JP2958043B2 - Mobile vehicle environment recognition device - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to an environment recognition apparatus for a mobile vehicle, for example, for recognizing and capturing a vehicle that may collide with the own vehicle across the front, and In particular, it relates to improving the efficiency and simplification of the recognition.

(Prior Art) A distance sensor such as an ultrasonic sensor or a laser radar is used for obstacle recognition in conventional traveling control of an autonomous vehicle. These sensors generally have the disadvantage that they can measure distances in a particular direction, but not over a wide range. In order to complement such a distance sensor, an obstacle recognition device equipped with a stereo vision camera and performing image recognition by inputting an image, for example,
This is proposed in Japanese Patent Application Laid-Open No. 64-26913. In this Japanese Unexamined Patent Publication (Kokai), two TV cameras are used to measure the distance of a wide field of view based on the principle of triangulation.

(Problems to be Solved by the Invention) However, in the image processing based on stereo vision, enormous processing is required for matching left and right images, and this is an obstacle to practical application when real-time environment recognition is performed. . In particular, it is not easy to predict the possibility of collision with the captured mobile.

For example, when performing a collision prediction based on image recognition, it is necessary to identify a moving object in an image and then calculate a moving direction and a route to the own vehicle, and perform calculations for all objects. It requires a huge amount of computation.

Therefore, the present invention has been proposed to eliminate the above-mentioned disadvantages of the conventional example, and an object of the present invention is to provide an environment recognizing device for a mobile vehicle capable of efficiently and easily recognizing a mobile object having a collision possibility. There is a suggestion.

(Means and Actions for Achieving the Object) Environment for Recognizing and Capturing a Moving Body Moving in a Second Direction Intersecting with a First Direction in which the Vehicle Moves of the Present Invention for Achieving the Object An image input unit configured to input image information of the external world in the first direction by a predetermined width along a predetermined horizontal line direction; a detection unit configured to detect a temporal change in the input external world image information; A recognition unit for recognizing a possibility of collision between the moving object and the own vehicle based on a time change.

When the moving body moving in the second direction has a possibility of colliding with the own vehicle, a certain characteristic appears in the time change.
With this feature, the possibility of collision can be predicted.

(Embodiment) A preferred embodiment of an environment recognition device of the present invention will be described below with reference to the accompanying drawings.

FIG. 2 illustrates the processing area 6 of the image processing used in the present embodiment. That is, FIG. 2 shows an image obtained by the camera of the own vehicle. In FIG. 2, reference numeral 6 denotes a processing area, which is not an external image.

In FIG. 2, reference numeral 1 denotes a traveling path of the vehicle, 2, 7 denotes a road end, 3 denotes a center line, and 4 denotes a vanishing point. Reference numeral 5 denotes another vehicle appearing from the right side of a road orthogonal to the traveling direction of the own vehicle. In this embodiment, the processing area is set to the right half on the horizontal line including the image vanishing point. In this embodiment, in order to simplify the description,
Although the processing area 6 is set to the right half on the horizontal line, it is needless to say that the processing area is set to the right side or the left side, and may be set to both sides.

FIG. 3 illustrates the relationship between the camera 10 of the vehicle and the processing area 6. The processing area 6 is composed of a total of n one-dimensionally arranged pixel blocks, and these pixels are represented by A ₁ , A ₂ , ‥‥‥, _An
Call.

The principle of predicting the possibility of collision in the present embodiment will be described with reference to FIGS. 4A and 4B. If the objects in the image are all stationary objects, the temporal change of each pixel obtained with the movement of the vehicle has a movement vector in the direction away from the vanishing point as shown in FIG. 4A. If there is a moving body in this, various moving vectors are shown according to the moving state. When this is considered only with one-dimensional information on the right side on the horizon including the vanishing point as in this embodiment, the vector of the fixed object always indicates rightward movement, and the vector of the moving object always indicates the moving speed direction. Indicates rightward movement, stillness, and leftward movement.

Figure 4B illustrates the vehicle 11 traveling from point Y ₁ along the Y-axis direction to Y _2, and another vehicle 5 traveling from point X ₁ to the point X ₂ along the X-axis direction. Other car 5 on X ₁ when own car 11 is on Y ₁
X ₂ when but visible in the direction of the angle alpha _1, the vehicle 11 is advanced to Y ₂
Other vehicles 5 it is assumed that appeared to angle α _2. If this other car 5
If he is likely to collide with his own vehicle, α ₁ ≒ α ₂ should hold. If α ₁ > α ₂ , the other vehicle will pass this zero point before the vehicle approaches the zero point. If α ₁ <α ₂ , the own vehicle will pass through the zero point before the other vehicle approaches the zero point. This embodiment utilizes this property.

As shown in FIG. 3, since the processing region 6 is set include an image vanishing point, each pixel of the A ₁ to A _n corresponds to the viewing angle from the camera 10. Therefore, if the image data of the A ₁ to A _n in which the vehicle has obtained when there at _one point Y is as shown in Figure 5A, i.e., data of pixels corresponding to the other vehicle is A _p ~ A _{p + 4}
If the other vehicle position shows a stable value and it seems that it is changing finely at other pixel positions, the pixel of the area 6 obtained by the own vehicle camera 10 at the time when the other vehicle moves as shown in FIG. 4B The data would look like Figure 5B. In other words, from Y ₁ point
When it proceeds to Y _2, the pattern of the image data in the region of 20
The pattern changes to a 20 'pattern, and the change at each pixel position is random. This means that regions 22 and 2
The same can be said for 2 '. Therefore, taking the difference between the corresponding pixels of the regions 20 and 20 'and taking the difference between the corresponding pixels of 22 and 22',
Large random changes will be obtained. Because,
This is because these areas are images of still objects. However, pixel areas 21 and 2 corresponding to other vehicles where a collision is expected
There is no significant change for 1 '. Therefore, when the difference is calculated between the corresponding pixels, the difference approaches “0”.

FIG. 1 is a conceptual configuration diagram of an embodiment apparatus configured according to this principle. The differential calculator 30 calculates the difference between the image data at time t and the image data at time t−1 for each pixel position. Reference numeral 31 denotes a portion for determining whether the difference value has a value equal to or greater than a predetermined value for each pixel, that is, whether the image data has changed to some extent. In Figure 5, 20 to 2
Since the area that changes like 0 'has changed to some extent, the output from this determination unit is "0". On the other hand, the judgment corresponding to the area 21 in FIG. 5 is “1”. In FIG. 1, “1” is output from the determination unit 31 for the pixel positions of A _{p to} A _{p + 4} .

FIG. 6 shows a circuit configuration in which the apparatus of FIG. 1 is further embodied. The differential operation unit 30 shown in FIG. 1 includes two m-bit shift registers 40 and 41 and an m-bit subtractor 42. It is assumed that image input is performed every T seconds. The shift register 41 stores image data of one pixel (assumed to have m-bit gradation) a predetermined time before (time t-1). The shift register 40 stores the current (time t) image data. The contents of the shift registers 40 and 41 are input to the subtracter 42 at regular intervals (every T seconds). On the other hand, every T / n seconds, the shift register 40 stores the pixel A _{i + 1}
The data of (t) is stored in the shift register 41 in the pixel A _{i + 1} (t
+1) is input.

The decision section 31 shown in FIG. 1 comprises an m-bit digital comparator 43, a shift register 45, and an AND gate 46 in FIG. The comparator 43 calculates the difference (A _i
(T) −A _i (t−1)) is compared with a predetermined threshold value. This determination result is a determination result for the i-th pixel position, and is shifted into the shift register 45. That is, the 3-bit shift register 45 holds the determination results for three pixels. In the example of FIG. 6, if the determination result for three pixels is "1", an output for determining that there is a possibility of collision is output to the gate 46.
Output from Of course, if the number of bits of the shift register 45 is increased, the one determined as “1” in a larger range will be determined as an obstacle.

As described above, a moving object that is likely to be an obstacle can be quickly and easily found. Incidentally, the position of the obstacle is determined by the determiner 31.
Can be determined by the pixel position where “1” is output from the pixel. Further, the length of the AND gate 45 may be variously changed to reduce erroneous determination.

The present invention can be variously modified without departing from the gist thereof.

This is because in the above embodiment, one camera is used and the possibility of collision in a horizontal plane including the camera is determined. Therefore, if a plurality of cameras mounted at different heights are installed and images are obtained from these cameras, prediction accuracy is improved.

As can be easily understood from the description of the above embodiment, since a one-dimensional image is sufficient for predicting a collision, a camera to be used does not need to have an expensive area sensor, and a line sensor is sufficient. .

(Effects of the Invention) As described above, the environment recognition device for recognizing and capturing a moving body that moves in a second direction that intersects the first direction in which the own vehicle moves according to the present invention is the first environment recognition apparatus. Input means for inputting image information of the external world in a predetermined direction along a predetermined horizontal line direction, detecting means for detecting a time change of the input external world image information, and performing the movement based on the time change. A recognition means for recognizing a possibility of collision between the body and the own vehicle is provided.

That is, when the moving body moving in the second direction may collide with the own vehicle, a certain characteristic appears in the time change. With this feature, the possibility of collision can be predicted. In addition, since high-level recognition processing is not required, the prediction is performed at high speed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of a preferred embodiment to which the present invention is applied, FIG. 2 is a diagram for explaining a processing region, FIG. 3 is a diagram for explaining a processing region for a camera, and FIG. FIG. 4B, FIG. 5A, and FIG. 5B are diagrams illustrating the prediction principle according to the present embodiment, and FIG. 6 is a diagram illustrating the prediction principle according to the present embodiment. It is a concrete circuit diagram. In the figure, 1 ... travel road, 2,7 ... road end, 3 ... center line, 4
...... Image vanishing point, 5 ... Obstacle (other car), 6 ... Target area, 10 ... Camera, 11 ... Own car, 20,20,20 ', 22,22' ...
... image area of a stationary object, 21, 21 '... image area of an obstacle, 3
0: Differential operation unit, 31: Change judgment unit, 40, 41, 45 ... Shift register, 42: Subtractor, 43: Comparator,
46 …… AND gate.

Claims (3)

(57) [Claims] A second direction intersecting a first direction in which the vehicle moves;
An environment recognition device for recognizing and capturing a moving object moving in a direction of: an image input means for inputting image information of the external world in the first direction by a predetermined width along a predetermined horizontal line direction; An environment recognition system for a mobile vehicle, comprising: detection means for detecting a time change of external image information; and recognition means for recognizing a possibility of collision between the moving body and the own vehicle based on the time change. apparatus. 2. The movement according to claim 1, wherein said image input means sets a target area of the input image to an area including a vanishing point and having the predetermined width in a horizontal line direction from the vanishing point. Car environment recognition device. A detecting means for detecting a moving object in the image information having the predetermined width, wherein a direction connecting the position of the vehicle and the recognized moving object is substantially constant for a predetermined period; The environment recognition apparatus for a mobile vehicle according to claim 2, wherein in a certain case, the mobile object is determined to have a possibility of collision.