JP7193942B2 - vehicle detector - Google Patents

Description

The present invention relates to vehicle detection devices.

Conventionally, there has been a demand for a technique for detecting vehicles approaching one's own vehicle. Against this background, pattern recognition is sometimes applied as a technique for detecting approaching vehicles. As an example of this pattern recognition, an image of an approaching vehicle viewed from the front is stored in advance as a feature, and it is determined whether or not there is a part that matches the feature in the camera image captured by the vehicle-mounted camera. There is

In the pattern recognition described above, recognition of whether or not the approaching vehicle is an approaching vehicle is performed using only an image of the approaching vehicle viewed from the front. Therefore, when the orientation of the approaching vehicle changes, it may not be possible to detect the approaching vehicle.

On the other hand, there has been proposed a technique of recognizing an approaching vehicle by combining pattern recognition with another image processing (see, for example, Patent Document 1). In this technique, a first area where the movement of the approaching vehicle is large relative to the background and a second area where the movement of the approaching vehicle is small relative to the background are set in the camera image. In the first area, an approaching vehicle is detected based on movement of feature points such as edges. In the second area, an approaching vehicle is detected by pattern recognition.

JP 2009-181557 A

In the technique disclosed in Patent Document 1, an approaching vehicle is detected using the fact that feature points move over time. In order to detect an approaching vehicle correctly, it is desirable to distinguish and classify the feature points of the approaching vehicle from the feature points of the background. Then, not only the feature points of the approaching vehicle but also the feature points of the static background must be extracted from the camera image. Therefore, there is a problem that the number of feature points increases and the processing load required of the system increases.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a vehicle detection device capable of reducing the processing load by using a bird's-eye view image based on an image captured by an in-vehicle camera.

The vehicle detection device according to the present invention performs pattern recognition based on the characteristics of other vehicles on an image obtained by imaging the surroundings of the own vehicle traveling on the road surface with an imaging unit that images the surroundings of the vehicle. A first vehicle detection unit for detecting the other vehicle and a plurality of images captured at different times by the image capturing unit are captured, and the other vehicle is viewed from a virtual viewpoint above the own vehicle. Based on the correlation between a plurality of bird's-eye view images converted from the viewpoint in the above manner, the confidence levels of the second vehicle detection unit for detecting the other vehicle and the first vehicle detection unit exceed a predetermined threshold, and and controlling the first vehicle detection unit to detect the other vehicle when the confidence of the second vehicle detection unit is equal to or less than the threshold, and the confidence of the first vehicle detection unit is the threshold and a detection control unit that controls the second vehicle detection unit to detect the other vehicle when the reliability of the second vehicle detection unit exceeds the threshold. .

According to this invention, it is possible to reduce the processing load required of the system by using the bird's-eye view image based on the image captured by the vehicle-mounted camera.

2 is a hardware block diagram showing hardware elements forming a vehicle detection system to which the vehicle detection device of Example 1 is applied; FIG. FIG. 10 is an explanatory diagram showing features when another vehicle is viewed from the front; It is explanatory drawing which shows the scene which applies a vehicle detection apparatus to the road of left-hand traffic. It is a flowchart which shows the flow of a series of processes performed by a vehicle detection apparatus. It is a figure which shows typically an example of the image which the camera imaged. 4 is a flowchart showing a specific flow of vehicle detection processing; It is a figure which shows an example of a bird's-eye view image typically. It is a figure for demonstrating a vehicle body difference registration process. It is explanatory drawing which shows the state which each detection result of the 1st vehicle detection part and the 2nd vehicle detection part became equal. It is explanatory drawing which shows an example of certainty. 4 is a flowchart showing a specific flow of vehicle body difference registration processing; 4 is a flowchart showing a specific flow of vehicle body position calculation processing; FIG. 4 is an explanatory diagram showing a procedure for calculating the vehicle length of another vehicle; FIG. 5 is an explanatory diagram showing a modification of the vehicle detection device according to the first embodiment;

Hereinafter, specific embodiments of a vehicle detection device according to the present invention will be described with reference to the drawings.

(Description of the configuration of the first embodiment)
FIG. 1 is a hardware block diagram showing hardware elements forming a driving assistance system 10 to which a vehicle detection device 100 of Embodiment 1 is applied. First, a system configuration to which the vehicle detection device 100 is applied will be described with reference to FIG.

A driving support system 10 of the present embodiment is mounted on a host vehicle 1 (FIG. 3). The driving support system 10 has a front camera 20F (imaging section), a left side camera 20L (imaging section), and a right side camera 20R (imaging section). Hereinafter, the front camera 20F, the left side camera 20L, and the right side camera 20R will simply be referred to as cameras 20 when there is no need to distinguish them.

A front camera 20</b>F is mounted on the front bumper and front grill of the vehicle 1 so as to face the front of the vehicle 1 . A left side camera 20</b>L is mounted on the left door mirror of the vehicle 1 toward the left side of the vehicle 1 . A right side camera 20R is mounted on the right door mirror of the vehicle 1 so as to face the right side of the vehicle 1. - 特許庁

A camera equipped with a wide-angle fisheye camera is adopted as the camera 20 . A fisheye camera performs distortion correction based on a known distortion correction function in order to acquire a wide-angle image. An image captured by the camera 20 is output to the vehicle detection device 100 .

The driving support system 10 further includes a vehicle state detection unit 30, a vehicle speed control unit 40, a display control unit 50, an audio output control unit 60, a distance measurement sensor 70, and a vehicle detection device 100.

The vehicle state detection unit 30 outputs to the vehicle detection device 100 a signal indicating the vehicle state such as the shift lever state, wheel speed (vehicle speed), parking method (parallel parking or parallel parking).

The vehicle speed control unit 40 controls an engine (not shown) and a brake (not shown) of the own vehicle 1 to accelerate or decelerate the own vehicle 1 .

The display control unit 50 performs control to display the detection results of the first vehicle detection unit 101 and the second vehicle detection unit 102 on a display (not shown).

The voice output control section 60 outputs a voice warning based on the detection results of the first vehicle detection section 101 and the second vehicle detection section 102 .

The ranging sensor 70 is an ultrasonic sensor, a laser radar, or the like, and includes a front ranging sensor 70F, a left ranging sensor 70L, and a right ranging sensor 70R. Among the ranging sensors 70, the ranging sensor 70 facing forward is called a forward ranging sensor 70F. Among the ranging sensors 70, the ranging sensor 70 directed to the left side is referred to as a left side ranging sensor 70L. Among the ranging sensors 70, the ranging sensor 70 directed to the right side is referred to as a right ranging sensor 70R. Hereinafter, the front range sensor 70F, the left side range sensor 70L, and the right side range sensor 70R are simply referred to as the range sensor 70 when there is no need to distinguish them.

A forward ranging sensor 70F is mounted on the front portion of the vehicle 1 . A left side ranging sensor 70L is mounted on the left side of the vehicle 1 . A right side ranging sensor 70R is mounted on the right side of the vehicle 1 . The distance measurement sensor 70 sequentially irradiates the surroundings of the own vehicle 1 with measurement waves of a predetermined frequency based on an instruction from the vehicle detection device 100 . The ranging sensor 70 successively receives the reflected waves reflected by objects within the irradiation range of the measurement waves.

(Schematic configuration of vehicle detection device)
FIG. 1 shows a schematic configuration of a vehicle detection device 100 according to one embodiment of the invention. A schematic configuration of the vehicle detection device 100 will be described below with reference to FIG. 1 .

Vehicle detection device 100 includes first vehicle detection unit 101 , second vehicle detection unit 102 , detection control unit 103 , edge candidate determination unit 104 , edge determination unit 105 , and storage unit 106 .

Inside the vehicle detection device 100, for example, a microprocessor and programs including peripheral devices, a RAM (Random Access Memory) that performs necessary processing, a ROM (Read Only Memory), a dedicated ASIC that performs image processing and signal processing (Application Specific Integrated Circuit) and other modules are mounted.

The first vehicle detection unit 101 detects the other vehicle 2 by performing pattern recognition based on the characteristics of the other vehicle 2 ( FIG. 3 ) on the image obtained by imaging the surroundings of the own vehicle 1 with the camera 20 . do.

The second vehicle detection unit 102 converts each of the plurality of images captured by the camera 20 at different times so that the other vehicle 2 is viewed from a virtual viewpoint above the own vehicle 1. Another vehicle 2 is detected based on the correlation between a plurality of bird's-eye view images.

When the certainty of the first vehicle detection unit 101 exceeds a predetermined threshold and the certainty of the second vehicle detection unit 102 is equal to or less than the threshold, the detection control unit 103 controls the first vehicle detection unit 101 is controlled to detect another vehicle 2, and when the certainty of the first vehicle detection unit 101 is equal to or less than the threshold and the certainty of the second vehicle detection unit 102 exceeds the threshold, the second vehicle detection Control is performed to cause the unit 102 to detect the other vehicle 2 .

Details of the edge candidate determination unit 104, the edge determination unit 105, and the storage unit 106 will be described later.

(Flow of a series of processes performed by the vehicle detection device 100)
Next, the flow of a series of processes performed by the vehicle detection device 100 will be described using the explanatory diagram shown in FIG. 3 and the flowchart shown in FIG. FIG. 3 shows a scene in which the vehicle detection device 100 is applied to a left-hand traffic road.

As shown in FIG. 3, a series of processing is performed by applying the vehicle detection device 100 (FIG. 1) to a scene in which another vehicle 2 is approaching the own vehicle 1 from behind.

As the operation condition of the vehicle detection device 100 in this scene, it is assumed that the own vehicle 1 is parallel parked and starts in the direction indicated by the white arrow F1 at a low speed of 2 km/h or less, for example. there is In this scene, the speed of another vehicle approaching the own vehicle 1 from behind the own vehicle 1 in the direction indicated by the white arrow F2 is assumed to be, for example, 25 km/h or less.

The detection control unit 103 controls the first vehicle detection unit 101 to detect the other vehicle 2 when the distance from the own vehicle 1 to the other vehicle 2 exceeds a predetermined distance. is equal to or less than a predetermined distance, the second vehicle detection unit 102 is controlled to detect the other vehicle 2 .

(Description of processing performed by the first vehicle detection unit 101)
First, processing performed by the first vehicle detection unit 101 will be described. The first vehicle detection unit 101 determines whether or not the template of the other vehicle 2 matches the object included in the target image. For example, the first vehicle detection unit 101 uses a pattern matching method that compares the feature points of the template and the target image. A template used for the pattern matching method is stored in the storage unit 106 in advance.

In this embodiment, for example, the features when the other vehicle 2 is viewed from the front are left and right lights, a license plate, and a windshield as shown in FIG. 2 (FIG. 2). Templates containing these features are defined in advance and stored in storage unit 106 . Thus, it is determined whether or not the image captured by the camera 20 has a portion that matches the feature. As a result of the determination, if there is a part that matches the characteristics, the other vehicle 2 is detected.

In this manner, the detection operation of the first vehicle detection unit 101 is performed using the characteristics of the other vehicle 2 when viewed from the front. Therefore, if the orientation of the other vehicle 2 changes from front facing to side facing or rear facing, the first vehicle detection unit 101 cannot detect the other vehicle 2 accurately. However, if a template is prepared according to the ever-changing movement of the other vehicle 2, the amount of template information to be stored in the storage unit 106 increases.

Therefore, in this embodiment, the detection operation of the first vehicle detection unit 101 is performed only when the other vehicle 2 exists farther than the predetermined distance from the own vehicle 1 to the other vehicle 2 . Then, when the other vehicle 2 exists in the vicinity where the distance from the own vehicle 1 to the other vehicle 2 is equal to or less than a predetermined distance, the detection operation of the second vehicle detection unit 102 is performed. The second vehicle detection unit 102 detects the other vehicle 2 using the difference between the frames of the bird's-eye view image. Specifically, the other vehicle 2 is detected by focusing on the fact that there is no difference in the stationary object (background) between the frames, but there is a difference in the other vehicle 2 .

(Description of processing performed by the second vehicle detection unit 102)
The detection area of the second vehicle detection unit 102 in this scene is set to a range in which it is difficult for the first vehicle detection unit 101 to detect the other vehicle 2 . Specifically, the detection area of the second vehicle detection unit 102 is set to, for example, 6 m corresponding to the width of two lanes on the right side of the vehicle 1, and to the rear of the vehicle 1, for example, 8 to 15 m. set. Here, 15 m indicates the longest detection distance at which the recognition rate by pattern matching (first vehicle detection) is high (90% or more) based on the image taken by the camera, and vehicle detection by pattern matching starts from here. be done. In addition, 8 m indicates the longest detection distance at which the recognition rate of the detection method (second vehicle detection) using the bird's-eye difference image is high (approximately 70% to 85%), and vehicle detection using only the bird's-eye difference image starts from here. be done. Since the width of one lane is usually about 3 m, the detection range of the camera (=distance from the camera installation position) is set to 6 m for two lanes (3 m×2).

The series of processes shown in FIG. 4 is started, for example, when an input instructing the start of execution of the vehicle detection process is sent to the vehicle detection device 100 by the driver's switch operation.

(Step S1)
First, the first vehicle detection unit 101 captures an image I Obtain (t−Δt), I(t). FIG. 5 is a diagram schematically showing an example of image I. As shown in FIG. Each image I(t−Δt) and I(t) is stored in storage unit 106 . Hereinafter, images I(t−Δt) and I(t) will simply be referred to as image I when there is no need to distinguish between them.

(Step S2)
Next, the first vehicle detection unit 101 detects whether or not the template of the other vehicle 2 and the other vehicle 2 present in the image I match. If the two match, the first vehicle detection unit 101 detects that the other vehicle 2 is present in the image I, and the process proceeds to step S3. If the two do not match, the first vehicle detection unit 101 detects that the other vehicle 2 does not exist in the image I, and the processing performed by the vehicle detection device 100 ends.

(Step S3)
Next, the detection control unit 103 stores the detection processing result of the first vehicle detection unit 101 in the storage unit 106 .

(Step S4)
Next, the second vehicle detection unit 102 performs vehicle detection processing for detecting the other vehicle 2 . A detailed flow of this vehicle detection processing will be described later.

(Step S5)
Next, the detection control unit 103 stores the detection processing result of the second vehicle detection unit 102 in the storage unit 106 .

(Step S6)
Next, the detection control unit 103 detects another vehicle 2 based on the detection processing result of the first vehicle detection unit 101 and the detection processing result of the second vehicle detection unit 102 .

(Description of specific flow of vehicle detection processing)
Next, a specific flow of the vehicle detection process shown in step S4 of FIG. 4 will be described using the flowchart of FIG.

(Step S11)
First, the second vehicle detection unit 102 converts each of the images I(t−Δt) and I(t) from a virtual viewpoint above the own vehicle 1 to a bird's-eye view of the other vehicle 2, Bird's-eye view images BI(t−Δt) and BI(t) are generated. In the following description, bird's-eye view images BI(t−Δt) and BI(t) are simply referred to as bird's-eye view images BI when there is no need to distinguish between them.

FIG. 7 is a diagram schematically showing an example of the overhead image BI. As shown in FIG. 7, the image I (FIG. 5) is mapped so that the edges of the vertical portion of the vehicle radiate from the camera 20 . Using the bird's-eye view image BI makes it easier to detect the position of the other vehicle 2 . In addition, even if there are a plurality of vehicles other than the other vehicle 2, each vehicle can be easily distinguished.

(Step S12)
The second vehicle detection unit 102 aligns the stationary object (background) reflected in the bird's-eye view image BI(t−Δt) with the stationary object (background) reflected in the bird's-eye view image BI(t). After the alignment, the second vehicle detection unit 102 performs a subtraction process of subtracting the bird's-eye view image BI(t−Δt) from the bird's-eye view image BI(t) to generate a difference image DP.

(Step S13)
The second vehicle detection unit 102 initializes a plurality of rays RL radially extending from the right side camera 20R. Initialization means erasing a plurality of radiation lines RL in order to start the vehicle body difference registration process in step S14.

(Step S14)
FIG. 8 is a diagram for explaining the vehicle body difference registration process. As shown in FIG. 8, the second vehicle detection unit 102 captures the differential image DP with a plurality of rays RL extending in the radial direction from the right side camera 20R at an equal angle (for example, 3 degrees) around the position of the right side camera 20R. degrees). The radiation RL serves as an index indicating the verticality of the other vehicle 2 existing in the differential image DP with respect to the road surface. As shown in FIG. 8, the other vehicle 2 included in the bird's-eye view image BI is divided into six solid object regions by six rays RL.

(Step S15)
The edge candidate determination unit 104 performs vehicle body difference registration processing. A detailed flow of the vehicle body difference registration process will be described later.

(Step S16)
The edge candidate determination unit 104 performs vehicle position calculation processing. A detailed flow of the vehicle position calculation process will be described later.

(Step S17)
When the certainty of the first vehicle detection unit 101 exceeds a predetermined threshold and the certainty of the second vehicle detection unit 102 is equal to or less than the threshold, the detection control unit 103 controls the first vehicle detection unit 101 is controlled to detect another vehicle 2, and when the certainty of the first vehicle detection unit 101 is equal to or less than the threshold and the certainty of the second vehicle detection unit 102 exceeds the threshold, the second vehicle detection Control is performed to cause the unit 102 to detect the other vehicle 2 . The detection control unit 103 creates a circle CI (Fig. 9). Here, the combined area is a section for switching tracking of the other vehicle 2 approaching from a distance from the first vehicle detection unit 101 to the second vehicle detection unit 102 . FIG. 9 is an explanatory diagram showing a state in which detection results of the first vehicle detection unit and the second vehicle detection unit are equal. As shown in FIG. 10, the reliability (=detection reliability) of the detection result by the first vehicle detection unit 101 gradually decreases. Conversely, as shown in FIG. 10, the degree of certainty (=detection certainty) of the detection result by the second vehicle detection unit 102 increases. Then, in the vicinity of the point where the line is drawn in FIG. 9, the reliability of both detection results becomes equal. In this section, the position of the approaching other vehicle 2 is determined based on the detection results of both of them, for example, by performing weighted addition according to the degree of certainty.

The detection control unit 103 determines whether or not the detection reliability of the first vehicle detection unit 101 has become zero. When the detection reliability of the first vehicle detection unit 101 becomes 0, the other vehicle 2 is detected only by the second vehicle detection unit 102 .

(Step S18)
When the detection certainty of the first vehicle detection unit 101 becomes 0 in step S17 (YES in step S17), the difference between the bird's-eye view image BI(t−Δt) and the bird's-eye view image BI(t) is used. Then, the second vehicle detection unit 102 detects the other vehicle 2 . The detection control unit 103 newly registers the position of the other vehicle 2 calculated by the edge candidate determination unit 104 in step S16.

(Step S19)
The detection control unit 103 tracks the other vehicle 2 whose position is newly registered in step S18. This tracking is performed until the other vehicle 2 advances, for example, about 8m from the newly registered position.

(Step S20)
On the other hand, when the other vehicle 2 does not exist within the circle CI in step S17 (NO in step S17), the detection control unit 103 calculates the certainty factor f by the detection processing of the second vehicle detection unit 102 . FIG. 10 is an explanatory diagram showing an example of certainty.

(Step S21)
The detection control unit 103 determines whether or not the certainty factor f calculated in step S20 exceeds a predetermined threshold value H (FIG. 10). If the certainty f exceeds the threshold H, the process proceeds to step S18. On the other hand, if the certainty f is equal to or less than the threshold H, the process proceeds to step S22.

(Step S22)
The detection control unit 103 removes the position of the other vehicle 2 calculated by the edge candidate determination unit 104 in the vehicle position calculation process of step S16 as noise.

After that, the process returns to the main routine (FIG. 4).

(Description of the specific flow of vehicle body difference registration processing)
Next, a specific flow of the vehicle body difference registration process shown in step S15 of FIG. 6 will be described using the flowchart of FIG.

(Step S31)
The edge candidate determination unit 104 scans the difference image DP at equal angular intervals centering on the position of the right camera 20R, and determines whether pixels exceeding a first predetermined number are arranged along the radial direction from the right camera 20R. judge. Here, the first predetermined number is a preset value for distinguishing other vehicles from other objects, and is preferably set to 7 to 8, for example.

If pixels exceeding the first predetermined number are lined up in step S31 (YES in step S31), the pixels lined up in the radial direction are determined to be a vertical edge and overlapped with the other vehicle 2, and the process proceeds to step S32. do. At this time, as shown in FIG. 8, straight lines (radiation lines) are drawn in a plurality of azimuths centering on the right side camera 20R, and the intersections of the lines overlapped with the other vehicle 2 are marked.

On the other hand, if the first predetermined number or less of pixels are aligned in step S31 (NO in step S31), the pixels aligned in the radial direction are determined to be noise and removed in step S35.

(Step S32)
The edge candidate determination unit 104 scans the difference image DP at equal angular intervals around the position of the right camera 20R, and determines whether or not there is at least one pixel along the radial direction from the right camera 20R. . If the number of three-dimensional object regions divided by the radiation RL is, for example, six as shown in FIG. Execute the number of iterations.

If at least one pixel exists (YES in step S32), the process proceeds to step S33. If no pixel exists (NO in step S32), the vertical edge determined in step S31 is determined as noise and removed in step S35.

(Step S33)
The edge candidate determination unit 104 determines whether pixels exceeding a second predetermined number are arranged along the radial direction from the right side camera 20R. Here, the second predetermined number is, like the first predetermined number described above, a preset value for distinguishing other vehicles from other objects, and is set, for example, to 7 to 8. is preferred.

If pixels exceeding the second predetermined number are arranged in step S33 (YES in step S33), the pixels arranged in the radial direction are determined as a vertical edge, and the process proceeds to step S34. On the other hand, if the number of pixels equal to or less than the second predetermined number are lined up in step S33 (NO in step S33), the process proceeds to step S36.

(Step S34)
The edge candidate determination unit 104 stores the vertical edges determined to have pixels exceeding the first predetermined number in step S31 and the vertical edges determined to have pixels exceeding the second predetermined number in step S33 to the storage unit 106. memorize to

(Step S36)
The edge candidate determination unit 104 stores in the storage unit 106 the region formed by the pixels sandwiched between the two vertical edges in step S34 as the vehicle body edge.

After that, the process returns to the main routine (FIG. 6).

(Explanation of specific flow of vehicle position calculation processing)
Next, a specific flow of the vehicle position calculation process shown in step S16 of FIG. 6 will be described using the flowchart of FIG.

(Step S41)
The edge determination unit 105 scans a vertical edge within a predetermined angle and determines whether or not there is a vertical edge. Here, the predetermined angle is a preset value for distinguishing other vehicles from other objects, and is preferably set to 30 degrees, for example.

If the vertical edge exists within the predetermined angle in step S41 (YES in step S41), the process proceeds to step S42. On the other hand, if there is no vertical edge within the predetermined angle in step S41 (NO in step S41), the vertical edge stored in storage unit 106 in step S34 shown in FIG. The stored vehicle body edges are removed as noise in step S52.

(Step S42)
Edge determination unit 105 determines whether or not at least two vertical edges exist within a predetermined angle. If two vertical edges do not exist (NO in step S42), the process proceeds to step S43. On the other hand, if two vertical edges exist (YES in step S42), the process proceeds to step S52 after the vertical edges are removed in step S53.

(Step S43)
The edge determination unit 105 determines whether a vehicle body edge is included within a predetermined angle. If the vehicle body edge is not included (NO in step S43), the process proceeds to step S44. On the other hand, if the vehicle body edge is included (YES in step S43), the process proceeds to step S47.

(Step S44)
The edge determination unit 105 scans the radiation while moving, for example, counterclockwise within a predetermined angle.

(Step S45)
The edge determination unit 105 determines whether scanning of radiation has been completed. If the radiation scanning is completed (YES in step S45), the process proceeds to step S52. On the other hand, if the radiation scanning has not been completed (NO in step S45), the process proceeds to step S46.
(Step S46)
The edge determination unit 105 determines whether or not a vehicle body edge exists on the radial line. If a vehicle body edge exists (YES in step S46), the process proceeds to step S43. On the other hand, if the vehicle body edge does not exist (NO in step S46), the process proceeds to step S52.

(Step S47)
The edge determination unit 105 officially registers the vehicle body edge within the group. This grouping of vehicle body edges includes a method of classifying vehicle body edges that are close to each other into the same group using the position information of the vehicle body edges. There is also a method of grouping the body edges based on the characteristics of the body edges, ie the direction or strength of the body edges. Also, the vehicle body edges can be grouped according to the degree of matching with the geometric figure designated as the object of extraction of the vehicle body edges.

(Step S48)
The edge determination unit 105 calculates the vehicle length of the other vehicle 2 using a well-known method. The vehicle length is, for example, the length of the shortest line connecting the tops of the closest edges (FIG. 13) of the vehicle body edges that are closest to the vertical edge.

(Step S49)
The edge determination unit 105 determines whether or not the vehicle length calculated in step S48 exceeds a predetermined threshold. If the vehicle length exceeds the threshold (YES in step S49), the process proceeds to step S50. On the other hand, if the vehicle length is equal to or less than the threshold (NO in step S49), the process proceeds to step S51.

(Step S50)
For example, the edge determination unit 105 classifies vehicle edges that are close to each other into the same group and stores them in the storage unit 106 . Specifically, the edge determination unit 105 compares the distances between the vehicle body edges, and groups the vehicle body edges whose distances are within a predetermined range.

(Step S51)
The edge determination unit 105 deletes the vehicle body edge formally registered in the group in step S47.

(Step S52)
When it is determined in step S46 that the vehicle body edge does not exist (NO in step S46), or when the process of step S50 or the process of step S51 ends, the edge determination unit 105 is stored in the storage unit 106. Calculate the center point of the vehicle edge belonging to each group. Specifically, the edge determination unit 105 calculates the center point of each vehicle edge in the longitudinal direction using, for example, a moving average method or an exponential smoothing method. The edge determination unit 105 regards the calculated center point of the vehicle edge as the center of the other vehicle 2 .

After that, the process returns to the main routine (FIG. 6).

In the above-described first embodiment, the detection control unit 103 detects when the certainty of the first vehicle detection unit 101 exceeds a predetermined threshold and the certainty of the second vehicle detection unit 102 becomes equal to or less than the threshold. , the first vehicle detection unit 101 is controlled to detect the other vehicle 2, the confidence of the first vehicle detection unit 101 is equal to or less than the threshold, and the confidence of the second vehicle detection unit 102 exceeds the threshold At this time, control is performed to cause the second vehicle detection unit 102 to detect the other vehicle 2 .

In the first embodiment described above, when the distance from the own vehicle 1 to the other vehicle 2 exceeds a predetermined distance, the other vehicle 2 is detected by the first vehicle detection unit 101 using the pattern matching method. 2 detections are made. When the other vehicle 2 exists in the vicinity where the distance from the own vehicle 1 to the other vehicle 2 is equal to or less than a predetermined distance, the difference between the field of view of the bird's-eye view image and the field of view of the image actually captured by the camera 20 is utilized. Then, the other vehicle 2 is detected using the difference between the bird's-eye view images. Thus, the other vehicle 2 that cannot be detected by the pattern matching method by the first vehicle detection unit 101 is detected by the second vehicle detection unit 102 . Accordingly, by using the bird's-eye view image based on the image captured by the camera 20 and taking the difference of the bird's-eye view image, it is possible to eliminate the motionless background portion. That is, by using the bird's-eye view image, there is no need to process the static background portion. As a result, the processing load can be reduced.

Although the embodiments of the present invention have been described above in detail with reference to the drawings, the embodiments are merely illustrative of the present invention, and the present invention is not limited only to the configurations of the embodiments. It goes without saying that even if there is a change in design within the scope of the gist of the invention, it is included in the present invention.

In the first embodiment described above, an example in which the predetermined angle is set to 30 degrees when scanning the vertical edge within the predetermined angle in the vehicle position calculation process has been described. The present invention is not limited to this aspect, and for example, the predetermined angle may be set to 3 degrees in order to distinguish pedestrians and bicycles smaller than other vehicles from other vehicles. As a result, when a plurality of edges exist within a range where the predetermined angle is 3 degrees or less, the edge determination unit 105 can determine that these edges are edges extracted from a pedestrian or a bicycle.

In the first embodiment described above, an example in which the vehicle detection device 100 is applied to a left-hand traffic road has been described. The present invention is not limited to this aspect, and for example, the vehicle detection device 100 may be applied to a right-hand traffic road.

In the first embodiment described above, the second vehicle detection unit 102 performs a subtraction process of subtracting the bird's-eye view image BI(t-Δt) from the bird's-eye view image BI(t) to generate the difference image DP. did. The present invention is not limited to this aspect. For example, the second vehicle detection unit 102 performs a subtraction process of subtracting the bird's-eye view image BI(t) from the bird's-eye view image BI(t−Δt) to generate the difference image DP. good too.

In addition, in Example 1 mentioned above, the example which applied the vehicle detection apparatus 100 to the scene which the own vehicle 1 parked in parallel and started from the state in which it stopped was shown. The present invention is not limited to this aspect, and as shown in FIG. 13, for example, the vehicle detection device 100 may be applied to a scene in which the own vehicle 1 is parked in parallel and started from a stopped state (FIG. 14). In this case, the detection of the other vehicle 2 is performed based on the correlation between the plurality of bird's-eye view images obtained by the viewpoint conversion for each of the plurality of images captured by the front camera 20F (imaging unit) at different times. will be As an operating condition of the vehicle detection device 100 in this scene, it is assumed that the own vehicle 1 starts at a low speed of 2 km/h or less in the direction indicated by the white arrow F1. In this scene, it is assumed that the speed of another vehicle approaching the vehicle 1 from the right side of the vehicle 1 in the direction indicated by the white arrow F2 is, for example, 25 km/h or less.

1 Own vehicle 2 Other vehicle 20 Camera (imaging unit)
20F: front camera (imaging unit)
20L: Left side camera (imaging unit)
20R: right side camera (imaging unit)
DESCRIPTION OF SYMBOLS 100... Vehicle detection apparatus 101... 1st vehicle detection part 102... 2nd vehicle detection part 103... Detection control part BI, BI (t-deltat), BI (t)... Bird's-eye view image DP... Difference images I, I(t-Δt), I(t)... Image RL... Radiation

Claims (1)

A first vehicle that detects the other vehicle by performing pattern recognition based on characteristics of other vehicles on an image obtained by imaging the surroundings of the own vehicle running on a road surface by an imaging unit that captures the surroundings of the vehicle. a detection unit;
For each of a plurality of images captured by the imaging unit at different times, correlation between a plurality of bird's-eye images obtained by viewpoint conversion so as to overlook the other vehicle from a virtual viewpoint above the own vehicle. a second vehicle detection unit that detects the other vehicle based on
When the certainty of the first vehicle detection unit exceeds a predetermined threshold and the certainty of the second vehicle detection unit is equal to or less than the threshold, the first vehicle detection unit detects the other vehicle. is detected, and when the certainty of the first vehicle detection unit is equal to or less than the threshold and the certainty of the second vehicle detection unit exceeds the threshold, the second vehicle detection unit A detection control unit that performs control to detect the other vehicle ,
The second vehicle detection unit divides the difference image between the plurality of bird's-eye view images into a plurality of three-dimensional object regions at intervals of a predetermined angle around the position of the imaging unit with radiation extending in a radial direction from the imaging unit. Evaluating the verticality of the other vehicle with respect to the road surface from among them, detecting the other vehicle based on the evaluation result,
an edge candidate determination unit for determining that, when a plurality of edges are present in the differential image along the radial direction, the plurality of edges are edge candidates extracted from the other vehicle;
The second vehicle detection unit detects the other vehicle based on the determination result of the edge candidate determination unit,
When a plurality of the edge candidates are determined by the edge candidate determination unit, the plurality of edge candidates are determined to be the other edge candidates if an angle between the plurality of edge candidates centered on the imaging unit is equal to or greater than a predetermined angle. An edge determination unit that determines that the edge is extracted from the vehicle,
The vehicle detection device , wherein the second vehicle detection unit detects the other vehicle based on the determination result of the edge determination unit .

Images