JP2012256159A - Approaching object detecting device and method for detecting approaching object - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide "an approaching object detecting device and a method for detecting an approaching object" capable of quickly determining whether it is an approaching object or not and reducing an output of an erroneous warning directed to a stationary object when applied to an approaching object warning system, by reducing an erroneous determination that the stationary object is the approaching object.SOLUTION: An approaching object detecting device includes: center-of-gravity calculating means 5 for calculating the center of gravity of a three-dimensional object that is shown in a latest detection result every time a latest detection result of the three-dimensional object is obtained by approaching object candidate detecting means 4; and moving direction calculating means 6 for calculating the moving direction of the center of gravity on the basis of the calculation result of the center of gravity. When the moving direction of the center of gravity is the direction showing an approach of the moving body, approaching object determination means 7 determines that it is the approaching object and, when the moving direction of the center of gravity is the direction showing a separation from the moving body, determines that it is not the approaching object.

Description

  The present invention relates to an approaching object detection apparatus and an approaching object detection method, and more particularly to an approaching object detection apparatus and an approaching object detection method suitable for detecting an approaching object based on a captured image of an imaging unit.

  Conventionally, an obstacle present on a road surface in a captured image is detected based on a captured image (input image from the camera) around the vehicle by a monocular (single) camera attached to the vehicle, and the detected obstacle A road obstacle warning system that outputs a warning for an object has been known.

  Various types of three-dimensional object detection methods for detecting a three-dimensional object (obstacle) on the road surface are applied to this type of system. For example, as shown in Patent Document 1, the amount of movement of the host vehicle is grasped. A method of extracting only a three-dimensional object (moving body) on the road surface by removing unnecessary areas such as a road surface and a distant view by comparing a past captured image by a monocular camera and a current captured image (so-called Background subtraction method) is known.

  Also, in this type of system, whether or not the three-dimensional object is an approaching object so that only a three-dimensional object (for example, a pedestrian or another vehicle) approaching the host vehicle from the blind spot of the user is targeted for warning. There was something to judge.

JP-A-10-222679 JP 2007-172540 A JP 2010-128869 A

  However, conventionally, there has been a problem that false alarms frequently occur for stationary objects (such as other vehicles) existing behind the host vehicle when the host vehicle is moving backward.

  Here, FIG. 6 and FIG. 7 are conceptual diagrams showing the reason why such a problem occurs.

  Specifically, FIG. 6A shows the captured image of the imaging region within the viewing angle behind the host vehicle by the back camera of the host vehicle and another vehicle (stationary object) stopped behind the host vehicle. The captured image including the other vehicle in the case where the distance to is relatively large is shown. FIG. 6B is a plan view schematically showing the positional relationship between the host vehicle and the other vehicle at this time. In the state of FIG. 6, when the other vehicle is detected by the background subtraction method or the like, as shown in FIG. 6B, the other vehicle is moved to the actual position (filled circular mark in the drawing). ) Is detected at a position farther from the host vehicle (a broken line circular mark portion in the figure). This is because the distance of the other vehicle from the host vehicle is large and the difference between the current captured image and the past captured image used for detection of the other vehicle as a three-dimensional object is small. However, a captured image used for detection of a three-dimensional object may be one after distortion correction taking into account distortion of a camera lens (fisheye lens or the like). In addition, the past captured image has been subjected to necessary deformation (coordinate conversion such as movement and rotation) according to the amount of movement of the host vehicle (camera) from the time of image capture to the present.

  On the other hand, FIG. 7A shows a captured image of the back camera when the host vehicle approaches the other vehicle by retreating from the state of FIG. Moreover, FIG.7 (b) is a top view which shows the positional relationship of the own vehicle and the said other vehicle at this time. In the state of FIG. 7, when the other vehicle is detected, the distance of the other vehicle from the own vehicle is small, and the current captured image and the past captured image (considering the movement amount of the own vehicle are taken into account). As shown in FIG. 7B, the other vehicle is detected at a position close to the actual position (solid circle mark portion in the figure).

  When such a large displacement of the detection position of the other vehicle occurs, the system side determines that the vehicle is approaching despite the other vehicle being stopped.

  In Patent Document 2, an average value of motion vector directions of pixels in a connected block is calculated, the calculated average value is specified as a moving direction of the connected block, and the specified moving direction and a predetermined approach direction are Is calculated, and the approach is determined according to the calculated angle difference. Patent Document 3 describes that an optical flow method is used to extract a flow area generated in the center direction on the screen and specify it as an approaching object area based on the relationship between the approaching object and the field of view of the camera. Has been. However, the configurations described in these documents have a problem that the amount of calculation is large and it is difficult to quickly determine an approaching object.

  Therefore, the present invention has been made in view of such problems, and can quickly determine whether or not a three-dimensional object is an approaching object, and erroneously state that a stationary object is an approaching object. The object of the present invention is to provide an approaching object detection device and an approaching object detection method capable of reducing the output of a false alarm for a stationary object when applied to an approaching object warning system with reduced determination. is there.

  In order to achieve the above-described object, an approaching object detection device according to the present invention is arranged at a predetermined position of a moving body, and a single imaging unit that images a predetermined imaging area around the moving body, and this imaging Based on the captured image of the means, an approaching object candidate detecting means for detecting a solid object in the imaging region as an approaching object candidate at predetermined intervals, and the three-dimensional object detected by the approaching object candidate detecting means is moved. An approaching object detection device comprising an approaching object determination unit that determines whether or not the object is approaching the body, each time the latest detection result of the three-dimensional object is acquired by the approaching object candidate detection unit In addition, a center-of-gravity calculating unit that calculates the center of gravity of the three-dimensional object indicated in the latest detection result, and a moving direction calculating unit that calculates the moving direction of the center of gravity based on the calculation result of the center of gravity by the center-of-gravity calculating unit. With the front The approaching object determining means determines that the approaching object is the approaching object when the moving direction of the center of gravity calculated by the moving direction calculating means is a direction indicating approach to the moving body, When the moving direction of the center of gravity is a direction indicating separation from the moving body, it is determined that the moving object is not the approaching object.

  In the approaching object detection method according to the present invention, a single imaging unit that images a predetermined imaging region around the moving body is arranged at a predetermined position of the moving body, and the approaching object detection method is based on the captured image of the imaging unit. An approaching object detection method for detecting a three-dimensional object in the imaging region as an approaching object candidate at predetermined intervals and determining whether the detected three-dimensional object is an approaching object approaching the moving object. And each time the latest detection result of the three-dimensional object is acquired, a first step of calculating the center of gravity of the three-dimensional object indicated in the latest detection result, and the calculation of the center of gravity in the first step Based on the result, when the second step of calculating the moving direction of the center of gravity and the moving direction of the center of gravity calculated in the second step are directions indicating the approach to the moving body, It is determined that the approaching object, Write, the direction of movement of the center of gravity, when the direction indicating the spacing from the moving body is characterized in that it comprises a third step for judging that there is the approaching object.

  According to the present invention, it is possible to easily, quickly and accurately determine whether or not a three-dimensional object is an approaching object based on the detected movement direction of the center of gravity of the three-dimensional object. .

  Further, in the approaching object detection device of the present invention, for a three-dimensional object determined to be the approaching object, an alarm is output for this, and for a three-dimensional object determined not to be the approaching object, You may provide the alarm output means which does not output the said alarm targeting this. Accordingly, in the approaching object detection method of the present invention, for the three-dimensional object determined to be the approaching object in the third step, an alarm is output for the three-dimensional object, and the third step is performed. For the three-dimensional object determined not to be the approaching object in step 4, a fourth step may be included in which the warning is not output for this object.

  In such a case, it is possible to reliably reduce the output of a false alarm for a stationary object.

  Furthermore, in the approaching object detection device of the present invention, the center-of-gravity calculation unit may use the pixel value of the three-dimensional object detected by the approaching object candidate detection unit for the calculation of the center of gravity. Accordingly, in the approaching object detection method of the present invention, the pixel value of the detected three-dimensional object may be used for calculating the center of gravity in the first step.

  In such a case, the calculation accuracy of the center of gravity can be improved by using the pixel value of the three-dimensional object.

  According to the present invention, it is possible to quickly determine whether or not a three-dimensional object is an approaching object, and to reduce an erroneous determination that a stationary object is an approaching object and to apply the approaching object warning system. In this case, the output of false alarms for stationary objects can be reduced.

The block diagram which shows embodiment of the approaching object detection apparatus which concerns on this invention The figure which shows a captured image and a solid-object detection result in embodiment of the approaching object detection apparatus which concerns on this invention. The figure which shows the output state of an alarm in embodiment of the approaching object detection apparatus which concerns on this invention. 1st process drawing which shows embodiment of the approaching object detection apparatus and approaching object detection method which concern on this invention 2nd process drawing which shows embodiment of the approaching object detection apparatus and approaching object detection method which concern on this invention 1st explanatory drawing for demonstrating the conventional problem 2nd explanatory drawing for demonstrating the conventional problem

  Hereinafter, an embodiment of an approaching object detection device according to the present invention will be described with reference to FIGS. 1 to 5.

  FIG. 1 is a block diagram showing an approaching object detection device 1 mounted on a host vehicle (vehicle) as a moving body as an embodiment of the approaching object detection device according to the present invention.

  As shown in FIG. 1, the approaching object detection device 1 according to the present embodiment includes an in-vehicle camera 2 as a single imaging unit. This in-vehicle camera 2 is attached to a predetermined position of the host vehicle, and an imaging region within a viewing angle around the host vehicle including a road surface as an imaging target surface is input to an input device (such as an operation button) (not shown). Imaging is performed at a predetermined frame rate by using a user operation or a predetermined driving operation of the vehicle as a start trigger. The road surface includes not only the road surface of the road but also the road surface of the parking lot, the ground, and other travel surfaces on which other vehicles are supposed to travel (the same applies hereinafter). And the approaching object detection apparatus 1 detects the solid object on the road surface in an imaging region based on such a captured image of the vehicle-mounted camera 2. The in-vehicle camera 2 is a wide viewing angle camera including a wide-angle lens such as a fisheye lens, and may be a digital camera including a solid-state imaging device (image plane) such as a CCD or CMOS. The in-vehicle camera 2 has a predetermined image centered on the rear of the host vehicle attached to the rear part of the host vehicle (for example, a rear license garnish unit) in a posture that looks down on the road surface behind the host vehicle from an oblique direction. The back camera which images an area | region may be sufficient. In addition to this, the in-vehicle camera 2 is a predetermined centered around the front of the host vehicle that is attached to the front portion (for example, an emblem portion) of the host vehicle in a posture that looks down on the road surface in front of the host vehicle from an oblique direction. It may be a front camera that images the imaging region. In addition to these, the in-vehicle camera 2 is mounted on the left side of the host vehicle attached to the left side of the host vehicle (for example, the left door mirror) in such a posture as to look down the road surface on the left side of the host vehicle from an oblique direction. It may be a left side camera that images a predetermined imaging area centered. In addition to these, the in-vehicle camera 2 is mounted on the right side of the host vehicle mounted on the right side of the host vehicle (for example, the right door mirror) in such a posture as to look down the road surface on the right side of the host vehicle from an oblique direction. It may be a right side camera that captures a predetermined imaging region at the center.

  As shown in FIG. 1, the approaching object detection device 1 includes a camera image acquisition unit 3, and images captured by the in-vehicle camera 2 are sequentially stored in the camera image acquisition unit 3 for each imaging. It is designed to be entered.

  Furthermore, as shown in FIG. 1, the approaching object detection apparatus 1 has an approaching object candidate detection unit 4 as an approaching object candidate detection unit. The approaching object candidate detection unit 4 sequentially acquires captured images of the in-vehicle camera 2 acquired by the camera image acquisition unit 3 from the camera image acquisition unit 3, and based on the acquired captured images, a road surface as an approaching object candidate. The upper three-dimensional object is detected by image recognition every predetermined time period (hereinafter referred to as a three-dimensional object detection period). The three-dimensional object detection cycle may be, for example, a time required for imaging for a predetermined number of frames by the in-vehicle camera 2 or may be the same time as the imaging cycle. Further, the detection of the three-dimensional object may be performed by a method that can detect which pixel region in the captured image corresponds to the three-dimensional object, and the above-described background difference method and other known three-dimensional object detection methods are applied. can do. When the three-dimensional object is detected in this way, the approaching object candidate detection unit 4 creates a detection image indicating the detected three-dimensional object as a detection result. Note that the approaching object candidate detection unit 4 uses the internal parameters (image center coordinates, lens focal length, lens distortion coefficient, etc.) of the in-vehicle camera 2 stored in advance in the approaching object detection device 1 for the captured image. A detected image may be created based on the captured image after distortion correction performed.

  Here, FIG. 2A shows a captured image after distortion correction including another vehicle on the road surface as an example of a captured image used for detecting a three-dimensional object. However, for the sake of convenience, the captured image shown in FIG. 2 is a limited image after distortion correction of a part (broken line frame portion) of the entire captured image schematically illustrated in FIG. Moreover, FIG.2 (b) shows an example of the detection image applicable to the solid object detected using the captured image of Fig.2 (a). The detected image of FIG. 2B is a captured image after distortion correction at the latest detection time (time) according to the three-dimensional object detection cycle (hereinafter referred to as the latest image). (Abbreviated as a captured image). In addition, the detected image in FIG. 2B is a distortion-corrected captured image (hereinafter, abbreviated as the previous captured image) at the previous detection time point that is earlier than the latest detection time. What has been subjected to transformation (coordinate conversion such as movement, rotation, etc.) according to the amount of movement of the host vehicle during the period from the previous detection point to the latest detection point (hereinafter referred to as the previous captured image taking into account the movement of the host vehicle) It is created using Specifically, as shown in FIG. 2B, the detected image has a predetermined difference (hereinafter referred to as the latest difference) between the latest captured image and the previous captured image in consideration of the movement of the host vehicle. The pixel value is included as a pixel (white portion in the figure). In addition, as shown in the gray portion in FIG. 2B, the detected image is obtained by replacing the past difference (previous difference, previous difference, etc.) obtained by the same procedure as the latest difference with the latest difference. The pixel is included as a pixel whose pixel value decreases as it goes back in the past at a position that does not overlap. In other words, the detected image is an image in which the latest difference is overwritten on the past difference while leaving the past difference in a range where the position (coordinates) does not overlap. More specifically, the pixel corresponding to the past difference shown in the figure becomes gray near black as it goes back to the past (lower pixel value). The reason why past differences are included in the detected image in this way is to increase the accuracy of calculating the center of gravity in consideration of variations in the detection result of the approaching object candidate detection unit 4. However, the extent to which the retroactive difference is included in the detected image can be variously changed according to the detection accuracy, processing speed, etc. of the approaching object candidate detection unit 4. It is also possible to use only the difference. Further, as shown in FIG. 2 (b), for the area where the difference did not appear and the area corresponding to the past difference which was included in the detected image but became old with the passage of time, the image is substantially It is a black part that does not exist.

  Returning to FIG. 1, the approaching object detection device 1 has a center-of-gravity calculation unit 5 as a center-of-gravity calculation means. The centroid calculating unit 5 calculates the centroid of the latest detected image every time the latest detected image (detection result) is created (acquired) by the approaching object candidate detecting unit 4.

  Here, the center of gravity may be calculated by the following equation.

xg = Σ _{(i = 1 to n)} xi × wi / n (1)
yg = Σ _{(i = 1 to n)} yi × wi / n (2)

  However, xg in the equation (1) is the x coordinate in the image coordinate system of the center of gravity. Further, n (natural number) in the expressions (1) and (2) is the number of pixels (number of pixels) corresponding to the difference. Furthermore, xi (integer) in equation (1) is the x coordinate in the image coordinate system for the i-th pixel among the n pixels corresponding to the difference. Furthermore, wi (real number) in the equations (1) and (2) corresponds to the pixel value of the i-th pixel among n pixels corresponding to the difference (the value increases as the pixel value increases). Weight coefficient (0 to 1). This calculation (weighting) using the weighting factor is substantially meaningful when a past difference whose pixel value is lower than the latest difference is also used as the difference, as shown in FIG. When not performing (that is, using only the latest difference), it is only necessary to unify wi = 1. Furthermore, yg in equation (2) is the y coordinate in the image coordinate system of the center of gravity. In addition, yi (integer) in equation (2) is the y coordinate in the image coordinate system for the i-th pixel among the n pixels corresponding to the difference.

  Moreover, as shown in FIG. 1, the approaching object detection apparatus 1 has the movement direction calculation part 6 as a movement direction calculation means. The movement direction calculation unit 6 calculates the movement direction of the center of gravity based on a plurality of successive calculation results of the center of gravity by the center of gravity calculation unit 5.

  Here, the moving direction of the center of gravity may be calculated using the following equation.

Δxg = xg (t) −xg (t−T) (3)
Δyg = yg (t) −yg (t−T) (4)

  However, Δxg in the equation (3) is the displacement amount of the x coordinate of the center of gravity according to the center of gravity calculation cycle. Further, xg (t) in the expression (3) is the x coordinate in the image coordinate system of the center of gravity calculated at a certain calculation time (time) t. Furthermore, T in the equations (3) and (4) is a center of gravity calculation period. Furthermore, xg (t−T) in the expression (3) is the x coordinate in the image coordinate system of the center of gravity calculated the previous time (before the calculation cycle T) of xg (t). In addition, Δyg in the equation (4) is a displacement amount of the y-coordinate of the center of gravity according to the center-of-gravity calculation cycle. Furthermore, yg (t) in equation (4) is the y coordinate in the image coordinate system of the center of gravity calculated at a certain calculation time t. Furthermore, yg (t−T) in the equation (4) is the y coordinate in the image coordinate system of the center of gravity calculated last time of yg (t).

  Then, the movement direction calculation unit 6 may calculate the movement direction of the center of gravity based on the signs of Δxg and Δyg calculated in this way.

  In order to improve the calculation stability of the moving direction, the average value of the center of gravity at a series of a plurality of calculation points may be used for calculating the moving direction. For example, in equation (3), instead of xg (t), an average value of xg (t), xg (t−T) and xg (t−2T) is used, and instead of xg (t−T) Alternatively, an average value of xg (t−T), xg (t−2T), and xg (t−3T) may be used. Similarly, in the formula (4), instead of yg (t), an average value of yg (t), yg (t−T) and yg (t−2T) is used, and yg (t−T) Instead, an average value of yg (t−T), yg (t−2T), and yg (t−3T) may be used.

  Furthermore, as shown in FIG. 1, the approaching object detection device 1 has an approaching object determination unit 7 as an approaching object determination unit. The approaching object determination unit 7 determines whether or not the three-dimensional object detected by the approaching object candidate detection unit 4 is an approaching object approaching the host vehicle. Specifically, the approaching object determination unit 7 determines that the detected three-dimensional object is an approaching object when the movement direction of the center of gravity calculated by the movement direction calculation unit 6 is a direction indicating approach to the host vehicle. On the other hand, when the moving direction of the center of gravity is a direction indicating separation from the host vehicle, it is determined that the detected three-dimensional object is not an approaching object. The determination that the object is not an approaching object may be a determination that the object is a stationary object.

  Furthermore, as shown in FIG. 1, the approaching object detection device 1 has an alarm output unit 8 as an alarm output means. The warning output unit 8 outputs a warning for a three-dimensional object determined to be an approaching object by the approaching object determination unit 7. As shown in FIG. 3, the alarm output is generated based on the captured image and displayed on the approaching object (another vehicle) in the vehicle periphery monitoring image displayed on the display unit 10 or a sound. You may make it perform by the audio | voice output of the alarm sound via the output part 11. FIG. On the other hand, the warning output unit 8 does not output a warning for a three-dimensional object that is determined not to be an approaching object by the approaching object determination unit 7.

  Each of the above-described components 3 to 8 of the approaching object detection device 1 includes a CPU that performs processing corresponding to the function of the approaching object detection device 1, a ROM that stores an execution program of the CPU, and a temporary processing result of the CPU. You may implement | achieve by RAM etc. which are used for a safe preservation | save.

  Here, FIG. 4 shows a case where another vehicle behind the host vehicle is detected based on the captured image of the back camera of the host vehicle as a first operation example of the present embodiment. The operation in the case of a stopped vehicle (stationary object) is shown.

  That is, in step 1 (ST1) of FIG. 4, the captured image of the back camera after distortion correction (the left half of the figure) and this captured image are created by the approaching object candidate detection unit 4 as the latest captured image. A detection image of another vehicle (the right half of the figure, the same applies hereinafter) is shown. In addition, the center of gravity G1 calculated by the center of gravity calculation unit 5 is shown on the detected image.

  Next, in Step 2 (ST2) of FIG. 4, the captured image of the back camera after distortion correction acquired when the three-dimensional object detection cycle has elapsed from the acquisition time of the captured image shown in Step 1 (ST1), and The detected image of the other vehicle created by the approaching object candidate detection unit 4 with the captured image as the latest captured image is shown. Further, the center of gravity G2 calculated by the center of gravity calculating unit 5 is shown on the detected image. Here, in step 2 (ST2), since the area of the detected image is larger in the left direction of the screen than in step 1 (ST1), the center of gravity G2 is further to the left than G1 in the case of step 1 (ST1). It is displaced in the direction.

  Next, in step 3 (ST3) of FIG. 4, the captured image of the back camera after distortion correction acquired when the three-dimensional object detection cycle has elapsed from the acquisition time of the captured image shown in step 2 (ST2), and The detected image of the other vehicle created by the approaching object candidate detection unit 4 with the captured image as the latest captured image is shown. In addition, the center of gravity G3 calculated by the center of gravity calculating unit 5 is shown on the detected image. Here, in step 3 (ST3), the area of the detected image is larger in the left direction of the screen than in step 2 (ST2), so that the center of gravity G3 is more left than G2 in step 2 (ST2). It is displaced in the direction.

  Next, in step 4 (ST4) of FIG. 4, the movement direction of the center of gravity calculated by the movement direction calculation unit 6 based on the series of center of gravity G1 to G3 calculated in step 1 (ST1) to step 3 (ST3). Are shown as movement vectors. As shown in FIG. 4, the movement vector in step 4 (ST4) is a vector in a direction away from the fixed coordinates corresponding to the own vehicle position (camera position) in the image coordinate system.

  In such a case, as shown in the next step 5 (ST5), the approaching object determination unit 7 determines that the object is not an approaching object, and finally, as shown in step 6 (ST6), an alarm is issued. The output of the alarm for the other vehicle by the output unit 8 is prohibited.

  Next, FIG. 5 shows a case where another vehicle behind the host vehicle is detected based on the captured image of the back camera of the host vehicle as a second operation example of the present embodiment. The operation in the case of an approaching vehicle (approaching object) is shown.

  That is, in Step 11 (ST11) of FIG. 5, the captured image of the back camera after distortion correction, and the detected image of the other vehicle created by the approaching object candidate detection unit 4 using this captured image as the latest captured image. It is shown. In addition, the center of gravity G11 calculated by the center of gravity calculation unit 5 is shown on the detected image.

  Next, in step 12 (ST12) of FIG. 5, the captured image of the back camera after distortion correction acquired when the three-dimensional object detection cycle has elapsed from the acquisition time of the captured image shown in step 11 (ST11), and The detected image of the other vehicle created by the approaching object candidate detection unit 4 with the captured image as the latest captured image is shown. Further, the center of gravity G12 calculated by the center of gravity calculating unit 5 is shown on the detected image. Here, in step 12 (ST12), since the area of the detected image is larger in the right direction of the screen than in step 11 (ST11), the center of gravity G12 is more right than G11 in the case of step 11 (ST11). It is displaced in the direction.

  Next, in step 13 (ST13) of FIG. 5, the captured image of the back camera after distortion correction acquired when the three-dimensional object detection cycle has elapsed from the acquisition time of the captured image shown in step 12 (ST12), and The detected image of the other vehicle created by the approaching object candidate detection unit 4 with the captured image as the latest captured image is shown. Further, the center of gravity G13 calculated by the center of gravity calculating unit 5 is shown on the detected image. Here, in step 13 (ST13), the area of the detected image is larger toward the right of the screen than in step 12 (ST12), so that the center of gravity G13 is more right than G12 when in step 12 (ST12). It is displaced in the direction.

  Next, in Step 14 (ST14) of FIG. 5, the movement direction of the center of gravity calculated by the movement direction calculation unit 6 based on the series of center of gravity G11 to G13 calculated in Step 11 (ST11) to Step 13 (ST13). Are shown as movement vectors. As shown in FIG. 5, the movement vector in step 14 (ST14) is a vector in a direction approaching a fixed coordinate corresponding to the own vehicle position (camera position) in the image coordinate system.

  In such a case, as shown in the next step 15 (ST15), the approaching object determination unit 7 determines that the object is an approaching object, and finally, as shown in step 16 (ST16), an alarm is issued. The output unit 8 outputs an alarm for the other vehicle.

  As described above, according to the present invention, it is possible to quickly determine whether or not a three-dimensional object is an approaching object, and to reduce the erroneous determination of a stationary object in the approaching object determination, thereby reducing the approaching object warning. When applied to the system, false alarms targeting stationary objects can be avoided in advance, and as a result, user discomfort due to unnecessary alarms can be reduced.

  In addition, this invention is not limited to embodiment mentioned above, A various change can be made in the limit which does not impair the characteristic of this invention.

DESCRIPTION OF SYMBOLS 1 Approaching object detection apparatus 2 Car-mounted camera 4 Approaching object candidate detection part 5 Center of gravity calculation part 6 Movement direction calculation part 7 Approaching object determination part

Claims (6)

A single imaging means that is arranged at a predetermined position of the moving body and images a predetermined imaging area around the moving body;
Based on the captured image of the imaging means, an approaching object candidate detecting means for detecting a solid object in the imaging region as an approaching object candidate for each predetermined period;
An approaching object detection device comprising: an approaching object determining means for determining whether or not the three-dimensional object detected by the approaching object candidate detecting means is an approaching object approaching the moving body,
Each time the latest detection result of the three-dimensional object is acquired by the approaching object candidate detection means, a centroid calculation unit that calculates the centroid of the three-dimensional object indicated in the latest detection result;
A moving direction calculating means for calculating a moving direction of the center of gravity based on a calculation result of the center of gravity by the center of gravity calculating means;
The approaching object determining means includes
When the moving direction of the center of gravity calculated by the moving direction calculating means is a direction indicating approach to the moving body, it is determined that the moving object is the approaching object, while the moving direction of the center of gravity is An approaching object detection device that determines that the object is not the approaching object when the direction is a direction indicating separation from the moving body. For the three-dimensional object determined to be the approaching object, an alarm is output for the three-dimensional object, and for the three-dimensional object determined not to be the approaching object, the alarm is output for the object. The approaching object detection device according to claim 1, further comprising an alarm output unit that does not perform the operation. The approaching object detection device according to claim 1, wherein the center-of-gravity calculation unit uses the pixel value of the three-dimensional object detected by the approaching object candidate detection unit for the calculation of the center of gravity. A single imaging unit that images a predetermined imaging area around the moving body is arranged at a predetermined position of the moving body, and based on the captured image of the imaging unit, the imaging area as the approaching object candidate in the imaging area An approaching object detection method for detecting a three-dimensional object at predetermined intervals and determining whether or not the detected three-dimensional object is an approaching object approaching the moving body,
A first step of calculating the center of gravity of the three-dimensional object shown in the latest detection result each time the latest detection result of the three-dimensional object is acquired;
A second step of calculating a moving direction of the center of gravity based on the calculation result of the center of gravity in the first step;
When the movement direction of the center of gravity calculated in the second step is a direction indicating approach to the moving body, it is determined that the object is an approaching object, while the movement direction of the center of gravity is And a third step of determining that the object is not the approaching object when the direction is a direction indicating separation from the moving body. For the three-dimensional object determined to be the approaching object in the third step, an alarm is output for this, and for the three-dimensional object determined not to be the approaching object in the third step The approaching object detection method according to claim 4, further comprising a fourth step of not outputting the alarm for the target. The approaching object detection method according to claim 4 or 5, wherein a pixel value of the detected three-dimensional object is used for calculating the center of gravity in the first step.