JP5430633B2 - Vehicle periphery monitoring device - Google Patents

Description

  The present invention relates to a vehicle periphery monitoring device that monitors the periphery of the vehicle using an image captured by an infrared camera or the like mounted on the vehicle, and particularly suitable for the periphery of the vehicle when applied at night or in a dark place. It relates to a monitoring device.

  Conventionally, as shown in Patent Document 1, an object such as a pedestrian who has a possibility of contact with the own vehicle from an image around the own vehicle (grayscale image and its binarized image) captured by an infrared camera. There is known a vehicle periphery monitoring device that detects an object and provides the detected object to a driver of the host vehicle.

  In the vehicle periphery monitoring device according to Patent Document 1, a high-temperature part is detected as an object in an image around the host vehicle captured by a pair of left and right infrared cameras (stereo cameras), and the object in the left and right images is detected. The distance to the target object is calculated by obtaining the parallax, and a pedestrian who is likely to affect the traveling of the host vehicle (possible contact) from the moving direction of the target object or the position of the target object Are detected and an alarm is output ([0014], [0018] of Patent Document 1).

  However, a vehicle periphery monitoring device using a pair of left and right infrared cameras is relatively expensive and is actually used in some high-end vehicles.

  In order to reduce costs, the vehicle periphery monitoring device according to Patent Document 2 uses a single infrared camera mounted on the vehicle to capture an object around the vehicle at least twice (2 frames) at predetermined time intervals. To do. The change in the size (size) of the object in the current captured image becomes larger as the relative speed between the object and the vehicle-periphery monitoring device is higher than the size (size) of the previous image. . An object existing in front of the vehicle has a shorter time to reach the vehicle as the relative speed between the object and the vehicle is higher. Thus, even with a single infrared camera, from the rate of change of the size of the image portion of the same object at a predetermined time interval, the arrival time to the vehicle, so-called TTC (Time To Contact or Time To Collation: The vicinity of the vehicle can be monitored by estimating the contact margin time ([0006], [0007], [0061], [0066] in Patent Document 2).

JP 2003-284057 A Japanese Patent No. 4267657

  By the way, the curved mirror attached to the utility pole, etc., to help the line of sight with the curve of the road, looks almost circular when viewed directly from the vehicle driver, and is similar to the pedestrian's head shape. It exhibits the shape.

  In this case, since the size of the curve mirror can be calculated instantaneously (every time an image is taken) based on the parallax of both cameras, the size of the curve mirror is the head of the pedestrian. In consideration of the fact that the curve mirror is larger than the pedestrian, erroneous detection (recognition) of the curve mirror can be instantaneously avoided.

  However, in the case of a single camera, as described above, it takes a predetermined time to calculate the distance, so it is difficult to ensure the distance accuracy as that of the stereo camera. Therefore, in the vehicle periphery monitoring device using a single camera, the possibility that the curve mirror is erroneously detected as the pedestrian is higher than that of the stereo camera, and there is room for improvement.

  The present invention has been made in consideration of such a problem, and even when the image is captured by a single imaging device, the possibility of erroneously detecting the curve mirror as a pedestrian can be reduced. An object of the present invention is to provide a vehicle periphery monitoring device.

  A vehicle periphery monitoring device according to the present invention is a curve mirror having a possibility of erroneously detecting a monitoring object in a vehicle periphery monitoring device that monitors the periphery of the vehicle using an image captured by an imaging device mounted on the vehicle. A curve mirror determination unit that determines whether or not the curve mirror determination unit extracts a substantially circular object or a substantially square object from the image as a curve mirror possibility object. And a pillar-shaped object detector for detecting the presence or absence of a pillar-shaped object in the vicinity of the extracted curved mirror possibility object, and the pillar-shaped object in the vicinity of the extracted curved mirror possibility object. When an object is detected, the curve mirror possibility object is determined to be the curve mirror.

  According to the present invention, when a pillar-shaped object is detected in the vicinity of a substantially circular or substantially quadrangular curve mirror object detected from an image captured by an imaging device mounted on a vehicle, the curve mirror Since it is determined that the possibility object is a curve mirror attached to the pillar-shaped object, even if the image is captured by a single imaging device, the possibility that the curve mirror possibility object is a walking target It is possible to reduce the possibility of erroneous detection of a person (head).

  In this case, the curve mirror determination unit further includes a width comparison determination unit, and when the width of the extracted object with the possibility of curve mirror is larger than the width of the detected pillar-shaped object, the curve mirror The possibility object may be determined to be the curve mirror. From the fact that the width of the pedestrian's head is equal to or less than the width of the torso of the pedestrian, the width of the extracted curved mirror possibility object is more than the width of the detected object of the pillar shape. Is larger, it is possible to more reliably determine that the object having the possibility of curve mirror is the curve mirror.

  Further, the curve mirror determination unit further includes an aspect ratio change determination unit, and along with a change in the relative angle between the imaging device and the extracted curve mirror possibility object as the vehicle moves, When the aspect ratio of the extracted curved mirror possibility object changes, the curve mirror possibility object may be determined to be the curve mirror. The curved mirror has a flat shape, and the pedestrian's head is generally cylindrical, so that the imaging device and the extracted curved mirror-possible object as the vehicle moves are relative to each other. When the aspect ratio of the extracted curved mirror possibility object changes as the angle changes, the curved mirror possibility object can be more reliably determined to be the curve mirror.

  The curve mirror determination unit further includes a relative position change determination unit, which is extracted in accordance with a change in a relative angle between the imaging device and the extracted curved mirror possibility object as the vehicle moves. When the distance between the curved mirror possibility object and the detected pillar-shaped object changes, the curved mirror possibility object may be determined to be the curve mirror. A finding that the distance between the pedestrian's head and the torso is small even if there is a change in the relative angle between the imaging device and the extracted pedestrian accompanying the movement of the vehicle. In consideration of the above, the pillar that has been detected as a possible object of the curved mirror is detected in accordance with a change in the relative angle between the imaging device and the potential object of the curved mirror that is extracted as the vehicle moves. When the distance between the object and the shape object changes, the curve mirror possibility object can be determined more reliably than the curve mirror.

  According to the present invention, when a pillar-shaped object is detected in the vicinity of a substantially circular or substantially quadrangular curve mirror object detected from an image captured by an imaging device mounted on a vehicle, the curve mirror Since the possibility object is determined to be a curve mirror attached to the pillar-shaped object, it is possible to falsely detect the possibility object of the curve mirror as an object to be monitored even if it is captured by a single imaging device. Sexuality can be reduced.

It is a block diagram which shows the structure of the vehicle periphery monitoring apparatus which concerns on one Embodiment of this invention. It is a schematic diagram of the vehicle carrying the vehicle periphery monitoring apparatus shown in FIG. It is a flowchart explaining the operation | movement by the image processing unit of a vehicle periphery monitoring apparatus. It is a detailed flowchart explaining a curve mirror determination process. It is explanatory drawing referred to a vertically long target object search process. It is explanatory drawing referred to width comparison processing. It is explanatory drawing referred to for a magnitude | size rule change determination process and a relative position change determination process. It is a characteristic figure with which it uses for description of a magnitude | size rule change determination process. FIG. 9A is a change characteristic diagram of the relative interval, and FIG. 9B is an explanatory diagram of the relative interval. It is a schematic diagram of the image of the pedestrian used for a pedestrian detection process. It is a schematic diagram of the vehicle by which the vehicle periphery monitoring apparatus which concerns on a modification is mounted.

  Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a block diagram showing a configuration of a vehicle periphery monitoring apparatus 10 according to an embodiment of the present invention. FIG. 2 is a schematic diagram of a vehicle (also referred to as a host vehicle) 12 on which the vehicle periphery monitoring device 10 shown in FIG. 1 is mounted.

  1 and 2, a vehicle periphery monitoring device 10 includes an image processing unit 14 that controls the vehicle periphery monitoring device 10 and a single (monocular) infrared camera 16 (imaging device) connected to the image processing unit 14. ), A vehicle speed sensor 18 for detecting the vehicle speed Vs of the vehicle 12, a brake sensor 20 for detecting the brake pedal operation amount (brake operation amount) Br by the driver, and a yaw rate sensor 22 for detecting the yaw rate Yr of the vehicle 12. A speaker 24 for issuing a warning or the like by voice and an image photographed by the infrared camera 16 are displayed, and a pedestrian or other object (moving object, monitoring object) having a high risk of contact is displayed on the vehicle 12. And an image display device 26 including a HUD (Head Up Display) 26a for causing the driver to recognize.

  The image display device 26 is not limited to the HUD (head-up display) 26a, but a display that displays a map or the like of a navigation system mounted on the vehicle 12, a display that displays fuel consumption or the like provided in a meter unit or the like ( Multi-information display) can be used.

  The image processing unit 14 detects an object to be monitored such as a pedestrian in front of the vehicle from an infrared image around the vehicle 12 and a signal indicating the running state of the vehicle (here, the vehicle speed Vs, the brake operation amount Br, and the yaw rate Yr). A detected image that is detected and issued with a warning (for example, a beeping sound) from the speaker 24 when it is determined that there is a high possibility of contact with the monitored object, and is displayed in gray scale on the HUD 26a. The object to be monitored is highlighted with a conspicuous color frame such as yellow or red. In this way, the driver's attention is drawn.

  Here, the image processing unit 14 performs input processing such as an A / D conversion circuit that converts an input analog signal into a digital signal, an image memory (storage unit 14m) that stores a digitized image signal, and various arithmetic processes. CPU (Central Processing Unit) 14c to be executed, RAM (Random Access Memory) used for storing data being calculated by CPU 14c, programs, tables, maps and templates executed by CPU 14c {pedestrian (human body) shape template: Storage unit 14m such as ROM (Read Only Memory) for storing right direction, left direction, front (rear) direction, curve mirror shape template, etc., a clock (clock unit), a timer (time unit), and a drive signal for speaker 24 And an output for outputting a display signal of the image display device 26 Includes a road or the like, the output signal of the infrared camera 16, the yaw rate sensor 22, a vehicle speed sensor 18, and the brake sensor 20 is configured to be inputted is converted into a digital signal by the CPU 14c.

  The CPU 14c of the image processing unit 14 functions as various functional units (also referred to as functional units) by taking in these digital signals and executing programs while referring to tables, maps, templates, and the like, and the speaker 24 and the image. A drive signal (audio signal or display signal) is sent to the display device 26. These functions can also be realized by hardware.

  In this embodiment, the image processing unit 14 includes the functional units such as a pedestrian detection processing unit 102, a curve mirror determination unit 104, a contact possibility determination unit 106, and an attention output generation determination unit 108, which will be described in detail later. Have. Here, the curve mirror determination unit 104 includes a curve mirror possibility object extraction unit 104a, a columnar shape object detection unit 104b, a width comparison determination unit 104c, an aspect ratio change determination unit 104d, a relative position change determination unit 104e, and the like.

  The image processing unit 14 basically includes an image acquired by the infrared camera 16 and the shape of an artificial structure such as a human body shape, an animal shape, a vehicle shape, and a pole including a power pole stored in the storage unit 14m. An object recognition processing (object detection processing) program for recognizing an object as compared with a schematic template such as is executed.

  As shown in FIG. 2, the infrared camera 16 has a front bumper portion of the vehicle 12 and a center portion in the vehicle width direction of the vehicle 12 in which the optical axis is arranged in parallel with the axle, and the temperature of the imaging target (object). Is higher, the output signal (imaging signal) level is higher (the luminance is increased).

  Further, the HUD 26a is provided on the front windshield of the vehicle 12 so that the display screen is displayed at a position that does not obstruct the driver's front view.

  Here, the image processing unit 14 converts the analog video signal output from the infrared camera 16 into digital data every frame clock interval / cycle of several tens ms, for example, 1 second / 30 frames [ms]. It has the above-described function of performing various arithmetic processes on the image ahead of the vehicle captured in the storage unit 14m (image memory) and captured in the storage unit 14m.

  The pedestrian detection processing unit 102 extracts an image portion of a monitoring target such as a pedestrian from the image in front of the vehicle captured in the storage unit 14m. The alert output generation determination unit 108 calculates a rate of change Rate of the size of the image portion of the same monitoring object between the images captured at the frame clock interval / period (predetermined time interval), and further, the rate of change The time T until the monitored object reaches the vehicle 12 is estimated using the Rate, the position of the monitored object in the real space is calculated, and the movement vector of the monitored object in the real space is calculated.

The time until the monitored object reaches the vehicle 12 (also referred to as a contact margin time) TTC, in other words, the time TTC until the vehicle 12 contacts the monitored object is obtained from the rate of change Rate (image). ) And an imaging interval (frame clock cycle) dT (known), which is a predetermined time interval, in a known manner, for example, as shown in Patent Document 2, can be obtained by the following equation (1): .
TTC = dT × Rate / (1-Rate) (1)

  Note that the rate of change Rate is the width or length W0 of the monitoring object in the image at the previous imaging of the monitoring object (which may be stored as the number of pixels, respectively) and the image at the time of the current imaging. Can be obtained by a ratio (Rate = W0 / W1) to the width or length W1 (number of pixels) of the same monitoring object.

Further, the distance Z to the monitoring object can be obtained by the following equation (2) as shown in Patent Document 2. In the equation (2), Vs is more accurately the relative speed between the monitored object and the vehicle 12. When the monitoring object is stopped, the relative speed is equal to the vehicle speed Vs.
Z = Rate × Vs × dT / (1-Rate) (2)

  Further, the attention output generation determination unit 108 detects the positional change amount Δx (horizontal direction), Δy of the image portion of the same monitoring object between the calculated contact margin time TTC and the images captured at a predetermined time interval. (Vertical direction) is calculated, and the possibility of contact between the monitoring object and the vehicle 12 is determined based on the calculated position change amounts (movement vectors) Δx and Δy.

  As will be described later, the curve mirror determination unit 104 is a pedestrian (or its head) or a curve mirror in the image portion of the monitoring object extracted from the image ahead of the vehicle captured in the storage unit 14m. In the case of a curve mirror, a process of excluding it from the monitoring target is performed.

  The detailed operation of the vehicle periphery monitoring apparatus 10 according to this embodiment configured and operated as described above will be basically described with reference to the flowchart of FIG.

  First, in step S1, the image processing unit 14 determines the traveling state (running or stopped) of the vehicle 12 from the vehicle speed Vs detected by the vehicle speed sensor 18, and the vehicle is stopped (step S1: NO). The process is stopped.

  When the vehicle is running (step S1: YES), in step S2, the image processing unit 14 captures an infrared image, which is an output signal for each frame in a predetermined field angle range in front of the vehicle, which is captured for each frame by the infrared camera 16. The obtained grayscale image is stored in the image memory (storage unit 14m), and the stored grayscale image is binarized, that is, a region brighter than the luminance threshold is “1” (white). Then, a conversion process into a binary image in which the dark area is “0” (black) is performed, and the converted binary image is also stored in the storage unit 14m for each frame in association with the frame. In this binarization process, depending on how the threshold is selected, the human body detects a lump (candidate human body object) consisting of the head, shoulders, torso, and two legs as a lump (aggregate) of “1”. Is done. Further, in a curved mirror attached to a pole such as a utility pole, the mirror body and the pole are detected as a lump (aggregate) of “1”.

  In this case, in step S3, the image processing unit 14 converts “1” (white) of the binarized image for each frame (image) into run length data for each scanning line in the x direction (horizontal direction). A line with a portion overlapping in the y direction (vertical direction) is regarded as one object, a label is attached to each of the objects, and a labeling process is performed as a candidate object. In this case, labels are attached to human body candidates (pedestrian candidates, monitoring objects) including head candidates and curve mirror candidates (monitoring unnecessary objects) including mirror main body candidates.

  Next, in step S4, the curve mirror determination unit 104 selects a curve mirror candidate from the labeled object candidate (a human body candidate including a head candidate and a curve mirror candidate including a mirror main body candidate) as an object (can be contacted). Curve mirror determination processing is performed to remove the object from the target object.

  FIG. 4 shows a detailed flowchart of the curve mirror determination process in step S4.

  First, the aspect ratio determination process in step S4a is performed by the curve mirror possibility object extraction unit 104a.

The curve mirror possibility object extraction unit 104a includes the head candidate that has a circular shape, an elliptical shape, or a quadrilateral shape among the candidate human body and the candidate mirror body each having a labeled head candidate, Both of the mirror body candidates are extracted as a curved mirror possibility object, and the vertical length (H) and the horizontal length for each of the head candidate and the mirror body candidate constituting the extracted curved mirror possibility object. The ratio H / W of (W) is estimated to be the head of the human body, the condition of the following formula (3) {the ratio of the vertical length H to the horizontal length W is between 1/2 or more and 2 Value} is determined.
1/2 ≦ H / W ≦ 2 (3)

  An object candidate that does not satisfy the condition of the expression (3) is excluded from the subsequent processing targets, assuming that it is not a head candidate but a curve mirror candidate including the mirror main body candidate.

  Next, the vertically long target object search process in step S4b is performed by the columnar object detection unit 104b.

  In this case, the strut-shaped object detection unit 104b is a grayscale image of the target candidate (the human body candidate including the head candidate and the curve mirror candidate including the mirror main body candidate) left (not excluded) in the process of step S4a. 5, for example, as shown in FIG. 5, object candidates having a size of height (height) H and width (width) W (this object candidate is the above-described head candidate or mirror body candidate). On the top and bottom of 30, the size is, for example, an area having a length H that is the same as the length H of the candidate object 30 and a width (width) W × 3 that is three times the width W of the candidate object 30 An upper mask region 32u and a lower mask region 32d are set.

  In general, in the case of a curved mirror installed on a utility pole, a pillar-shaped object that is a vertically long object exists in the upper mask area 32u and the lower mask area 32d in contact with the object candidate 30, and the upper mask area The feature of the pillar-shaped object extracted in 32u should be similar to the feature of the pillar-shaped object extracted in the lower mask region 32d.

  Here, the feature is, for example, a vertical edge or a luminance profile, and the similarity determination by the columnar shape object detection unit 104b is, for example, the vertical edge widths hwcu and hwcd calculated in the upper mask region 32u and the lower mask region 32d. This difference is determined based on whether or not the difference between the two is within the threshold value Δth (| hwcu−hwcd | ≦ Δth) or whether the difference in the luminance profile width is within the threshold value.

  When the vertically long objects 34u and 34d (also referred to as the vertically long object 34 or the columnar object 34) are detected above and below the object candidate 30 in step S4b, the curve mirror determination unit 104 is shown in FIG. The object candidate 30, that is, the mirror main body candidate, is determined to be a curve mirror candidate installed on the utility pole, and is excluded from the subsequent processing targets.

  Next, the width comparison process in step S4c is performed by the width comparison determination unit 104c. In this case, the width comparison determination unit 104c determines the vertical edge width hwcd (referred to as the width W1 of the vertically long object 34d) extracted in the lower mask region 32d in step S4b and the height H (or width) of the object candidate 30. W) and the height H (or width W) of the object candidate 30 is larger than the width W1 of the vertically long object 34d (H> W1 = hwcd), the object candidate 30 is not a human body candidate. However, it is regarded as a curve mirror candidate and excluded from the subsequent processing targets. In general, the height H (or width W) of the mirror main body is larger than the width (thickness) of the pole (pole) or the electric pole to which the mirror main body is attached, and in the human body, the width of the body is larger than the height and width of the head. Since it is large, it is not a human body candidate but a curve mirror candidate, and is excluded from subsequent processing targets.

  Note that the determination in step S4c is preferably applied to a situation (situation) in which the vehicle 12 is traveling toward the T-junction 36, as shown in FIG. In this case, the vehicle 12 travels toward the front of the mirror main body 40 installed on the support column 38 (including the power pole) constituting the curve mirror 37, and the captured image of the infrared camera 16 acquired during the travel. , The change in the positional relationship between the support column 38 and the mirror body 40 constituting the curved mirror 37 is small. Therefore, it is possible to determine whether or not the object is the curve mirror 37 by comparing the height H (or width W) of the object candidate 30 with the width W1 of the vertically long object 34d. For example, when H> W1 = hwcd, the width comparison determination unit 104c can determine that the curve mirror 37 is used.

  Next, the size rule change determination process in step S4d is performed by the aspect ratio change determination unit 104d. In step S4d, as shown in FIG. 7, when the vehicle 12 is traveling toward the L-shaped road 42, the vehicle 12 includes a mirror mirror 40 provided with a mirror body 40 installed on the support column 38. Keep in mind that you drive towards the side of the car.

  In this case, as can be seen from the characteristic 44 in FIG. 8, in the image captured by the infrared camera 16, the object candidate 30 is located in the mirror main body 40 (for convenience of understanding) far from the L-shaped path 42 (in the past). For example, it is assumed that the shape is circular.), The aspect ratio H / W is H / W >> 1 because it looks like an ellipse, and as it approaches (current position), Since the relative angle θ (see FIG. 7) between the infrared camera 16 and the curved mirror possibility object (object candidate 30) increases, the aspect ratio H / W of the mirror body 40 (object candidate 30) is a perfect circle. The pattern approaches the corresponding value 1. Therefore, when the aspect ratio H / W has the characteristic 44, the object candidate 30 is determined to be the curved mirror 37 including the mirror body 40, and is excluded from the subsequent processing. If the object candidate 30 is a head candidate, the head of the human body can be approximated in a spherical shape or a cylindrical shape. Therefore, even if the vehicle 12 travels in the direction indicated by the arrow in FIG. There is little change in / W.

  Next, the relative position change determination process in step S4e is performed by the relative position change determination unit 104e. Note that the determination in step S4e is the same as the determination in step S4c, and the vehicle 12 travels toward the side surface of the curve mirror 37 including the mirror main body 40 installed on the support column 38, as shown in FIG. doing.

  In this case, in the image captured by the infrared camera 16, as shown in FIG. 9B, the object candidate (head candidate or mirror main body candidate) that appears in the shape of a vertically long ellipse far from the L-shaped path 42 (in the past). The relative distance (relative width) RW between the center of the object 30 and the axis of the vertically long object 34 is narrow, but as it approaches (currently), the center of the object candidate 30 and the object The relative interval RW between the axis of the object 34 is wide. FIG. 9A shows the characteristic 46 of the relative interval RW.

  In the human body, the distance between the center of the head (the axis of the cylinder) and the axis of the torso does not change so much, so the center of the candidate object 30 and the axis of the vertically long object 34 When the relative interval RW changes greatly, it can be excluded from the subsequent processing targets as a curve mirror candidate.

  Thus, after finishing the curve mirror determination process of step S4 consisting of steps S4a to S4e, in step S5 of FIG. 3, the object candidate regarded as a curve mirror having a mirror main body installed on a utility pole or support column. 30 is regarded as not a pedestrian candidate (human body candidate), and is excluded from the subsequent processing targets (step S5: YES).

  Next, in step S6, pedestrian detection is performed on the remaining image portion from which the curve mirror candidate is excluded by the processing up to step S5 (the remaining object candidate 30 in which the curve mirror is excluded from the object candidate 30). The process part 102 implements the pedestrian detection process which detects the pedestrian as a monitoring target object in a well-known way.

  The pedestrian detection processing unit 102 has a mask area 53 (in FIG. 10) that is slightly larger than the circumscribed rectangle 52 of the human body candidate (also referred to as a pedestrian candidate) 62 having the head 54 shown in FIG. 10 labeled in step S3. , The pixel value in the mask area 53 is scanned from the upper side to the lower side and the pixel value is scanned from the left side to the right side. Then, when the scanning portion continues, it is determined that the scanning portion is a boundary between the pedestrian candidate 62 (consisting of the head portion 54, the body portion 60, and the leg portion 59) in the image and the road surface 51, and the lower end of the object. OBbm.

  Further, in step S6, the pedestrian detection processing unit 102 searches while scanning pixel by pixel from the left side to the right side with respect to the image in the mask area 53 on the upper side, conversely from the lower end OBbm of the object, When a change section of the horizontal edge of luminance in the vertical direction (in the binarized image, a section in which a pair of “1” and “0” is substantially continuous) is detected, the scanning part of the change section is the human body candidate 62 and the background. The upper end OBtm of the object that is the edge of the boundary.

  Next, in step S <b> 7, the contact possibility determination unit 106 determines the contact possibility between the pedestrian candidate 62 detected in step S <b> 6 and the host vehicle 12.

  That is, as described above, the contact possibility is determined by considering each contact margin time TTC of the above-described equation (1) and each movement vector of the pedestrian candidate 62 with respect to the pedestrian candidate 62 (including the distance Z). In addition, the vehicle 12 can contact the pedestrian candidate 62 based on the brake operation amount Br, the vehicle speed Vs, and the yaw rate Yr that are the outputs of the brake sensor 20, the vehicle speed sensor 18, and the yaw rate sensor 22, respectively. If it is determined whether or not there is a possibility of contact (step S7: YES), in step S8, the alert output generation determination unit 108 generates an alert output, and the driver To alert the driver (provide information to the driver). Specifically, a pedestrian in the gray scale image is highlighted with a conspicuous color frame or the like and displayed on the HUD 26a, and an alarm is generated through the speaker 24 to alert the driver of the vehicle 12.

  Here, on the display of the HUD 26a, the attention output generation determination unit 108 generates an output for encouraging attention by surrounding the pedestrian candidate 62 with the above-described conspicuous color frame such as red or yellow.

[Outline of Embodiment]
As described above, the vehicle periphery monitoring device 10 according to the above-described embodiment excludes the curve mirror 37 from the image captured by the single infrared camera 16 mounted on the vehicle 12 and performs monitoring that requires a warning output. Detect pedestrians (including animals) that are objects.

  In this case, the vehicle periphery monitoring device 10 includes a curve mirror determination unit 104 that determines whether or not the vehicle is a curve mirror 37 that may be erroneously detected as a pedestrian as a monitoring target.

  The curve mirror determination unit 104 extracts a substantially circular object or a substantially square object from the image as a curve mirror possibility object, that is, a curve mirror possibility object extraction unit 104a that extracts a human head candidate and a mirror body candidate, A columnar object detection unit 104b that detects the presence or absence of a columnar object (in the above embodiment, the vertically long object 34) in the vicinity of the curved mirrorable object, and the extracted curve mirror is possible When the pillar-shaped object is detected in the vicinity of a sex object, the curve mirror possibility object is determined to be the curve mirror.

  That is, when a pillar-shaped object is detected in the vicinity of a substantially circular or substantially square curved mirror possibility object detected from an image captured by a single infrared camera 16 mounted on the vehicle 12, the curve Since it is determined that the mirror possibility object is the curve mirror 37 attached to the pillar-shaped object, even if it is captured by a single infrared camera 16, the curve mirror possibility object is determined to be a pedestrian ( The possibility of erroneous detection of the head) can be reduced.

  When the pole-shaped object is a utility pole, the curved mirror 37 is accurately recognized by comparing the calculated utility pole width with the mirror body width based on the knowledge that the utility pole width is equivalent to the head width. You can also. Since it can be recognized only by the image, the curve mirror 37 can be easily recognized regardless of whether the distance accuracy is good or bad.

  In this case, the curve mirror determination unit 104 further includes a width comparison determination unit 104c, and the width W of the extracted curve mirror possibility object (target object candidate 30) is detected as the columnar object (vertically long). When the width W1 of the object 34) is larger than the width W1 (see FIG. 5), the curve mirror possibility object (object candidate 30) may be determined to be the curve mirror 37 (mirror body 40). Based on the fact that the pedestrian's head width hwm (see FIG. 10) is equal to or smaller than the pedestrian's torso width Wm (see FIG. 10), the extracted curve mirror possibility object (object) When the width W of the candidate 30) is larger than the width W1 of the detected pillar-shaped object (vertically long object 34), the curve mirror possibility object (object candidate 30) is moved to the curve mirror 37 (mirror body 40). ) Can be determined more reliably.

  Further, the curve mirror determination unit 104 further includes an aspect ratio change determination unit 104d, and the relative relationship between the infrared camera 16 and the extracted object (possible object 30) that can be curved mirrors as the vehicle 12 moves. When the aspect ratio H / W of the extracted curved mirror possibility object (object candidate 30) changes as shown by the characteristic 44 in FIG. 8 with the change of the angle θ (see FIG. 7), You may make it determine the said curve mirror possibility object (target object candidate 30) as the curve mirror 37 (mirror main body 40). The curved mirror 37 (mirror main body 40) has a flat shape, and the pedestrian's head 54 has a substantially cylindrical shape (height hh × diameter hwm: see FIG. 10), so that the vehicle 12 moves. The aspect ratio of the extracted curved mirror possibility object (target object 30) in accordance with the change in the relative angle θ between the infrared camera 16 and the extracted curved mirror possibility object (target object candidate 30). When the H / W changes, it is possible to more reliably determine that the curved mirror possibility object (target object 30) is the curved mirror 37 (mirror body 40).

  The curve mirror determination unit 104 further includes a relative position change determination unit 104e. The curve mirror determination unit 104 has a relative angle θ between the infrared camera 16 and the extracted object (possible target object 30) that is likely to be a curve mirror as the vehicle 12 moves. Along with the change, the relative distance RW, which is the distance between the extracted curved mirror possibility object (object candidate 30) and the detected pillar-shaped object (vertically long object 34), is the characteristic of FIG. 9A. When it changes as shown in 46, you may make it determine the said curve mirror possibility object (target object candidate 30) to be the curve mirror 37. FIG. The distance between the vertical axis of the pedestrian's head 54 and the vertical axis of the body 60 is the relative angle θ between the infrared camera 16 and the extracted pedestrian as the vehicle 12 moves. The relative angle θ between the infrared camera 16 and the extracted curved mirror possibility object (object candidate 30) that accompanies the movement of the vehicle 12 in consideration of the knowledge that the amount of change is small even if there is a change in When the relative distance RW, which is the distance between the extracted object (possible object 30) and the detected pillar-shaped object (vertically long object 34), changes with the change. The curved mirror possibility object (object candidate 30) can be more reliably determined to be the curve mirror 37.

  As described above, according to the vehicle periphery monitoring apparatus 10 according to the above-described embodiment, a substantially circular or substantially quadrangular curve mirror-possible object (target object candidate) detected from an image captured by the infrared camera 16 mounted on the vehicle 12. 30), when a pillar-shaped object (vertically long object 34) is detected, the curved mirror possibility object (object candidate 30) is attached to the pillar-shaped object (vertically long object 34). Since it is determined to be the mirror body 40, even if the image is captured by the single infrared camera 16, the possibility that the object having the curved mirror (target candidate 30) is erroneously detected as a pedestrian (head) is likely. Can be reduced.

  Note that the present invention is not limited to the above-described embodiment, and it is needless to say that various configurations can be adopted based on the contents described in this specification.

  For example, as shown in FIG. 11, a vehicle periphery monitoring device including a pair of left and right infrared cameras 16R and 16L mounted on a vehicle 12A may be used. Infrared cameras 16R and 16L as so-called stereo cameras are arranged on the front bumper portion of the vehicle 12A at positions substantially symmetrical with respect to the center in the vehicle width direction of the vehicle 12A, and the two infrared cameras 16R and 16L. The optical axes are parallel to each other, and the height from both road surfaces is fixed. As is well known, in the vehicle periphery monitoring apparatus provided with the left and right set of infrared cameras 16R and 16L, a part having a high temperature in the image around the host vehicle captured by the set of left and right infrared cameras 16R and 16L is used as an object. In addition, the distance to the object is calculated by the principle of triangulation using the parallax of the same object in the left and right images, and from the moving direction (movement vector) of the object and the position of the object ( The own vehicle) detects an object (object that may be contacted) that is likely to affect the traveling of 12A, and outputs a warning output.

  In addition, a general digital video camera (imaging device) that captures a normal visible region can be used as a single camera or a stereo camera as in the embodiment without using an infrared camera.

  Furthermore, when the vehicles 12 and 12A are equipped with a navigation device, the vehicle periphery monitoring device 10 (the image processing unit 14 thereof) is configured to be able to cooperate with the navigation device, and is obtained from the navigation device. The vehicle (own vehicle) 12, 12A is traveling toward the T-junction 36 based on information such as the vehicle location information, map information, and facility information (FIG. 6) or toward the L-junction 42. When it is recognized that the vehicle is traveling (FIG. 7), or when such a situation is received from the navigation device, the image processing unit 14 determines the width by the width comparison determination unit 104c. Triggering (starting) comparison processing, aspect ratio change determination processing by the aspect ratio change determination unit 104d, and relative interval change determination processing by the relative position change determination unit 104e You may be prompted to select.

DESCRIPTION OF SYMBOLS 10 ... Vehicle periphery monitoring apparatus 12, 12A ... Vehicle 14 ... Image processing unit 14c ... CPU
14m ... Storage unit 16, 16R, 16L ... Infrared camera 102 ... Pedestrian detection processing unit 104 ... Curve mirror determination unit 106 ... Contact possibility determination unit 108 ... Attention output generation determination unit

Claims (5)

In a vehicle periphery monitoring device that monitors the periphery of the vehicle using an image captured by an image capturing device mounted on the vehicle,
A curve mirror determination unit that determines whether or not the object is a curve mirror that has the possibility of erroneous detection as a monitoring object,
The curve mirror determination unit
A curve mirror possibility object extraction unit for extracting a substantially circular object or a substantially square object from the image as a curve mirror possibility object;
A pillar-shaped object detection unit that detects the presence or absence of a pillar-shaped object in the vicinity of the extracted curved mirror possibility object ;
If the extracted issued an object of the strut shape in the vicinity of the curved mirror potential object is detected, caused by the movement of the vehicle, the relative angle between said imaging device, extracting said curved mirrors possibility object was An aspect ratio change determination unit that determines that the curved mirror possibility object is the curve mirror when the aspect ratio of the extracted curved mirror possibility object is changed,
Vehicle periphery monitoring apparatus characterized by having a. In a vehicle periphery monitoring device that monitors the periphery of the vehicle using an image captured by an image capturing device mounted on the vehicle,
A curve mirror determination unit that determines whether or not the object is a curve mirror that has the possibility of erroneous detection as a monitoring object,
The curve mirror determination unit
A curve mirror possibility object extraction unit for extracting a substantially circular object or a substantially square object from the image as a curve mirror possibility object;
A pillar-shaped object detection unit that detects the presence or absence of a pillar-shaped object in the vicinity of the extracted curved mirror possibility object ;
If the extracted issued an object of the strut shape in the vicinity of the curved mirror potential object is detected, caused by the movement of the vehicle, the relative angle between said imaging device, extracting said curved mirrors possibility object was When the distance between the extracted curved mirror possibility object and the detected pillar-shaped object changes in accordance with the change of the relative position which determines the curved mirror possibility object to be the curve mirror A position change determination unit;
Vehicle periphery monitoring apparatus characterized by having a. In a vehicle periphery monitoring device that monitors the periphery of the vehicle using an image captured by an image capturing device mounted on the vehicle,
A curve mirror determination unit that determines whether or not the object is a curve mirror that has the possibility of erroneous detection as a monitoring object,
The curve mirror determination unit
A curve mirror possibility object extraction unit for extracting a substantially circular object or a substantially square object from the image as a curve mirror possibility object;
A pillar-shaped object detection unit that detects the presence or absence of a pillar-shaped object in the vicinity of the extracted curved mirror possibility object;
When the pillar-shaped object is detected above and below the extracted curved mirror possibility object, the curved mirror possibility object is determined to be the curve mirror.
The vehicle periphery monitoring apparatus characterized by the above-mentioned. In the vehicle periphery monitoring device according to claim 3,
The curve mirror determination unit
When the strut-shaped object whose width difference is within a threshold value is detected above and below the extracted curved mirror possibility object, the curved mirror possibility object is determined to be the curve mirror.
The vehicle periphery monitoring apparatus characterized by the above-mentioned. In the vehicle periphery monitoring device according to any one of claims 1 to 4 ,
The curve mirror determination unit
A width comparison / determination unit that determines that the curved mirror possibility object is the curve mirror when the width of the extracted curved mirror possibility object is larger than the detected width of the pillar-shaped object. The vehicle periphery monitoring apparatus characterized by the above-mentioned.

Images