JP2019105550A - Object detection device, control method and control program for object detection device - Google Patents

Abstract

To provide an object detection device that can distinguish a background even when an object with a low reflectance moves.SOLUTION: The object detection device includes a light emitting/receiving unit 100, a processing unit PROC and a storage unit. The light emitting/receiving unit 100 emits a laser beam into a space while performing a scan, and receives a reflected laser beam from an object. The processing unit PROC obtains a value of distance to an object using a time difference between the emission time of a laser beam and the light receiving time of a laser beam reflected from the object, and generates a distance image showing the distribution of values of distance in a space that is obtained by laser beam scanning. The storage unit stores a background image. The processing unit PROC compares a background image and a present image acquired at a present time point; clusters different regions of the present image from the background image to identify them; checks whether or not there is an overlapped part of the region of an identified object and an infinity region that was present in the background image; determines, when the proportion of the infinity region to the region of an identified object is equal to or more than a predetermined value, that the identified object is an object overlapping with the infinity region that was present in the background image.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an object detection device, a control method of the object detection device, and a control program of the object detection device.

  The object detection apparatus emits laser light toward space, and measures the distance to an object in space from the time from the emission to reception of reflected light. The apparatus using this laser is called laser radar, and it may be called LiDER (Light Detection and Ranging or Laser Imaging Detection and Ranging). In such an object detection apparatus, information on a plurality of directions facing the space can be obtained by scanning the space two-dimensionally by sequentially changing the delivery direction of the laser light. The obtained information is referred to as a distance image, and includes information on the direction of the object as viewed from the laser light transmitting / receiving unit and the distance to the object.

  In such an object detection apparatus, a distance image (background image) serving as a background is obtained in advance, and the background image is compared with the input distance image (current image) to extract and change pixels whose distance has changed. A so-called background subtraction method is used to obtain a region.

  Conventionally, when acquiring this background, it is performed to update the background using data acquired at a constant time cycle (Patent Document 1).

  In this background subtraction method, reflection of laser light is used to obtain a distance image, so it is not possible to recognize an object that does not reflect laser light. For example, if there is an object that does not reflect laser light or an object with low reflectance within the detection range, the laser light does not return from the portion where such an object exists, or can not be detected even if it is reflected. That part is recognized as infinity. For this reason, when the background is updated by the above-mentioned conventional technology, a distance image in which an infinite distance area where the reflection of the laser light can not be detected becomes a background image.

  Then, when an object that does not reflect laser light or an object with a low reflectance moves, the object is detected by comparing the background image with the current image by the background subtraction method, the back of the infinite distance region (laser light emission An object farther from the position is detected as if it appeared there.

  Conventionally, as processing for an object that does not reflect the laser light or an object with a low reflectance, the laser beam is emitted while scanning the laser beam into the monitoring area set to include the crossing path, and no light reception is received when the reflected laser beam is not received. By associating the distance data indicating, an object that does not reflect laser light or an object with a low reflectance is detected. This is because, for example, it is determined that an object that does not reflect such a laser beam or an object with a low reflectance intrudes into the monitoring space when no light is received from the state of light reception (Patent Document 2) .

JP 2005-300259 A JP 2007-126025 A

  However, in the technique of Patent Document 2, since the object is detected by the change from the reception of the reflected laser light to the absence of the light reception, the object or the reflectance that does not reflect the laser light that becomes an infinite distance region in the background image When there is an object with a lower value of, when such an object moves, it becomes from the absence of reception of the reflected laser light to the presence of light reception. For this reason, the movement of an object that does not reflect laser light or an object that does not reflect laser light appears as movement of an object that does not reflect laser light appears as a moving object. There still remains the problem of not being able to distinguish between it and its background.

  The present invention has been made in view of the above circumstances, and it is an object of the present invention to provide an object which does not reflect laser light, which has been present as an infinite area in a background image, or which has low reflectance. The object is to provide an object detection device which can distinguish an object which has been in a far area and its background.

  In addition, another object of the present invention is to provide an object that is an infinite area and its background even when an object that does not reflect laser light, which has been present as an infinite area in a background image, or an object with a low reflectance moves. It is providing the control method of the object detection apparatus which can distinguish.

  Furthermore, another object of the present invention is to provide an object that is an infinite area and its background even when an object that does not reflect laser light, which has been present as an infinite area in a background image, or an object with a low reflectance moves. It is providing the control program of the object detection apparatus which can distinguish.

  The above object of the present invention is achieved by the following means.

(1) an emitting unit for emitting a laser beam;
A scanning unit that scans the laser light in space;
A light receiving unit that receives laser light reflected from an object in the space;
The distance value to the object is determined based on the time difference between the emission time when the laser light is emitted from the emission unit and the light reception time when the laser light reflected from the object is received by the light receiving unit, the distance value by the scanning unit A processing unit that generates a distance image indicating a distribution of distance values in the space obtained by scanning of laser light;
A storage unit that stores a background image acquired based on the distance image generated by the processing unit;
Have
The processing unit compares the background image with a current image which is the distance image acquired at the current time, clusters regions different from the background image in the current image for each region, and recognizes a recognition object. It is extracted and checked for each recognition object whether or not there is a portion overlapping with the infinity region existing in the background image, and the ratio of the infinity region to the region of the recognition object is a predetermined value or more. An object detection device that determines that the recognition object is an object overlapping an infinite distance area in the background image;

  (2) The processing unit is configured to continuously recognize the recognition object of a predetermined number of frames or more among the plurality of frames in which the distance images are arranged in time series, and to select the recognition object of the plurality of frames. The object according to (1), wherein the recognition object is determined to be a moving object when the distance from the position in the frame at the time of first appearance to the position after a predetermined number of frames has passed is a predetermined distance or more Detection device.

(3) The processing unit updates the background image stored in the storage unit at a predetermined timing,
The processing unit stores the recognition object in the storage unit as the background image including the recognition object when the recognition object is not determined to be the moving object by the timing of updating the background image. The object detection apparatus as described in 2.).

  (4) The processing unit adds identification information to indicate that the recognition object is in the infinite distance region when it is determined that the recognition object is an object overlapping the infinite distance region in the background image. The object detection apparatus as described in any one of (1)-(3).

(5) an emitting unit for emitting a laser beam;
A scanning unit that scans the laser light in space;
A light receiving unit that receives laser light reflected from an object in the space;
The distance value to the object is determined based on the time difference between the emission time when the laser light is emitted from the emission unit and the light reception time when the laser light reflected from the object is received by the light receiving unit, the distance value by the scanning unit A processing unit that generates a distance image indicating a distribution of distance values in the space obtained by scanning of laser light;
A storage unit that stores a background image acquired based on the distance image generated by the processing unit;
A control method of an object detection device having
Determining whether there is an area different from the background image in the current image by comparing the background image with the current image which is the distance image acquired at the current time,
If there are different regions in step (a), the different regions are clustered for each region and extracted as a recognition object, and each recognition object has a portion overlapping the infinite distance region that was present in the background image (B) determining whether or not the recognition object is an object overlapping the infinity area in the background image if the ratio of the infinity area to the area of the recognition object is not less than a predetermined value; ,
Control method of an object detection apparatus having:

  (6) After the step (b), among the plurality of frames in which the distance images are arranged in time series, the recognition object is continuously recognized for at least a predetermined number of frames, and among the plurality of frames Determining that the recognition object is a moving object when the distance from the position in the frame when the recognition object first appears to the position after a predetermined number of frames is greater than a predetermined distance (c) The control method of the object detection apparatus according to the above (5), comprising:

(7) The method further includes the step (d) of updating the background image stored in the storage unit at a predetermined timing.
(E) storing the recognition object in the storage unit as the background image including the recognition object when the recognition object is not determined to be the moving object by the timing of updating the background image; 6) A control method of the object detection device according to the above.

  (8) In the step (b), when it is determined that the recognition object is an object overlapping the infinity region in the background image, identification for indicating that the object of the recognition object is in the infinity region The control method of the object detection device according to any one of the above (5) to (7), including the step (f) of adding information.

  (9) A control program of an object detection device which causes a processing unit in the object detection device to execute the control method of the object detection device according to any one of (5) to (8).

  According to the present invention, in the case where the object (recognition object) recognized in the current image is compared with the infinity region existing in the background image, the region of the recognition object includes a predetermined proportion of the infinity region In this case, it is determined that the recognition object is an object overlapping the infinity area in the background image. This makes it possible to distinguish an object that has overlapped with the infinite area in the background image as an object that has appeared as the infinite area moves.

It is a sectional view of a monitoring device concerning an embodiment. It is a block diagram which shows the structure of a process part. It is explanatory drawing for demonstrating the state which scans the inside of the monitoring space of a monitoring apparatus. It is explanatory drawing explaining the state which looked at the monitoring space from the side. It is explanatory drawing explaining the state which looked at the monitoring space from the side, and is a figure for demonstrating the subject at the time of acquiring a distance image. It is explanatory drawing explaining the state which looked at the monitoring space from the side, and is a figure for demonstrating the subject at the time of acquiring a distance image. It is a flowchart which shows the process sequence for determining whether it is an object which overlaps with the infinity area which existed in the background image. FIG. 8 is a flowchart showing a processing procedure for determining whether or not the object is an object overlapping the infinity area existing in the background image, following FIG. 7;

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. In the description of the drawings, the same elements will be denoted by the same reference symbols, without redundant description. Also, the dimensional proportions of the drawings are exaggerated for the convenience of the description, and may differ from the actual proportions.

  In the present embodiment, an example in which the object detection device of the present invention is used as a monitoring device will be described.

  FIG. 1 is a cross-sectional view of the monitoring device according to the present embodiment.

  The monitoring device MD includes an emitting unit LPS that emits a laser beam, a light receiving unit RPS that handles laser light reflection from an object, and a mirror unit MU, which constitute the light emitting and receiving unit 100.

  The emitting unit LPS includes a pulse-type semiconductor laser LD that emits a laser beam, and a collimator lens CL that converts divergent light from the semiconductor laser LD into parallel light.

  The mirror unit MU scans and projects the laser light collimated by the collimator lens CL toward the monitoring space by the rotating mirror surface, and reflects the reflected light from the object. The mirror unit MU is rotated by the motor MT that is rotationally driven.

  The light receiving unit RPS has a lens LS for collecting the light reflected from the object reflected by the mirror unit MU, and a photodiode PD for receiving the light collected by the lens LS. The photodiode PD has a plurality of pixels arranged in the Z direction.

  It is preferable that the optical axes of the light emitting unit LPS and the light receiving unit RPS be orthogonal to the rotation axis RO of the mirror unit MU.

  The monitoring device MD includes a processing unit PROC that obtains distance information according to the time difference between the emission timing of the semiconductor laser LD and the light reception timing of the photodiode PD.

  FIG. 2 is a block diagram showing the configuration of the processing unit PROC.

  The processing unit PROC is a computer, and is provided with a CPU 11, which is an arithmetic unit, a work area, a RAM 12 used for temporary storage, a ROM 13 storing basic programs, and as necessary, stores programs and parameter data. The nonvolatile memory 14 is provided and connected to one another by a bus. In addition, the light emitting and receiving unit 100 is connected to the bus (directly or via an interface for connecting an external device), and under the control of the CPU 11, emission of laser light, reception of light, and control of rotation of the mirror unit are performed. There is. Since these are configurations and functions of known computers, detailed description will be omitted. Further, a display (not shown) may be connected to the monitoring device MD so that a distance image (video) can be viewed. In addition, another computer or the like may be connected so that data from the monitoring device MD can be stored or used.

  Each part is accommodated in a housing CS. The casing CS has a box shape fixed to the rigid wall WL or the like, and includes an upper wall CSa, a lower wall CSb opposite thereto, and a side wall CSc connecting the upper wall CSa and the lower wall CSb. Have.

  An opening CSd is formed in part of the side wall CSc, and a transparent plate TR is attached to the opening CSd.

The mirror unit MU has a shape in which two square pyramids are joined in an opposite direction and integrated, that is, four pairs (but not limited to four pairs) of mirror surfaces M1 and M2 inclined in opposite directions. ) Have.
The mirror surfaces M1 and M2 are preferably formed by vapor deposition of a reflective film on the surface of a resin material (for example, polycarbonate) in the shape of a mirror unit.

  The mirror unit MU is connected to the shaft MTa of the motor MT fixed to the housing CS, and is rotationally driven. The axis (rotational axis) of the axis MTa extends in the Z direction, which is the vertical direction, and the XY plane formed by the X and Y directions orthogonal to the Z direction is a horizontal plane. However, the axis of the axis MTa is vertical You may incline to the direction.

  The object detection principle of the monitoring device MD will be described.

  In FIG. 1, a part of the scanned and projected light flux that strikes the object and is reflected again passes through the transparent plate TR and is incident on the second mirror surface M2 of the mirror unit MU in the housing CS. , And is reflected by the first mirror surface M1, condensed by the lens LS, and detected for each pixel by the light receiving surface of the photodiode PD.

  Then, the processing unit PROC, which is a processing unit, obtains distance information (distance value) according to the time difference between the emission time of the semiconductor laser LD and the light reception time of the photodiode PD. In this manner, the object is detected in the entire area in the monitoring space, and a frame FL (see FIG. 3) as a distance image having distance information for each pixel can be obtained.

  It will be described in more detail.

  The divergent light intermittently emitted in a pulse form from the semiconductor laser LD is converted into a parallel light beam by the collimator lens CL, is incident on the first mirror surface M1 of the rotating mirror unit MU, is reflected there, and is further transmitted After being reflected by the mirror surface M2, the light is transmitted as a laser spot light having, for example, a vertically long rectangular cross section, through the transparent plate TR and directed toward the outside monitoring space.

  The direction in which the emitted laser spot light is reflected by the object and returns as the reflected light is referred to as a light emitting / receiving direction. The laser spot light traveling in the same light emitting and receiving direction is detected by the same pixel.

  FIG. 3 is an explanatory diagram for describing a state in which the inside of the monitoring space of the monitoring device MD is scanned. In FIG. 3, the laser spot beam SB (shown by hatching) is emitted according to the rotation of the mirror unit MU.

  Here, the mirror unit MU serves as a scanning unit, and the crossing angles are different in the combination of the first mirror surface M1 and the second mirror surface M2.

  The laser light is sequentially reflected by the rotating first mirror surface M1 and the rotating second mirror surface M2.

  First, the laser light reflected by the first mirror surface M1 and the second mirror surface M2 of the first pair is horizontally shifted from left to right in the first area Ln1 of the monitoring space according to the rotation of the mirror unit MU. Is scanned.

  Next, the laser light reflected by the first mirror surface M1 and the second mirror surface M2 of the second pair is horizontally oriented from the left in the second region Ln2 from the top of the monitoring space according to the rotation of the mirror unit MU. It is scanned to the right.

  Next, the laser light reflected by the third mirror surface M1 and the second mirror surface M2 of the third pair horizontally from the left in the third region Ln3 from the top of the monitoring space according to the rotation of the mirror unit MU It is scanned to the right.

  Next, the laser light reflected by the fourth mirror surface M1 and the second mirror surface of the fourth pair horizontally from left to right in the lowermost region Ln4 of the monitoring space according to the rotation of the mirror unit MU. It is scanned.

  This completes one scan of the entire space that can be monitored by the monitoring device MD. One frame FL is obtained by combining the images obtained by scanning the regions Ln1 to Ln4.

  Then, after the mirror unit MU makes one rotation, if the first mirror surface M1 and the second mirror surface M2 of the first pair return, again from the first area Ln1 to the lower area Ln4 of the monitoring space The scan is repeated to obtain the next frame FL.

  In FIG. 3, when the object is a white wall surface of the building BL (the incident position P1 of the laser beam), the reflection condition is uniform since the surface is a flat surface having approximately the same reflectance, and the distance value varies from scan to scan. It is hard to occur. Thus, approximately equal distance values can be obtained.

  On the other hand, when the object is a tree or the like (the incident position P2 of the laser light), the reflection condition is not stable because the surface is uneven or the color of the object is locally changed. Therefore, if the distance value is determined based on the reflected light from them, the distance value will vary for each scan, but the distance value can be obtained.

  If the object is an edge etc. of the building BL (the incident position P3 of the laser light), the laser spot light hits the building BL at one time due to the deflection of the mirror unit MU etc. I hit it. In these cases, two groups of data (i.e., a distance value to a building or a distance value to a tree) are obtained for each scan on the same pixel. Thus, a pixel having a distance value of two groups in a plurality of scans may be called a blank pixel. Even in such a case, the distance value can be obtained.

  On the other hand, when the object is a black painted part of the building BL or the like (the incident position P4 of the laser light), the reflected light may not be detected by the same pixel of the photodiode PD depending on the irradiation condition of the laser light. If it can not be detected, the distance value becomes unknown. An area in which a pixel whose distance value is unknown is present is an infinite area. Therefore, the position of the incident position P4 of the laser beam in FIG. 3 is an infinite distance region.

  When the amount of light received by the photodiode PD is equal to or less than a predetermined value, the processing unit PROC sets the pixel as the unknown distance value, and recognizes an area in which such a pixel is present as an infinite area. Since the pixel in the infinity area has no laser light reflection, the distance information (distance value) of the pixel is, for example, sky data, a large distance value which can not be measured in this measurement, or-(minus) as data indicating infinity. A distance value etc. are stored.

  In addition to the black painted part of the building, there are, for example, black cars and objects that transmit laser light (for example, windows made of glass or resin), etc. in addition to the black painted part of the building.

  Next, detection of an object of the monitoring device MD using the background subtraction method will be described. FIG. 4 is an explanatory view illustrating the monitoring space viewed from the side.

  In the background subtraction method, an image as a reference is stored (also referred to as registration) in the RAM 12 as a background image. Therefore, the RAM 12 is a storage unit. Specifically, as preparation for monitoring (preprocessing), registration of the background image is performed at startup of the monitoring device MD and at a predetermined time cycle. The background image ideally scans the laser spot light SB from the monitoring device MD in the absence of moving objects such as cars, humans and animals, as shown in FIG. 4A. At this time, a distance image in which only an object serving as a background (referred to as a background object BG) ideally exists is registered as a background image.

  The background image is newly acquired, for example, at start-up of the monitoring device MD, for a predetermined time period, or at a predetermined timing such as an instruction from the user, and stored in the RAM 12. If the background image is already stored in the RAM 12, it is updated by the newly acquired distance image.

  At the time of monitoring, a distance image at the current time is acquired. This is called the current image. For example, as shown in FIG. 4B, when the intruder (person) OBJ appears in front of the background object BG, the reflected light RB2 from the intruder OBJ is newly generated. In the figure, an example of a car is shown as the intruder OBJ, but it is not limited to a car, and may be, for example, a bicycle, a person, an animal or the like.

  The processing unit PROC compares the background image acquired in FIG. 4A with the current image acquired in FIG. 4B, and if there is a different area, the different area is basically set as the background. Is recognized as an object (foreground) that did not exist. Here, it is recognized as an intruder OBJ. If the intruder OBJ is recognized, for example, an alarm is issued or a moving object is tracked. For the alarm, an alarm device (a lamp or a sound) may be connected to the processing unit PROC, and the alarm device may be activated from the processing unit PROC by recognizing the intruder OBJ. The movement tracking is performed by clustering the intruder OBJ, and if the intruder OBJ is moving, the intruder OBJ is tracked because the position of the intruder OBJ clustered in multiple frames obtained by repeating scanning changes. To determine the movement direction and speed.

  Next, the subject at the time of acquiring a distance image is demonstrated in detail.

  FIG. 5 and FIG. 6 are explanatory diagrams for explaining a state in which the monitoring space is viewed from the side, and are diagrams for describing a problem when acquiring a distance image.

  In the case of FIG. 5, there is almost no reflection of the laser light in front of the automobile OB1 (reflecting object) with high laser light reflectivity (closer to the surveillance apparatus MD when viewed from the surveillance apparatus MD). Hereinafter, it is a case where the low reflective object is parked.

  In such a case, the processing unit PROC recognizes that portion as an infinite distance region because there is no reflection from the black automobile OB2. When there is a low reflection object that is recognized as this infinite area, it happens that when it comes to timing to acquire a background image, there is an infinite area in the background image stored at that timing.

  From this state, as shown in FIG. 6, when the car OB2 moves, the reflected light RB2 from the car OB1 hidden from the car OB2 as viewed from the monitoring device MD is detected. Then, since the reflected light RB2 from the automobile OB1 is detected in the part of the background image at infinity, it is determined that the automobile OB1 has entered at that time. However, in actuality, since only the automobile OB1 parked before the background image acquisition time is detected, there is no intruder at least compared to the background image acquisition time. Therefore, determining that the vehicle OB1 is an intruder at this point is an erroneous determination.

  In the present embodiment, when an infinite area is registered in the background image, when an object with high reflectance appears in the current image, the infinite area in which the appearing object was present in the background image It is determined whether the object overlaps with the object.

  FIG. 7 and FIG. 8 are flowcharts showing processing procedures for determining whether or not the object is an object overlapping the infinite area existing in the background image. The processing procedure is implemented by the processing unit PROC executing a program created as a control method of the monitoring device MD. In this processing procedure, not only it is determined whether or not the object overlaps the infinite area existing in the background image, and it is also determined whether or not the overlapped object is a moving object.

  First, the processing unit PROC registers a background image (S10). The registration of the background image in S10 is, for example, at the time of activation of the monitoring device MD. In addition, registration of the background image for every fixed period will be performed in a later step in this procedure.

  Subsequently, the processing unit PROC acquires a current image (S11).

  Subsequently, the processing unit PROC compares the registered background image with the current image using the background subtraction method, and if there is an area different from the background image in the current image, extracts the different area (S12).

  Subsequently, the processing unit PROC clusters the extracted different regions (S13). When there are a plurality of different areas, clustering is performed for each area. As a result, objects that were not present in the background image by the background subtraction method, that is, objects that appeared in the current image are clustered and recognized. Here, this clustered object is called a recognition object. When there are a plurality of clustered objects, there are also a plurality of recognition objects.

  Subsequently, the processing unit PROC determines, for each of the recognized objects, whether or not a portion overlapping the infinity area in the background image is included in the area at a predetermined ratio or more (S14).

  In the determination of S14, first, the pixel of the recognition object in the current image and the pixel in the area at the same position in the background image are compared, Determine if Then, if there is a pixel in the infinite area, the number of pixels is counted to obtain a ratio to the number of pixels of the recognition object. Here, for example, when 80% (80%) or more of the pixels of the recognition object overlap the pixel at infinity, it is determined that the recognition object in the current image is an object overlapping the infinity region in the background image. That is, YES is obtained in S14. Of course, this judgment ratio is arbitrary, and may be, for example, 90% (90%) or more or 100% (100%). Conversely, it may be determined that the object overlaps the infinite distance area even if there are few portions overlapping the infinite distance area, such as 30% (30%) or more and 50% (50%) or more.

  This ratio is, for example, when there is a wide infinity area in the background image, for example, when the sky is always visible, when the recognition object is compared with the infinity object which is the sky, the recognition object and the infinity object are compared. The percentage of area overlap is small. On the other hand, as shown in FIG. 5, in the case of the black car OB2, since the car OB1 behind appears when it moves, the area of the car OB1 which is a recognition object and the car OB2 which is an infinite area at the background image. Will be compared. In such a case, the ratio of the overlap becomes large, and is 100% here. As described above, it can be determined that the larger the ratio, the higher the probability that the object is behind the infinite area and the object which has been in the infinite area is moved.

  If the processing unit PROC determines that the region of the recognition object and the infinity region overlap with each other by a predetermined ratio or more in S14 (S14: YES), the recognition object is an object overlapping the infinity region of the background image (S16) . An object overlapping such an infinite distance region is referred to as an infinite distance region overlapping object. As a result, since the low reflection object which has been in the infinity region is not moving, the object behind it is recognized as an object overlapping the infinity region in the current image.

  Subsequently, the processing unit PROC adds identification information to the recognized object which is determined to be an infinite area overlapping object (S17). The processing unit PROC attaches identification information to this recognition object and stores it. Also, when displaying a recognized object that is considered to be an infinite area overlapping object on the screen, for example, a line surrounding the recognized object (the color of the line is changed from another object) or an arrow is attached Also, it is preferable to color or distinguish the recognition object itself.

  If NO is determined in S14, and after S16, the processing unit PROC subsequently calculates the center position of the recognized object (calculates the center position of the object from the clustering data), and identifies the identified object with the identification code ( ID) is assigned (S15).

  Subsequently, the processing unit PROC determines whether the recognition object is already determined to be a moving object (S18). Here, when it is determined that the recognition object is already a moving object (S18: YES), the processing unit PROC continues the recognition object to be a moving object (S25). In the recognition object which is a moving object, the ID of the recognition object is stored in association with the fact that the object is a moving object (described later).

  If it is determined that the recognition object is not a moving object in S18 (S18: NO), the processing unit PROC determines whether the recognition object is continuously detected until the number of frames reaches a predetermined number (S19). ). For example, if there is the same cluster in the previous frame and the object clustered in the frame in this step, it is determined to be the same object. Then, it is determined whether or not the cluster has continued for a predetermined number of frames. The predetermined number of frames is, for example, two frames or three frames. The predetermined number of frames is the number of frames for confirming whether or not it is a moving object.

  If the processing unit PROC determines that the recognition object is not continuously detected until the predetermined number of frames is reached in S19 (S19: NO), the processing returns to S11. Here, it may be time to periodically update the background image until the predetermined number of frames elapses. However, as described above, the predetermined number of frames is, for example, 2 frames or 3 frames. Therefore, since the time until the predetermined number of frames is reached is small, it is safe to return to S11 and continue the process (if restart or the like, the process is restarted from S10).

  If it is determined in S19 that the recognition object is continuously detected until the predetermined number of frames is reached (S19: YES), the processing unit PROC then determines whether the recognition object is an infinite distance area overlapping object. To do (S20). This determination is made based on whether or not identification information as an object overlapping the infinity area is attached in S17.

  If the processing unit PROC determines in S20 that the object is not an infinite area overlapping object (S20: NO), it determines the recognition object as a moving object (S25). In this step, if the recognition object is not an infinite area overlap object, it is an object that has moved from somewhere, that is, a moving object, since the object is an object that did not exist in the background image. The recognition object determined to be a moving object for the first time in this step S25 is stored as a moving object linked to the ID.

  If the processing unit PROC determines that the recognition object is an infinite area overlapping object in S20 (S20: YES), then the processing unit PROC determines that the position at which the predetermined number of frames of the recognition object has elapsed is the first appearance It is determined whether or not a predetermined distance or more has been moved from the position of (S21). Here, if it is determined that the recognition object has moved a predetermined distance or more from the first appearance (S21: YES), the recognition object is determined as a moving object (S25). This is because, in this step, that the recognition object is moving by a predetermined distance or more when the predetermined number of frames has passed since the first appearance, it means that the recognition object is moving, so it is a moving body. The processing unit PROC associates the ID of the recognition object determined to be a moving object for the first time in this step S25 with the fact that it is a moving object and stores it. The predetermined distance varies depending on the object to be detected as a moving object. For example, in the case of a human being, it is preferable that the monitoring target (object) be about several cm to 50 cm. In the case of an automobile having a high speed, it may be about 10 cm to 1 m.

  If it is determined that the recognition object has not moved by a predetermined distance or more from the first appearance in S21 (S21: NO), the processing unit PROC then determines whether it is time to update the background image (S22). The timing of the background image update in this step is to register the background image at a constant cycle described above. If the processing unit PROC determines in S22 that it is not the timing of updating the background image (S22: NO), it determines that the recognition object is a stationary object (S23). On the other hand, if the processing unit PROC determines that it is the timing of updating the background image in S22 (S22: YES), it includes the recognition object and registers it as a background image (S24). Coming up to this step, stationary objects that are not moving objects and are not classified as stationary objects will be registered as a background image.

  After S23, S24, and S25, the process returns to S11 to repeat the process.

  As described above, according to the present embodiment, the region of the object (recognized object) recognized by the background subtraction method in the current image includes a predetermined proportion or more of the infinity region in the background image. The object to be recognized was determined to be an object overlapping with an infinite area in the background image. As a result, since the recognition object appearing in the current image overlaps the infinity area in the background image by a predetermined ratio or more, it can be determined that the object hidden behind appears due to the movement of the infinity area. Therefore, according to the present embodiment, it is possible to distinguish an object hidden in an object regarded as an infinite region in the background image.

  Furthermore, in the present embodiment, after it is determined whether the object (recognized object) recognized by the background difference method is an object hidden behind an infinite distance area, is it moved after a predetermined number of frames have elapsed since the first appearance? By determining whether it is, it is determined whether an object hidden behind an infinite area is a moving object or a stationary object. Thereby, in the monitoring apparatus, even if an object behind an infinite area is detected, it is not determined as a moving body unless the object is moving. Therefore, for example, after an object such as a window glass or a black car that does not return reflected light is registered in the background image, even if the window glass or the black car moves, the stop object behind them (stationary object) However, moving objects can be detected only for moving objects without detection of moving objects. As a result, it is possible to prevent a false alarm on a car that is only stopped.

  Although the embodiments to which the present invention is applied have been described above, the present invention is not limited to these embodiments. The present invention can be variously modified based on the configuration described in the claims, and is also within the scope of the present invention.

11 CPU,
12 RAM,
13 ROM,
14 non-volatile memory,
100 emitting and receiving units,
CL collimate lens,
CS case,
CSa top wall,
CSb lower wall,
CSc side wall,
CSd opening,
FL frame,
FS fence,
LD semiconductor laser,
LPS emitting part,
LS lens,
M1, M2 mirror surface,
MD monitoring device,
MT motor,
MTa axis,
MU mirror unit,
OBJ intruders,
OB1 laser light reflective car (reflective object),
OB2 black car (low reflection object),
PD photodiode,
PROC processing unit,
RB1, RB2 reflected light,
RO rotation axis,
RPS light receiver,
SB laser spot light.

Claims (9)

An emitting unit for emitting laser light;
A scanning unit that scans the laser light in space;
A light receiving unit that receives laser light reflected from an object in the space;
The distance value to the object is determined based on the time difference between the emission time when the laser light is emitted from the emission unit and the light reception time when the laser light reflected from the object is received by the light receiving unit, the distance value by the scanning unit A processing unit that generates a distance image indicating a distribution of distance values in the space obtained by scanning of laser light;
A storage unit that stores a background image acquired based on the distance image generated by the processing unit;
Have
The processing unit compares the background image with a current image which is the distance image acquired at the current time, clusters regions different from the background image in the current image for each region, and recognizes a recognition object. It is extracted and checked for each recognition object whether or not there is a portion overlapping with the infinity region existing in the background image, and the ratio of the infinity region to the region of the recognition object is a predetermined value or more. An object detection device that determines that the recognition object is an object overlapping an infinite distance area in the background image;   The processing unit is configured such that the recognition object is continuously recognized as a predetermined frame or more among a plurality of frames in which the distance images are arranged in time series, and the recognition object is first displayed among the plurality of frames. The object detection apparatus according to claim 1, wherein the recognition object is determined to be a moving object when the distance from the position in the frame to the position after a predetermined number of frames has elapsed is a predetermined distance or more. The processing unit updates the background image stored in the storage unit at a predetermined timing.
The processing unit, when there is the recognition object not determined to be the moving object by the timing of updating the background image, stores the recognition object in the storage unit as the background image. The object detection apparatus according to claim 1.   The processing unit adds identification information to indicate that the recognition object is in an infinite distance region, when it is determined that the recognition object is an object overlapping the infinite distance region in the background image. The object detection device according to any one of 3. An emitting unit for emitting laser light;
A scanning unit that scans the laser light in space;
A light receiving unit that receives laser light reflected from an object in the space;
The distance value to the object is determined based on the time difference between the emission time when the laser light is emitted from the emission unit and the light reception time when the laser light reflected from the object is received by the light receiving unit, the distance value by the scanning unit A processing unit that generates a distance image indicating a distribution of distance values in the space obtained by scanning of laser light;
A storage unit that stores a background image acquired based on the distance image generated by the processing unit;
A control method of an object detection device having
Determining whether there is an area different from the background image in the current image by comparing the background image with the current image which is the distance image acquired at the current time,
If there are different regions in step (a), the different regions are clustered for each region and extracted as a recognition object, and each recognition object has a portion overlapping the infinite distance region that was present in the background image (B) determining whether or not the recognition object is an object overlapping the infinity area in the background image if the ratio of the infinity area to the area of the recognition object is not less than a predetermined value; ,
Control method of an object detection apparatus having:   After the step (b), the recognition object is continuously recognized for a predetermined number of frames or more among the plurality of frames in which the distance images are arranged in time series, and the recognition is performed among the plurality of frames. Determining that the recognition object is a moving object, when the distance from the position in the frame when the object first appears to the position after a predetermined number of frames has exceeded the predetermined distance, The control method of the object detection apparatus according to claim 5. The method further comprises the step (d) of updating the background image stored in the storage unit at a predetermined timing.
When the recognition object is not determined to be the moving object by the timing of updating the background image, the method further includes the step (e) of storing the recognition object and the background image as the background image. 6. The control method of the object detection apparatus according to 6.   If it is determined in the step (b) that the recognition object is an object overlapping the infinity area in the background image, identification information is added to indicate that the object of the recognition object is in the infinity area. The control method of the object detection device according to any one of claims 5 to 7, comprising the step (f).   A control program of an object detection device for causing a processing unit in an object detection device to execute the control method of an object detection device according to any one of claims 5 to 8.