound-eye distance calibration unit for calibrating the distance between the calibration by the calibration coefficients calculated by the coefficient calculation unit, the vehicle and the object calculated by the compound-eye distance calculating means,
The vehicular compound eye distance measuring device according to claim 1, further comprising:   The compound eye distance calculating means includes a parallax calculation unit that calculates parallax of the two or more cameras, and a distance calculation unit that calculates a distance between the host vehicle and the object based on the parallax calculated by the parallax calculation unit; The vehicular compound eye distance measuring device according to any one of claims 1 to 3.   5. The camera according to claim 1, wherein a camera that captures an image in the reference distance calculator is one of two or more cameras that capture an object in the compound eye distance calculator. The compound eye distance measuring device for vehicles as described in the item.   The vehicle according to any one of claims 1 to 5, wherein in the reference distance calculation means, the license plate image for calculating the reference distance is the largest license plate image in the photographed image. Compound eye distance measuring device. In the reference distance calculation means, when the license plate image in the photographed image cannot be detected or is less than a predetermined size, the measurement distance calibration means sets the calibration coefficient to 1 or calibration in the previous distance measurement. It is set as a coefficient , The compound eye distance measuring device for vehicles according to any one of claims 1 to 6 characterized by things. A license plate image detection step for detecting the size of an image of a license plate attached to another vehicle in the vicinity of the host vehicle (referred to as a license plate image) taken by a camera mounted on the host vehicle;
The distance between the host vehicle and the other vehicle is calculated using the relationship between the distance between the host vehicle and the other vehicle registered in advance and the size of the license plate image detected in the license plate image detection step. A reference distance calculation step to be a reference distance,
The distance between the vehicle and the object is calculated based on the principle of triangulation from the imaging of the object in the vicinity of the vehicle by two or more cameras mounted on the vehicle at a predetermined interval. A compound eye distance calculating step;
The host vehicle and the license plate calculated from the reference distance and the imaging of the license plate taken by the two or more cameras simultaneously with the imaging of the license plate image in the reference distance calculating step in the compound eye distance calculating step A calibration coefficient is calculated from the ratio of the distance to the distance, and the distance between the subject vehicle and the object calculated in the compound eye distance calculation process using the calibration coefficient is calibrated using the reference distance, ,
A compound eye distance measuring method comprising:

Images