JP2009258015A - Apparatus and method for generating range image data for vehicle - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an apparatus and method for generating range image data for vehicles, capable of continuously grasping a situation ahead of own vehicle. <P>SOLUTION: The apparatus includes a projector 5, image intensifier 7b, a high-speed camera 8, a timing controller 9 for controlling the image intensifier and the high-speed camera, and an image processing section 10 which generates range image data representing a distance to an object per pixel, based on brightness of an identical pixel in a plurality of pick-up images with different target distances obtained from the image intensifier 7b and the high-speed camera 8. The image processing section 10 is always luminescent and provided with a processing step S5 for eliminating pixel data existing in all pick-up images as noise. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention belongs to the technical field of a vehicle distance image data generation device.

In Patent Literature 1, whether or not an object such as an obstacle exists at the target distance is based on an image that is projected in accordance with the timing of reflected light that is projected in front of the host vehicle and returned from the target distance. Techniques for detection are disclosed.
US Pat. No. 6,732,123

  However, in the above prior art, objects other than the target distance cannot be detected. That is, there is a problem that the situation is intermittently grasped and the situation ahead of the host vehicle cannot be grasped continuously.

  An object of the present invention is to provide a vehicular distance image data generation apparatus and a vehicular distance image data generation method capable of continuously grasping the situation ahead of the host vehicle.

In order to achieve the above object, in the vehicle distance image data generation device of the present invention,
A projection unit that projects pulsed light in front of the host vehicle at a predetermined period, an imaging unit that captures reflected light returning from the target distance at an imaging timing set according to the target distance, and the target distance are continuous. Represents the distance to the object for each pixel based on the luminance of the same pixel in a plurality of captured images with different target distances obtained by the imaging means, and a timing control means for controlling the imaging timing so as to change A distance image data generating means for generating distance image data, and the distance image data generating means is a noise removing means for removing pixel data existing in all captured images as noise because it always emits light. It is characterized by having.

  Therefore, in the present invention, since the distance image data representing the distance to the object for each pixel is generated based on the luminance of the same pixel in the plurality of captured images having different target distances, the situation in front of the host vehicle is continuously displayed. Can be grasped.

  The best mode for realizing the vehicle distance image data generating apparatus and the vehicle distance image data generating method of the present invention will be described below with reference to the drawings.

First, the configuration will be described.
FIG. 1 is a block diagram illustrating a configuration of an obstacle detection apparatus 1 according to a first embodiment to which the vehicle distance image data generation apparatus according to the present invention is applied. The obstacle detection apparatus 1 according to the first embodiment generates distance image data. A device 2, an object recognition processing unit 3, and a determination unit 4 are provided.

The distance image data generation device 2 includes a projector (projecting unit) 5, an objective lens 6, a light multiplying unit 7, a high-speed camera (imaging unit) 8, a timing controller (timing control unit) 9, and an image. And a processing unit (distance image data generating means) 10.
The projector 5 is a near-infrared LED disposed at the front end of the vehicle, and outputs pulsed light for a predetermined light projecting time tL (for example, 5 ns) in accordance with the pulse signal output from the timing controller 9. The cycle of the pulse signal is the projection cycle tP of the projector 5, and the projection cycle tP is, for example, an interval of 1/100 s or less.
The objective lens 6 is for receiving reflected light from an object, and is disposed adjacent to the projector 5. For example, the optical system is set to have an angle of view capable of capturing a predetermined range in front of the host vehicle.

The light multiplication unit 7 includes a gate 7a and an image intensifier 7b.
The gate 7a opens and closes in response to an opening / closing command signal from the timing controller 9. Here, in Example 1, the opening time (gate time) tG of the gate 7a is set to 5 ns, which is the same as the light projection time tL. Here, the gate time tG corresponds to the imaging target width of the imaging area (target distance). The longer the gate time tG, the longer the imaging target width of the imaging area. In the first embodiment, since the gate time tG = 5 ns, the imaging target width is 1.5 m from the speed of light (about 3 × 10 ⁸ m / s) × gate time (5 ns).

The image intensifier 7b temporarily converts extremely weak light (reflected light from an object, etc.) into electrons and then electrically amplifies it, and returns it to a fluorescent image again, thereby doubling the amount of light and producing an image with contrast. It is a viewing device. Photoelectrons launched from the photocathode of the image intensifier 7b by a photoelectric phenomenon are accelerated at a high voltage of the order of kV, and are emitted into the phosphor screen on the anode side, thereby emitting fluorescence having a photon number of 100 times or more. The fluorescence generated on the phosphor screen is guided to the image sensor of the high-speed camera 8 by the fiber optic plate without being scattered while maintaining the same positional relationship.
The high-speed camera 8 captures an image emitted from the light doubling unit 7 in response to a command signal from the timing controller 9 and outputs a captured image (color image) to the image processing unit 10. In the first embodiment, a camera having a resolution of 640 × 480 (horizontal: vertical), a luminance value of 1 to 255 (256 levels), and 100 fps or more is used.

The timing controller 9 is the time from the light projection start time of the projector 5 until the gate 7a is opened so that the captured image captured by the high-speed camera 8 is the timing of the reflected light returning from the target imaging area. The imaging timing is controlled by setting a certain delay time tD and outputting an open / close command signal corresponding to the delay time. That is, the delay time tD is a value that determines the distance from the host vehicle to the imaging area (imaging target distance), and the relationship between the delay time tD and the imaging target distance is as follows.
Imaging distance = speed of light (approx. 3 x 10 ⁸ m / s) x delay time tD / 2
FIG. 2 shows a temporal relationship between the operation of the projector 5 (light projection operation) and the operation of the gate 7a (camera gate operation) when imaging one imaging area.

  The timing controller 9 increases the imaging range of the high-speed camera 8 by increasing the delay time tD by a predetermined interval (for example, 10 ns) so that the imaging area continuously moves from the front side of the vehicle to the front side. Change to the side. The timing controller 9 starts the imaging operation of the high-speed camera 8 immediately before the gate 7a is opened, and ends the imaging operation after the gate 7a is completely closed.

  Further, in the first embodiment, as illustrated in FIG. 3, when the imaging target distance is continuously changed from B1 → B2 → B3 →..., The imaging target distance is increased more than the imaging target width A of the imaging area. By increasing the amount (B2-B1), the increase amount of the imaging target distance is set so that a part of the imaging area changes while overlapping.

  FIG. 4 is a schematic diagram showing temporal luminance changes when the amount of increase in the imaging target distance is reduced to the limit, in other words, when imaging is performed with an infinite increase in the imaging area. By overlapping each other, the luminance value of the same pixel in a plurality of consecutive captured images gradually increases, and after the peak, the luminance value gradually decreases. In practice, the number of imaging areas is limited (1 to n), but by overlapping a part of continuous imaging areas, the temporal luminance change becomes close to the characteristics of FIG.

  The timing controller 9 outputs an image processing command signal to the image processing unit 10 when all the captured images for one frame, that is, the set predetermined range (area 1, area 2,..., Area n) are captured. To do.

  The image processing unit 10 generates distance image data representing the distance to the object for each pixel based on the luminance of the same pixel in the captured image for one frame captured by the high-speed camera 8, and the generated distance image data Is output to the object recognition processing unit 3.

The object recognition processing unit 3 identifies an object included in the distance image data. A well-known technique such as pattern matching can be used as an object identification method.
Based on the relationship (distance, relative speed, etc.) between the object (person, car, sign, etc.) identified by the object recognition processing unit 3 and the host vehicle, the determination unit 4 presents information to the driver by an alarm, Determine whether vehicle control such as automatic braking is necessary.

[Distance image data generation control processing]
FIG. 5 is a flowchart illustrating the flow of the distance image data generation control process executed by the image processing unit 10 according to the first embodiment. Each step will be described below. This process is repeatedly executed at a predetermined calculation cycle.

  In step S1, luminance value data having the lowest luminance of the captured image obtained by imaging the front of the host vehicle without light projection by the projector 5 is stored as minimum value data, and the process proceeds to step S2.

  In step S2, a captured image is input, and the process proceeds to step S3.

  In step S3, the minimum value data stored in step S1 is compared with the lowest luminance value data among the luminance value data of each pixel of the captured image input in step S2, and the luminance value data is more than the minimum value data. Determine whether it is low. If YES, the process proceeds to step S4. If NO, the process proceeds to step S5.

  In step S4, luminance value data lower than the minimum value data is stored as new minimum value data, and the process proceeds to step S5.

  In step S5, processing for excluding the corresponding pixel data as noise data is performed depending on whether the updated and stored minimum value data exceeds a preset threshold value.

  In step S6, the difference between the luminance value data and the minimum value data is obtained for each pixel, and the current data in the pixel is obtained, and the process proceeds to step S7.

  In step S7, the difference between the previous data (previous control cycle) and the current data is calculated for each pixel, and the process proceeds to step S8.

  In step S8, it is determined whether or not the difference obtained in step S7 is equal to or smaller than a threshold value (noise determination threshold value). If YES, the process proceeds to step S10. If NO, the process proceeds to step S9. In the first embodiment, the threshold value is set to 32.

  In step S9, the pixel of the data whose difference with the immediately preceding data exceeds the threshold is excluded (for example, the luminance value is 0), and the process proceeds to step S10.

  In step S10, it is determined whether image input for one frame (area 1, area 2,..., Area n) has been completed. If YES, the process proceeds to step S11. If NO, the process proceeds to step S2.

  In step S11, the brightness value data of the same pixel of the n captured images is compared, distance image data is generated from the frame number (delay time tD) of the captured image corresponding to the highest brightness value data, and the process returns. Transition.

Next, the operation will be described.
[Distance image data generation function]
The timing controller 9 controls the imaging timing of the high-speed camera 8 by setting the delay time tD so that the captured image captured by the high-speed camera 8 becomes the timing of the reflected light returning from the target imaging area. To do. When an object is present in the target imaging area, the time for the light emitted from the projector 5 to return from the imaging area is the time for the light to reciprocate the distance between the host vehicle and the imaging area (imaging target distance). Therefore, the delay time tD can be obtained from the imaging target distance and the speed of light.

  In the captured image of the high-speed camera 8 obtained by the above method, when an object exists in the imaging area, the luminance value data of the pixel corresponding to the position of the object is affected by the reflected light, and the luminance of other pixels Indicates a value higher than the value data. Thereby, based on the luminance value data of each pixel, the distance from the object existing in the targeted imaging area can be obtained.

  Furthermore, in the first embodiment, captured images of the imaging areas 1 to n are acquired while changing the delay time tD. Subsequently, the brightness value data of the same pixel in each captured image is compared, and the distance data (distance image data) of all the pixels (640 × 480) is generated using the highest brightness value data as the distance of the pixel.

  Conventional distance detection methods using laser radars and stereo cameras are susceptible to rain, fog, snow, etc., and the noise level with respect to the signal level is large (the SN ratio is small), so the reliability in bad weather is low. . When using a millimeter wave radar that is not easily affected by bad weather, the reliability of distance detection is high, but it is difficult to perform object recognition (object identification) from the millimeter wave radar signal. Is required. Further, since the camera image becomes unclear during bad weather, it is difficult to perform accurate object recognition.

On the other hand, in the first embodiment, only reflected waves returning from the target imaging area are reflected in the captured image, so that the light bent due to the influence of rain, fog, snow, or the like, that is, the noise mixing level is kept low. High signal-to-noise ratio can be obtained. That is, high distance detection accuracy can be obtained regardless of bad weather or night.
Since the distances of all the pixels are known from the generated distance image data, the distance to the object can be instantly grasped when performing object recognition using a technique such as pattern matching.

  Furthermore, in the first embodiment, the captured area is continuously changed to acquire a plurality of captured images, and the captured images are compared to detect the distance of each pixel. In addition, it can be grasped over a wide range. For example, even in the situation where a pedestrian has jumped between the host vehicle and the preceding vehicle, the distance between the preceding vehicle and the pedestrian can be grasped at the same time. Control can be performed.

FIG. 6 shows a situation in which four pedestrians A to D exist at different positions in front of the host vehicle, and the relationship between the distance between the host vehicle and each pedestrian is A <B <C <D. .
At this time, in Example 1, a part of the imaging area is overlapped so that the reflected light from one object is reflected in the pixels of the captured image in a plurality of continuous imaging areas. For this reason, the temporal luminance change of the pixel corresponding to each pedestrian shows a triangular characteristic that takes a peak at the position of the pedestrian, as shown in FIG.

Since the distance image data is data used for alarms and vehicle control, it is impossible to set an imaging area infinitely finely as long as a certain calculation speed is required. By making the reflected light from the plurality of captured images included, as shown in FIG. 8, the temporal luminance change of the pixel is approximated to the above characteristics, and the imaging area corresponding to the peak of the triangular portion is By setting the distance of the object in the pixel, detection accuracy can be increased.
As an example of the distance image, FIG. 11 and FIG. 12 show the color dot display in which the distance to the object in each pixel is related to red and near to blue and the color wavelength.

[Noise elimination]
In the first embodiment, an object having an imaging target distance is grasped by obtaining a bright image portion. And what is in the distance with the own vehicle other than imaging distance is processed by making it the image part without a brightness | luminance.
For this reason, if there is light from the outside, such as outdoor lights, car headlights, brake lamps, and building windows, unless they are excluded, the object is captured at each distance. Is determined to exist.

In the first exemplary embodiment, pixel data that captures an image that is constantly emitting light is excluded as noise by the processing in step S5 performed by the image processing unit 10.
Specifically, the brightness determined by the processing of steps S3 and S4 performed first to process the brightness data of the pixel (same pixel) at the same XY position (position determined by the vertical and horizontal coordinates) of the image data of each distance. The first value data of the minimum value is used. This minimum value data is the minimum luminance value in the same pixel at the same position in the pixel data of each distance.
Therefore, if the threshold value set in advance is compared with the minimum value data, a pixel that captures an image that is constantly emitting light has a high minimum luminance value, and is determined by exceeding the threshold value. Then, by being excluded as noise, the distance image data generated thereafter becomes more accurate. In addition, even when a process for specifying an object is performed using this distance image, the object can be specified with higher accuracy.
For example, if the luminance value data of one pixel has a resolution of 0 to 256, the minimum luminance value data is excluded when it exceeds 48 thresholds.

FIG. 9 is an image when the front of the host vehicle is captured by a general camera. In this situation, when the method of the first embodiment is used, captured images in the respective imaging areas are illustrated in FIGS. 10 (a) to 10 (j). It becomes. Here, in FIG. 10A to FIG.
Thus, the part which always emits light exists as a part with high brightness | luminance in each image of Fig.10 (a)-(j).

In the distance image data generated based on the images (FIGS. 10 (a) to 10 (j)) obtained by capturing the always light emitting portion, the always light emitting portion is actually always emitting light. A distance that is different from the imaging target distance in which is present is displayed (see the portion P in FIG. 11).
On the other hand, the noise removal of Example 1 is performed, and the distance data of the constant light emission part is deleted as indicated by the symbol P in FIG. 12, thereby improving the accuracy of the distance image data. Improve the accuracy of processing to identify the object.
9 to 12 are gray scales for convenience, they are actually color images.

Next, the effect will be described.
The distance image data generation device 2 according to the first embodiment has the following effects.

  (1) A projector 5 that projects pulsed light in front of the host vehicle at a predetermined period, an image intensifier 7b that captures reflected light returning from the target distance at an imaging timing set according to the target distance, and a high speed The same pixel in a plurality of captured images with different target distances obtained by the camera 8, the timing controller 9 that controls the imaging timing so that the target distance changes continuously, the image intensifier 7b, and the high-speed camera 8 An image processing unit 10 that generates distance image data representing a distance to an object for each pixel based on luminance is provided. Since the image processing unit 10 always emits light, pixel data present in all captured images is obtained. Since the process of step S5 to remove as noise is provided, the situation in front of the host vehicle can be grasped continuously, The light noise can be removed to improve the accuracy of the distance image data, which can contribute to improving the accuracy of the subsequent object specifying process.

  (2) In (1), the process of step S5 performed by the image processing unit 10 selects the minimum luminance value for the luminance of the same pixel in a plurality of captured images with different target distances, and sets the minimum luminance value in advance. If the threshold is exceeded, data is removed as noise, so that always emitting light is picked up with high brightness on the same pixel of a plurality of captured images. By excluding it as noise, it is possible to remove the noise that is always emitted and improve the accuracy of the distance image data, thereby contributing to the accuracy improvement of the subsequent object specifying process.

(Other examples)
The best mode for carrying out the present invention has been described above based on the embodiment. However, the specific configuration of the present invention is not limited to the configuration of the embodiment, and it relates to each claim of the claims. Design changes and additions are allowed without departing from the spirit of the invention.

  For example, the light projection period, the light projection time, the gate time, the imaging target width, the amount of change in the imaging target distance, and the number of imaging areas in one frame are appropriately set according to the performance of the imaging means and the performance of the distance image data generation means can do.

It is a block diagram which shows the structure of the obstruction detection apparatus 1 of Example 1. FIG. It is a figure which shows the temporal relationship between the operation | movement (light projection operation | movement) of the light projector 5, and the operation | movement (camera gate operation | movement) of the gate 7a at the time of imaging one imaging area. It is a figure which shows the state which a part of imaging area overlaps. It is a schematic diagram which shows a temporal luminance change at the time of imaging by increasing an imaging area infinitely. 3 is a flowchart illustrating a flow of distance image data generation control processing executed by the image processing unit 10 according to the first embodiment. It is a figure which shows the condition where four pedestrians AD exist in the different position ahead of the own vehicle. It is a schematic diagram which shows the temporal luminance change of the pixel corresponding to each pedestrian A-B. It is a figure which shows the distance image data generation effect | action of Example 1. FIG. It is an image when the front of the host vehicle is imaged with a general camera. It is a figure which shows the captured image of each imaging area. It is distance image data in the case of not performing noise removal of Example 1. It is distance image data which shows the noise removal effect | action of Example 1. FIG.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Obstacle detection apparatus 2 Distance image data generation apparatus 3 Object recognition process part 4 Judgment part 5 Light projector (light projection means)
6 Objective lens 7 Photomultiplier 7a Gate 7b Image intensifier 8 High-speed camera (imaging means)
9 Timing controller (timing control means)
10 Image processing unit (distance image data generating means)
P (Indicates the constant light emission location of distance image data)

Claims (2)

Light projecting means for projecting pulsed light in front of the host vehicle at a predetermined period;
Imaging means for imaging reflected light returning from the target distance at an imaging timing set according to the target distance;
Timing control means for controlling the imaging timing so that the target distance changes continuously;
Distance image data generation means for generating distance image data representing a distance to an object for each pixel based on the luminance of the same pixel in a plurality of captured images with different target distances obtained by the imaging means;
With
The distance image data generation means includes
Since it always emits light, it has a noise removal means to remove pixel data present in all captured images as noise,
A vehicular distance image data generating apparatus characterized by the above. The vehicle distance image data generation device according to claim 1,
The noise removing means is
Select the minimum luminance value for the luminance of the same pixel in a plurality of captured images with different target distances, and if the minimum luminance value exceeds a preset threshold, remove data as noise,
A vehicular distance image data generating apparatus characterized by the above.