JP7152961B2 - Parking space recognition device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a parking space recognition device for detecting a parking position of a vehicle from images captured by a camera mounted on the vehicle.

In order to reduce the burden on the driver who parks the vehicle, there has been proposed a parking assistance device that detects a parking space in which the vehicle can be parked and controls the vehicle so that the vehicle is parked within the detected parking space. ing.

For example, Patent Literature 1 discloses a parking guidance device that detects an empty parking space from an image captured by a camera installed in a vehicle. When a parking marker having a predetermined pattern is detected from an image captured by a camera, the parking guidance device determines that the vehicle can be parked in a parking section where the detected parking mark is drawn.

JP 2017-27405 A

As described above, the conventional parking assist system detects a parking space from each of four images captured by four cameras installed in front, rear, right, and left of the vehicle. In other words, the parking assist system has a problem that the processing load for detecting parking spaces increases in proportion to the increase in the number of cameras.

SUMMARY OF THE INVENTION An object of the present invention is to provide a parking space recognition device capable of reducing the processing load when detecting a parking space.

In order to solve the above-mentioned problems, a first invention is a parking space recognition device, which includes an image acquiring unit, a detection setting unit, a detection area designating unit, and a parking space detecting unit. The image acquisition unit uses a plurality of cameras including a left side camera that captures the left side of the vehicle, a right side camera that captures the right side of the vehicle, and a rear camera that captures the rear side of the vehicle. Get multiple images generated. A detection setting unit selects a camera to be used for detecting a parking space from among a plurality of cameras based on the running state of the vehicle, and selects a detection mode for detecting a parking space from an image generated by the selected camera. is selected from the first mode and the second mode based on the running state of the vehicle. When the detection setting unit selects the first mode, the detection area designating unit designates a previously designated first detection area in the image generated by the camera selected from among the plurality of images obtained by the image obtaining unit. If the detection setting unit selects the second mode, a second detection area narrower than the first detection area is specified in the image generated by the selected camera. When the detection setting unit selects the first mode, the parking space detection unit detects the parking space from the first detection area specified by the detection area specifying unit with a first accuracy, and the detection setting unit selects the second mode. If selected, the parking space is detected with the second accuracy from the second detection area designated by the detection area designation unit.

According to the first invention, both the detection mode and the image of the parking space to be detected are changed according to the running state of the vehicle. Thereby, the detection load of the parking space can be reduced.

A second invention is the first invention, wherein the parking space detector includes a first detector and a second detector. The detection setting unit allocates one or more cameras among the plurality of cameras to each of the first detection unit and the second detection unit. The first detection section and the second detection section detect parking spaces in parallel in time.

According to the second invention, the parking spaces can be detected from a wide range around the vehicle by detecting the parking spaces in parallel with each other in terms of time by the first detection unit and the second detection unit.

A third invention is the second invention, wherein the detection setting unit assigns the left side camera and the right side camera to the first detection unit, and assigns the rear camera to the second detection unit. The first detection unit alternately repeats a process of detecting a parking space from the image generated by the left side camera and a process of detecting a parking space from the image generated by the right side camera. A second detector detects a parking space from the image generated by the rear camera.

According to the third invention, the left side camera and the right side camera are assigned to the first detection section, and the rear camera is assigned to the second detection section. As a result, parking spaces positioned to the sides and rear of the vehicle can be detected in real time.

A fourth invention is the second or third invention, wherein the detection setting unit, when the speed of the vehicle is within a predetermined speed range and the steering angle of the vehicle is smaller than a predetermined angle, The side camera and the right side camera are assigned to the first detection unit, and the rear camera is assigned to the second detection unit. When the speed of the vehicle is within a predetermined speed range and the steering angle is equal to or greater than a predetermined angle, the detection setting unit selects the camera positioned outside the turning trajectory of the vehicle out of the left side camera and the right side camera. The first detector is assigned, and the rear camera is assigned to the second detector. The first detection unit and the second detection unit detect parking spaces from the frames generated by the cameras assigned by the detection setting unit.

According to the fourth invention, when the speed of the vehicle is within the predetermined speed range and the steering angle of the vehicle is smaller than the predetermined angle, the left side camera and the right side camera are assigned to the first detection unit, Assign the rear camera to the second detector. As a result, when the vehicle is parked in a parking space by backward parking, an empty parking space can be quickly detected. Further, when the speed of the vehicle is within a predetermined speed range and the steering angle is greater than or equal to a predetermined angle, the camera positioned outside the turning trajectory of the vehicle out of the left side camera and the right side camera is the first detection. and the rear camera is assigned to the second detection unit. As a result, it is possible to detect with high accuracy a parking section in which the vehicle is highly likely to park.

A fifth invention is any one of the first to fourth inventions, wherein the parking space detection unit includes a boundary line detection unit and a space determination unit. The boundary line detection unit selects a target pixel in the specified first detection region or the specified second detection region, and determines whether or not an edge is formed by the selected target pixel. , to detect a boundary line that is at least a part of the contour of the lane marking the extent of the parking bay. The section determination unit determines a parking section based on the detected boundary line. The selection interval of the pixels of interest when the second mode is selected is narrower than the selection interval of the pixels of interest when the first mode is selected.

According to the fifth invention, when the second mode is selected, the parking space can be detected with higher accuracy than when the first mode is selected.

A sixth invention is any one of the first to fifth inventions, wherein when the first mode is selected, the parking space detection unit detects at least a part of the contour of the parking space indicating the range of the parking space. A boundary line is detected from the image with a first accuracy, and a parking reference position is determined based on the detected boundary line. When the second mode is selected, the detection area designating section designates the second detection area based on the already detected end points of the boundary line. When the second mode is selected, the parking space detection unit detects the boundary line with the second accuracy from the second detection area designated by the detection area designation unit, and determines the parking reference position based on the detected boundary line. do. The parking reference position is a reference position for stopping the vehicle within the parking space.

According to the sixth invention, when the second mode is selected, the parking section detection unit designates the detection area based on the already detected parking reference position, and based on the boundary line detected from the detection area. Determine the parking reference position. Thereby, the parking reference position of the parking space detected in the first mode can be determined with higher accuracy.

A seventh invention is a parking space recognition method, comprising a) step, b) step, c) step, and d) step. a) step is performed by a plurality of cameras including a left side camera capturing a left side view of the vehicle, a right side camera capturing a right side view of the vehicle, and a rear camera capturing a rear side view of the vehicle; Get multiple images generated. b) for selecting a camera to be used for parking space detection from among the plurality of acquired images based on the running state of the vehicle, and for detecting the parking space from the image generated by the selected camera; detection mode is selected from the first mode and the second mode based on the running state of the vehicle. c) designating a first pre-specified detection area in an image generated by the selected camera of the plurality of acquired images if the first mode is selected and the second mode is selected; If so, specify a second detection area in the image generated by the selected camera that is smaller than the first detection area. d) detecting a parking space from a first specified detection area with a first accuracy if the first mode is selected; and detecting from a second specified detection area if the second mode is selected; Parking spaces are detected with a second precision.

A seventh invention is used for the first invention.

ADVANTAGE OF THE INVENTION According to this invention, the parking space recognition apparatus which can reduce the processing load at the time of detecting a parking space can be provided.

1 is a functional block diagram showing the configuration of a parking assistance system that includes a parking space recognition device according to an embodiment of the present invention; FIG. 2 is a diagram showing a world coordinate system set by the parking space recognition device shown in FIG. 1; FIG. 2 is a diagram showing an example of a parking space recognized by the parking space recognition device shown in FIG. 1; FIG. 2 is a functional block diagram showing the configuration of the parking space recognition device shown in FIG. 1; FIG. FIG. 2 is a functional block diagram showing the configuration of a parking section detection unit shown in FIG. 1; 2 is a diagram showing an example of changes in cameras selected by the parking space recognition device 1 shown in FIG. 1; FIG. 2 is a diagram showing an example of a detection area designated by a frame generated by the camera shown in FIG. 1; FIG. 3 is a diagram showing another example of a detection area designated by a frame generated by the camera shown in FIG. 1; FIG. 2 is a flow chart showing the operation of the parking space recognition device shown in FIG. 1; 3 is a diagram showing an example of a setting table stored in a memory shown in FIG. 1; FIG. It is the figure which expanded a part of parking lot shown in FIG. 2 is a diagram showing an example of a frame generated by the left side camera shown in FIG. 1; FIG. FIG. 10 is a flowchart of a parking section detection process shown in FIG. 9; FIG. 6 is a diagram showing an example of edge lines generated by the boundary line detection unit shown in FIG. 5; FIG. FIG. 6 is a diagram showing an example of a parking section detected by the section determination unit shown in FIG. 5; FIG. FIG. 5 is a diagram showing an example of a partition list shown in FIG. 4; FIG. FIG. 14 is a flowchart of the partition list update process shown in FIG. 13; FIG. 5 is a diagram showing a detection area based on a detailed mode specified by a detection area specifying unit shown in FIG. 4; FIG. 3 is a diagram showing an example of a parking route of the vehicle shown in FIG. 2; FIG. 3 is a diagram showing another example of a parking route of the vehicle shown in FIG. 2; FIG. FIG. 4 is a diagram showing a modified example of the parking section shown in FIG. 3; FIG. 4 is a diagram showing another modified example of the parking space shown in FIG. 3; FIG. 23 is a diagram showing another method of determining the parking reference position shown in FIG. 22;

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The same reference numerals are given to the same or corresponding parts in the drawings, and the description thereof will not be repeated.

{1. Configuration of Parking Support System 100}
{1.1. overall structure}
FIG. 1 is a functional block diagram showing the configuration of a parking assistance system 100 including a parking space recognition device 1 according to an embodiment of the invention. Referring to FIG. 1, parking assistance system 100 is mounted on a vehicle such as an automobile. Parking assistance system 100 assists the driver in parking the vehicle in a parking bay. A definition of a parking space will be described later.

The parking assistance system 100 includes a parking space recognition device 1 , a vehicle control device 2 , four cameras 3 , a display device 4 , a ranging sensor 5 and a parking instruction switch 6 .

The parking space recognition device 1 and the vehicle control device 2 are CAN (Controller Area Network) nodes and are connected via a bus 7 . Although not shown in FIG. 1 , the bus 7 is also connected to an ECU (Electronic Control Unit) other than the parking space recognition device 1 and the vehicle control device 2 .

When the parking space recognition device 1 detects a parking space from the frames generated by the four cameras 3, it notifies the vehicle control device 2 of the detected parking space. When receiving a desired parking space from the vehicle control device 2, the parking space recognizing device 1 generates image data including the received desired parking space. The parking space recognition device 1 outputs the generated video data to the display device 4 . The desired parking space is the parking space selected by the driver, as described below.

The vehicle control device 2 selects a desired parking space from the parking spaces notified by the parking space recognition device 1 according to the driver's operation. The vehicle control device 2 outputs the selected desired parking space to the parking space recognition device 1 . The vehicle control device 2 assists the driver's parking operation by controlling the steering angle of the front wheels when moving the vehicle into the selected desired parking space.

Each of the cameras 3 has a lens and an imaging device, and photographs the surroundings of the vehicle. Cameras 3 include a front camera 3F, a rear camera 3B, a left side camera 3L, and a right side camera 3R. Each of the cameras 3 generates a moving image including a plurality of frames, and supplies the generated moving image to the parking space recognition device 1 frame by frame.

The display device 4 is, for example, a liquid crystal display device. The display device 4 receives the image data from the parking space recognition device 1 and displays an image around the vehicle based on the received image data.

The ranging sensor 5 is, for example, an ultrasonic sonar. Although only one range sensor is shown in FIG. 1 , the parking assistance system 100 actually includes a plurality of range sensors 5 . The distance measuring sensor 5 is arranged on the side surface or the rear end surface of the vehicle. A distance sensor 5 detects the distance from the vehicle to the obstacle.

The parking instruction switch 6 is used when the driver instructs the parking assistance system 100 to start parking assistance.

{1.2. Definition of world coordinate system}
FIG. 2 is a diagram showing a world coordinate system set by the parking space recognition device 1. As shown in FIG. Referring to FIG. 2, vehicle 9 is equipped with parking assistance system 100 shown in FIG. In the world coordinate system, the origin Ow of the world coordinate system is the middle point of the rotation axis of the rear wheels of the vehicle 9 .

The X-axis is a straight line that extends in the straight-ahead direction of the vehicle 9 and passes through the center of gravity of the vehicle 9 . The positive direction of the X-axis is the direction from the rear end surface of the vehicle 9 toward the front end surface. The Y-axis extends in a direction perpendicular to both the straight-ahead direction and the vertical direction of the vehicle 9 and coincides with the rotational axis of the rear wheels of the vehicle 9 . The positive direction of the Y-axis is the direction from the left side to the right side of the vehicle 9 . That is, the origin Ow is the midpoint of the rotation axis of the rear wheels of the vehicle 9 .

{1.3. Definition of terms}
The terms used to describe the parking space recognition device 1 will be described below. FIG. 3 is a diagram showing an example of a parking lot in which parking spaces are set. Although the parking section L is actually in contact with the white line W, the parking section L is shown separated from the white line W in FIG. 3 for convenience of explanation. The same applies to figures other than FIG.

A parking section L is an area set for parking a vehicle in a parking lot. The parking space recognition device 1 recognizes the parking space L regardless of whether or not the vehicle is parked. In this embodiment, the parking bay L is rectangular and the long side of the parking bay L is perpendicular to the track A. As shown in FIG.

The white line W is a line for indicating the range of the parking section L to the driver. In this embodiment, an example in which the marking lines are white will be described, but the marking lines may not be white.

A boundary line B is a longitudinal side of the white line W and defines the width of the parking space L. As shown in FIG. The boundary line B is part of the long side of the parking space L and part of the long side of the white line W. In other words, the boundary line B is part of the outline of each of the parking space L and the white line W. In FIG. 3, some of the boundary lines B are labeled.

The parking reference position S is set as a guideline for stopping the vehicle 9 when the vehicle 9 is parked in one parking section L. Specifically, the parking reference position S is the position of the origin Ow when the vehicle 9 has completed parking in one parking section L. As shown in FIG. When the origin Ow coincides with the parking reference position S and the X axis of the world coordinate system, which will be described later, coincides with the center line of the parking section L, the parking assist system 100 performs the process of parking the vehicle 9 in the parking section L. judged to be complete. The parking reference position S is set for each parking section L regardless of whether or not another vehicle is parked in the parking section L.

The parking space recognition device 1 uses the parking reference position and the starting point of the boundary line B as parameters indicating the position of the parking space L. FIG. The starting point of one boundary line B is the end point closer to the track A, out of the two end points of the one boundary line B.

{1.4. Arrangement of camera 3}
Referring to FIG. 2, the front camera 3F is provided in the vicinity of the license plate attachment position at the front end of the vehicle 9, and its optical axis 3Fa is oriented in the positive direction of the X axis. The rear camera 3B is provided near the license plate mounting position at the rear end of the vehicle 9, and its optical axis 3Ba is directed in the negative direction of the X axis. The mounting positions of the front camera 3F and the rear camera 3B are desirably on the X-axis, but they may be positioned slightly off the X-axis.

The front camera 3F captures a scene in front of the vehicle 9 to generate a front frame 31F, and supplies the generated front frame 31F to the parking space recognition device 1. - 特許庁The rear camera 3B captures the scenery behind the vehicle 9 to generate a rear frame 31B, and supplies the generated rear frame 31B to the parking space recognition device 1. - 特許庁

The left side camera 3L is provided on the left door mirror 93, and its optical axis 3La is directed in the negative direction of the Y axis. The left side camera 3L captures the scenery in the left direction of the vehicle 9 to generate a left frame 31L, and supplies the generated left frame 31L to the parking space recognition device 1.

The right side camera 3R is provided on the right door mirror 93, and its optical axis 3Ra is directed in the positive direction of the Y axis. The right side camera 3R captures a rightward view of the vehicle 9 to generate a right frame 31R, and supplies the generated right frame 31R to the parking space recognition device 1. FIG.

A lens of the camera 3 is a wide-angle lens having an angle of view of 180° or more. Therefore, the parking assistance system 100 is capable of photographing the entire surroundings of the vehicle 9 .

{1.5. Configuration of parking space recognition device 1}
FIG. 4 is a functional block diagram showing the configuration of the parking space recognition device 1 shown in FIG. Referring to FIG. 4, parking space recognition device 1 includes image acquiring unit 11, detection setting unit 12, detection area designating unit 14, parking space detecting unit 15, display control unit 16, and memory 17. Prepare.

The image acquisition unit 11 acquires frames from each of the cameras 3 . The image acquisition unit 11 outputs frames captured by the camera selected by the detection setting unit 12 among the acquired frames to the parking space detection unit 15 .

The detection setting unit 12 selects two cameras to be used for detecting the parking space L from among the cameras 3 based on the running state of the vehicle 9 . The running state includes, for example, the speed of the vehicle 9, the rotation angle of the steering wheel of the vehicle 9, the traveling direction of the vehicle 9, the position of the vehicle 9, and the like. The traveling direction indicates whether the vehicle 9 is moving forward or backward. The detection setting unit 12 notifies the image acquiring unit 11 and the detection area designating unit 14 of the selected camera.

The detection setting unit 12 also determines the detection mode of the parking space detection unit 15 based on the running state of the vehicle 9 . Specifically, the detection setting unit 12 determines one of the simple mode and the detailed mode as the detection mode of the parking space detection unit 15 based on the running state of the vehicle 9 . The detailed mode is more accurate in detecting parking spaces than the simple mode. Further, the simple mode detects the parking space L in a wider range than the detailed mode. Specifically, when the simple mode is selected, the detection area designating unit 14 designates the detection area L previously designated in the area table 32, and the parking space detection unit 15 detects the parking space from the designated detection area. Detect L. On the other hand, when the detailed mode is selected, the detection area specifying unit 14 specifies the detection area based on the parking reference position S that has already been detected. Thereby, the parking reference position S of the parking section L detected in the detailed mode can be determined with higher accuracy. Details of the simple mode and the detailed mode will be described later.

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera selected by the detection setting unit 12 . The detection area is determined based on the detection mode of the parking space detector 15 .

When the detection mode is the simple mode, the parking space detection unit 15 detects the parking space L from the frame received from the image acquisition unit 11 with the first accuracy. When the detection mode is the detailed mode, the parking space detection mode detects the parking space L from the frame received from the image acquisition unit 11 with the second accuracy. Specifically, the parking section detection unit 15 detects the boundary line B from the detection area designated by the detection area designation unit 14, and detects the parking section L based on the detected boundary line B. FIG. The parking space detection unit 15 generates a space list 18 in which the detected parking spaces are recorded, and outputs the list to the vehicle control device 2 .

When the vehicle 9 moves backward, the display control unit 16 generates a parking assistance image including the rear frame 31B and outputs the generated parking assistance image to the display device 4 . The display control unit 16 acquires the traveling direction of the vehicle 9 from, for example, a shift position sensor (not shown).

The memory 17 is a non-volatile storage device and stores a partition list 18, a setting table 19, and an area table 32. FIG. The section list 18 records the parking section L detected by the parking section detection unit 15 . The setting table 19 records the camera and detection mode used when detecting the parking space L. FIG. The area table 32 records the detection areas specified when the parking area detection unit 15 detects the parking area L in the simple mode.

The parking space recognition device 1 selects at least one camera from among the cameras 3 based on the running state of the vehicle 9, and detects the parking space L from the frame generated by the at least one selected camera. As a result, the parking space recognition device 1 can prevent the detection accuracy of the parking space L from deteriorating even when all the cameras cannot be used to detect the parking space L.

The parking space recognition device 1 also changes the detection mode based on the running state of the vehicle 9 . As a result, the parking space recognition device 1 can reduce the load of detection processing of the parking space L without lowering the detection accuracy of the parking space L when the vehicle 9 is parked.

{1.6. Configuration of Parking Section Detection Unit 15}
FIG. 5 is a functional block diagram showing the configuration of the parking space detector 15 shown in FIG. 4. As shown in FIG. Referring to FIG. 5 , parking space detector 15 includes a first detector 21 and a second detector 22 .

The first detection unit 21 and the second detection unit 22 detect the parking space L from the frame. The first detection unit 21 and the second detection unit 22 are microcontrollers, and each can detect the parking space L independently. In addition, the first detection unit 21 and the second detection unit 22 can reconfigure the connections and settings of the programmable FPGA (Field Programmable Gate Array) and the circuit cells inside the integrated circuit after the integrated circuit is manufactured. Any reconfigurable processor may be used.

The first detection unit 21 detects a parking space from a frame generated by one of the two cameras selected by the detection setting unit 12 . First detection unit 21 includes boundary line detection unit 211 , section determination unit 212 , list management unit 213 , and position correction unit 214 .

The boundary line detection unit 211 detects the boundary line B from the frame generated by one of the cameras, and outputs the position of the detected boundary line B to the section determination unit 212 . The position of boundary B is described in the world coordinate system and includes the positions of the two endpoints that boundary B has.

The section determination unit 212 receives the position of the boundary line B from the boundary line detection unit 211 and determines the parking section L based on the received position of the boundary line B. FIG. The section determination unit 212 outputs the determined position of the parking section L to the list management unit 213 . The position of the parking bay L is described in the world coordinate system and includes the parking reference position S and the positions of the starting points of each of the two boundary lines.

The list management unit 213 updates the positions of the parking spaces recorded in the space list 18 based on the positions of the parking spaces L determined by the space determination units 212 and 222 .

The position correction unit 214 corrects the position of the parking space L added to the space list 18 based on the amount of movement of the vehicle 9 during the period from when the frame was generated by the camera 3 to the current time. Further, the position correction unit 214 corrects the position of the parking section L already recorded in the section list 18 based on the amount of movement of the vehicle 9 during the period from the previous correction time to the current time. The amount of movement of the vehicle 9 is specified based on the speed of the vehicle 9 and the rotation angle of the steering wheel of the vehicle 9 . The position correction unit 214 stores the partition list 18 in which the position of the parking partition L has been corrected in the memory 17 and outputs it to the vehicle control device 2 .

The second detection unit 22 detects the parking space L from the frame generated by the other of the two cameras selected by the detection setting unit 12 . The second detection section 22 includes a boundary line detection section 221 and a section determination section 222 .

The boundary line detection unit 221 detects the boundary line B from the frame generated by the other camera. The section determining section 222 determines the parking section L based on the position of the boundary line B detected by the boundary line detecting section 221 and outputs the determined position of the parking section L to the list managing section 213 .

Although FIG. 5 illustrates an example in which the first detection unit 21 includes the list management unit 213 and the position correction unit 214, the present invention is not limited to this. The second detector 22 may include a list manager 213 and a position corrector 214 . In this case, when the load of the first detection unit 21 is higher than the predetermined threshold, the second detection unit 22 updates the section list 18 and detects the parking information recorded in the section list 18 instead of the first detection unit 21 . Correct the position of section L.

{2. Outline of operation}
FIG. 6 is a diagram showing changes in cameras used by the parking space recognition device 1. FIG. Referring to FIG. 6, it is assumed that vehicle 9 moves from position Pa to position Pb. Shooting areas 7L, 7R, and 7B shown in FIG. 6 indicate areas that can be shot by each of the left-side camera 3L, right-side camera 3R, and rear camera 3B for convenience. not shown.

The parking space recognition device 1 switches the cameras used for detecting the parking space L according to the running state of the vehicle 9 . For example, when the vehicle 9 is moving forward at the position Pa and the speed of the vehicle 9 is 10 km/h, the parking space recognition device 1 selects the left side camera 3L and the right side camera 3R, the left side frame 31L and the right side frame 31L. A parking space L is detected from the square frame 31R. When the vehicle 9 moves forward and the speed of the vehicle 9 is 10 km/h, it is unlikely that the vehicle 9 will immediately start reverse parking. The parking space recognition device 1 uses a left side camera 3L and a right side camera 3R in order to detect the parking space L from a wide area in the left-right direction of the vehicle 9 .

At position Pb, when the vehicle 9 is moving forward, the speed of the vehicle 9 is 2 km/h, and the steering wheel of the vehicle 9 is turned 250° to the right, the parking space recognition device 1 detects the left side camera 3L and the rear A parking space L is recognized using the camera 3B. In this case, there is a high possibility that the vehicle 9 will reverse after stopping and move into the parking space L located to the left rear of the vehicle 9 . The parking space recognition device 1 uses the left side camera 3L and the rear camera 3B to detect the parking space L from the left rear area of the vehicle 9. As shown in FIG. That is, the parking space recognition device 1 selects a camera that captures an area outside the turning locus of the vehicle 9 from among the cameras 3 .

Thus, the parking space recognition device 1 selects a camera to be used for detecting the parking space L according to the running state of the vehicle 9 . As a result, the parking space recognition device 1 can reduce the processing load of detecting the parking space L. FIG. In addition, the parking space recognition device 1 detects the parking space L according to the position of the vehicle 9, the speed of the vehicle 9, the rotation angle of the steering wheel of the vehicle 9, the traveling direction of the vehicle 9, the position of the vehicle 9, and the like. You can change the image used.

Further, the parking space recognition device 1 changes the contents of the detection processing of the parking space L based on the running state of the vehicle 9 . Specifically, the parking space recognition device 1 determines one of the simple mode and the detailed mode as the parking space detection mode based on the running state of the vehicle 9 . The parking space recognition device 1 detects the parking space L from the frame generated by the selected camera in the determined detection mode.

The parking space recognition device 1 designates an area of the parking space L to be detected. The detection area specified by the frame differs between the detection mode and the detail mode.

FIG. 7 is a diagram showing an example of the detection area designated by the left frame 31L when the simple mode is set. Referring to FIG. 7, when the simple mode is set, the parking space recognition device 1 sets an area 31K in which the vehicle body of the vehicle 9 is not reflected in the left frame 31L as a detection area.

FIG. 8 is a diagram showing an example of the detection area designated by the left frame 31L when the detailed mode is set. Referring to FIG. 8, when the detailed mode is set, parking space recognition device 1 sets a plurality of areas 31N where white lines W are assumed to appear in left frame 31L as detection areas. The area 31N is designated based on the past detection results of the parking space L. FIG.

That is, the simple mode detects the parking lot L in a wider range than the detailed mode, and detects the parking lot L more roughly than the detailed mode. When the simple mode is determined, it becomes easier for the parking space recognition device 1 to select a new parking space. On the other hand, in the detailed mode, the parking space L is detected in a narrower range than in the simple mode, and the parking space L is detected with higher precision than in the simple mode. Therefore, the processing load of the parking space recognition device 1 is the same in both the simple mode and the detailed mode.

Thus, the parking space recognition device 1 uses either the simple mode or the detailed mode according to the running state of the vehicle 9 . As a result, the processing load can be reduced without lowering the detection performance of the parking space L.

{3. Operation of Parking Support System 100}
{3.1. Operation of parking space recognition device 1}
FIG. 9 is a flow chart showing the operation when the parking space recognition device 1 shown in FIG. 1 detects the parking space L. As shown in FIG. The process shown in FIG. 9 is started when the speed of the vehicle 9 is equal to or less than a preset start reference value. A starting reference value is, for example, 20 km/h. At the time when the processing shown in FIG. 9 is started, the camera has not been selected and the detection mode has not been determined.

The parking space recognition device 1 may start the processing shown in FIG. 9 regardless of the speed of the vehicle 9 when the ignition switch is turned on.

Referring to FIG. 9, detection setting unit 12 acquires the running state of vehicle 9 (step S11). For example, the detection setting unit 12 acquires the speed of the vehicle 9 from a speed sensor (not shown) and acquires the rotation angle of the steering wheel of the vehicle 9 from a rotation angle sensor (not shown). Further, the detection setting unit 12 acquires the traveling direction of the vehicle 9 from a shift position sensor (not shown).

The detection setting unit 12 determines whether or not the speed of the vehicle 9 obtained in step S11 is greater than the start reference value (step S12). If the speed of the vehicle 9 is greater than the start reference value (Yes in step S12), the parking space recognition device 1 ends the processing shown in FIG. On the other hand, if the speed of the vehicle 9 is equal to or less than the start reference value (No in step S12), the detection setting unit 12 proceeds to step S13.

The detection setting unit 12 selects two cameras to be used for detecting the parking space L from among the cameras 3 based on the driving state acquired in step S11 (step S13). At step S13, the setting table 19 stored in the memory 17 is referred to. The number of cameras selected is less than the number of cameras 3 . Thus, even if the parking space detector 15 cannot simultaneously detect the parking space L from the images generated by the cameras 3, the parking space can be detected.

FIG. 10 is a diagram showing an example of the setting table 19 stored in the memory 17. As shown in FIG. Referring to FIG. 10, for example, when vehicle 9 is moving forward at a speed of 10 km/h, detection setting unit 12 determines that the running state of vehicle 9 corresponds to mode ID "1", and the left side camera Select 3L and right side camera 3R. Based on the setting table 19, the left side camera 3L is assigned to the first detection section 21, and the right side camera 3R is assigned to the second detection section 22. FIG. That is, the detection setting unit 12 determines the frame in which the parking section L is detected by each of the first detection unit 21 and the second detection unit 22 in step S13.

The detection setting unit 12 determines the detection modes of the first detection unit 21 and the second detection unit 22 based on the running state acquired in step S11 (step S14). At step S14, the setting table 19 stored in the memory 17 is referred to. Referring to FIG. 10 , for example, when the running state of vehicle 9 corresponds to mode ID “1”, detection setting unit 12 sets the detection modes of both first detection unit 21 and second detection unit 22 to simple mode. mode.

The detection area specifying unit 14 specifies the detection area of the frame in which each of the first detection unit 21 and the second detection unit 22 detects the parking section L (step S15). For example, when the first detection unit 21 detects the parking space L from the left frame 31L, the detection area designated by the left frame 31L differs between the simple mode and the detailed mode. The details of step S15 will be described later.

The parking space detection unit 15 acquires the frame generated by the camera selected in step S13 from the image acquisition unit 11, and detects the parking space L from the detection area designated in the acquired frame (step S16). Each of the first detection unit 21 and the second detection unit 22 detects the parking space L from the detection area of the obtained frame according to the detection mode determined in step S14. The details of step S16 will be described later.

The parking section detection unit 15 updates the section list 18 based on the detection result of the parking section L in step S16 (step S17). Details of step S17 will be described later. The parking section detection unit 15 outputs the updated section list 18 to the vehicle control device 2 .

{3.2. Parking section detection (step S16)}
Before describing the specification of the detection area (step S15) in detail, the detection of the parking space L (step S16) will be described. When specifying the detection area in the detailed mode, the detection area specifying unit 14 uses the parking reference position S of the parking section L detected in the past and the position of the starting point of the boundary line B. FIG. For this reason, first, a method for specifying the positions of the parking reference position S and the starting point of the boundary line B will be specifically described.

Detection of the parking space L will be described below by taking as an example the case of detecting the parking space L using the left frame 31L generated by the left side camera 3L.

FIG. 11 is an enlarged view of part of the parking lot shown in FIG. Referring to FIG. 11, it is assumed that vehicle 9 is traveling along track A of a parking lot in the direction of arrow C1. The parking space recognition device 1 detects parking spaces L1 to L3 from the left frame 31L generated by the left side camera 3L. The parking space recognition device 1 detects the parking space L2 even when another vehicle is parked in the parking space L2.

The white lines W forming the parking section L1 are referred to as white lines W1 and W2. The white lines W forming the parking section L3 are referred to as white lines W3 and W4. Parking section L2 is formed by white line W2 and white line W3.

FIG. 12 is a diagram showing an example of the left frame 31L generated by the left side camera 3L. Referring to FIG. 12, the positions of pixels in left frame 31L are expressed in a two-dimensional coordinate system with the upper left vertex as origin Oi. The x-axis extends rightward from the origin Oi and the y-axis extends downward from the origin Oi. A coordinate system defined by the left frame 31L is referred to as an "image coordinate system".

(Selection of target pixel)
FIG. 13 is a flowchart of the parking space detection process (step S16). Referring to FIG. 13, boundary detection unit 211 selects a pixel of interest from the pixels of left frame 31L (step S201). The simple mode and the detailed mode have different selection criteria for the pixel of interest.

In the simple mode, the boundary detection unit 211 intermittently selects the pixel of interest. For example, the boundary detection unit 211 selects a target pixel at a rate of 1 out of 5 pixels in the x-axis direction, and selects a target pixel at a rate of 1 out of 3 pixels in the y-axis direction. In other words, the boundary detection unit 211 thins out four pixels in the x-axis direction before selecting the target pixel, and thins out two pixels in the y-axis direction before selecting the target pixel. By intermittently selecting the pixel of interest in the simple mode, it is possible to prevent an increase in the processing load caused by detecting the parking space L from a wide area.

In the detailed mode, the boundary detection unit 211 selects all pixels included in the detection area as pixels of interest. As a result, the boundary line detection unit 211 can accurately detect the direction and the position of the starting point of the boundary line B, which will be described later, so that the detection accuracy of the parking section L can be improved.

Note that the boundary line detection unit 211 does not have to select all the pixels included in the detection area as pixels of interest in the detailed mode. It suffices that the interval for selecting target pixels in detailed mode is narrower than the interval for selecting target pixels in simple mode.

(detection of edge points)
The boundary line detection unit 211 identifies adjacent pixels adjacent to the target pixel selected in step S201. The boundary line detection unit 211 detects an edge point based on the absolute value of the difference between the brightness of the target pixel and the brightness of the adjacent pixels (step S202). An edge point is a general term for plus edge points and minus edge points. A plus edge point is a point where the luminance increases in the positive direction of the x-axis or the y-axis and the calculated absolute difference value is greater than a preset edge reference value. A minus edge point is a point where the luminance decreases in the positive direction of the x-axis or the y-axis and the calculated absolute difference value is greater than the edge reference value.

The boundary detection unit 211 transforms the positions of the detected edge points from the image coordinate system to the world coordinate system (step S203).

Boundary line detection unit 211 repeats steps S201 to S203 until the selection of the target pixel is completed (Yes in step S204).

(Generation of edge lines)
The boundary line detection unit 211 generates an edge line by connecting the edge points converted into the world coordinate system in step S203 (step S205). An edge line is a general term for a plus edge line and a minus edge line. A positive edge line is generated by connecting two or more positive edge points arranged at a distance shorter than a preset connection reference value. A negative edge line is generated by connecting two or more negative edge points arranged at a distance shorter than the connection reference value. The edge line constitutes part of the outline of the white line W.

The boundary line detection unit 211 identifies, among the generated plus edge lines, those plus edge lines having bends larger than a preset bend reference value. The boundary line detection unit 211 cuts the specified plus edge line at a plus edge point where a bend larger than the bend reference value occurs. The bend reference value is, for example, 10°. The boundary line detection unit 211 generates linear edge lines in order to detect the boundary line B, which is the contour of the white line W and the parking section L in the long side direction.

FIG. 14 is a diagram showing an example of edge lines generated in step S205. In FIG. 14, the width of the white lines W1 to W3 is exaggerated, and the display of the white line W4 and the parking spaces L1 to L3 is omitted.

In the example shown in FIG. 14, the plus edge line CP1 is generated by connecting the plus edge points PE1 to PE4. Similarly, plus edge lines CP2 and CP3 are generated. In the plus edge lines CP2 and CP3, the display of the plus edge points other than the plus edge points PE5 and PE6 is omitted.

A minus edge line CM2 is generated by connecting the minus edge points ME1 to PE4. Similarly, minus edge lines CM1 and CM3 are generated. In FIG. 14, the minus edge points forming the minus edge lines CM1 and CM3 are omitted except for the minus edge points ME5 and ME6.

Although the edge lines shown in FIG. 14 have the same length, the edge lines generated in step S205 do not necessarily have the same length. When the entire white line W2 appears in the left frame 31L like the white line W2 in FIG. 12, the boundary line detection unit 211 can detect an edge line corresponding to the length of the long side of the white line W2. However, when part of the white line W3 is hidden by another vehicle like the white line W3 shown in FIG. Detect lines. In this case, the lengths of the edge lines generated in step S205 do not match.

(Determination of Parking Section L)
The partition determination unit 212 acquires edge lines generated by the boundary line detection unit 211 . The division determining unit 212 generates pairs of parallel plus edge lines and minus edge lines from the acquired edge lines (step S206).

Specifically, the division determination unit 212 selects one plus edge line and one minus edge line that are adjacent to each other. If the angle formed by the two selected edge lines is smaller than a preset parallel reference value, the partition determination unit 212 determines that the two selected edge lines are parallel, and determines that the two selected edge lines are paired. to generate The parallel reference value is, for example, 2 degrees.

The division determination unit 212 determines the position of the starting point of each of the two edge lines forming the pair generated in step S206 (step S207). For example, when the vehicle 9 is outside the parking section L, the section determination unit 212 determines the end point closer to the vehicle 9, of the two end points of the positive edge line forming the pair, as the starting point of the positive edge line. The starting points of negative edge lines forming a pair are similarly determined.

Also, when the vehicle 9 is in the parking section L, the section determination unit 212 determines the end point with the larger X coordinate among the two end points of the positive edge line forming the pair as the starting point of the positive edge line. The starting points of negative edge lines forming a pair are similarly determined.

The section determination unit 212 determines whether or not the distance between the two edge lines forming a pair corresponds to the width of the parking section L (step S208). Specifically, the division determination unit 212 calculates the distance from the starting point of the positive edge line to the starting point of the negative edge line as the distance between the two edge lines forming a pair. The section determination unit 212 determines whether the calculated distance is within a predetermined range corresponding to the width of the parking section L. The predetermined range is determined based on the width of an ordinary passenger car, and is, for example, 2-2.5 m.

If the calculated distance is within the predetermined range (Yes in step S208), section determination unit 212 determines that parking section L has been detected based on the two edge lines forming a pair (step S209). . In other words, the section determination unit 212 determines that the two edge lines forming a pair are the boundary line B of the parking section L. FIG.

In this way, when the two detected boundary lines B are parallel, the section determination unit 212 determines that the two boundary lines are the outline of the parking section L based on the interval between these two boundary lines B. determines whether or not to form In other words, the section determination unit 212 detects the parking section based on the positional relationship between the detected two boundary lines. Since it is not necessary to use the length of the two boundary lines to detect the parking space, the parking space can be detected even if part of the white line is hidden by an obstacle. For example, parking spaces where other vehicles are already parked can be detected.

FIG. 15 is a diagram showing an example of detection of the parking section L. As shown in FIG. Referring to FIG. 15, division determination unit 212 generates a pair of plus edge line CP1 and minus edge line CM2 because adjacent plus edge line CP1 is parallel to minus edge line CM2. A plus edge point PE1 is determined as the starting point of the plus edge line CP1. The minus edge point ME1 is determined as the starting point of the minus edge line CM2. A distance D1 from the plus edge point PE1 to the minus edge point ME1 is within a predetermined range. Therefore, the section determination unit 212 determines that the parking section L1 has been detected based on the plus edge line CP1 and the minus edge line CM2. The plus edge line CP1 and the minus edge line CM2 are boundary lines B in the long side direction of the parking section L1.

In addition, since the adjacent plus edge line CP1 and minus edge line CM1 are parallel, the division determination unit 212 forms a pair formed by the plus edge line CP1 and the minus edge line CM1. A minus edge point ME5 is determined as the starting point of the minus edge line CM1. A distance D1 from the plus edge point PE1 to the minus edge point ME5 is outside the predetermined range. This is because the distance D1 is the width of the white line W1, which is significantly different from the width of an ordinary automobile. Therefore, the section determination unit 212 determines that the pair of the plus edge line CP1 and the minus edge line CM1 does not form the parking section L.

Similarly, the section determination unit 212 detects the parking section L2 based on the plus edge line CP2 and the minus edge line CM3. A parking space L3 shown in FIG. 11 is also detected in the same manner.

(Determination of parking reference position S)
The section determination unit 212 determines the parking reference position S of the parking section L detected in step S209 (step S210). Referring to FIG. 15, step S210 will be described by taking the case of determining the parking reference position S1 of the parking section L1 as an example.

The section determination unit 212 determines the center line K1 of the parking section L1 based on the plus edge line CP1 and the minus edge line CM2 that form the boundary line B of the parking section L1. Specifically, the division determination unit 212 identifies a line segment E1 connecting a positive edge point PE1, which is the starting point of the positive edge line CP1, and a negative edge point ME1, which is the starting point of the negative edge line CM2. The section determination unit 212 determines a straight line passing through the midpoint M1 of the identified line segment E1 and parallel to either one of the plus edge line CP1 and the minus edge line CM2 as the center line K1 of the parking section L1.

The section determination unit 212 determines a point on the center line K1 located in the parking section L1 and at a position of a length Lg from the midpoint M1 as the parking reference position S1 of the parking section L1. The length Lg is determined based on the vehicle length of the vehicle 9 . The length Lg is, for example, the distance from the front end surface of the vehicle 9 to the rotation axis of the rear wheels of the vehicle 9 .

The section determination unit 212 determines the reliability of the detected parking section L (step S211). The reliability indicates the probability of the detected parking space L1, and is a numerical value of 0 or more and 1 or less, for example. The closer the reliability is to 1, the more likely it is that the detected parking space L exists.

For example, the confidence is determined based on the length of each of the two edge lines used to identify the parking bay L. The longer each of the two edge lines used to identify parking bay L1, the closer the confidence is to one. Alternatively, the confidence level may be determined based on the density of each of the two edge lines used to identify the parking bay L. For example, the shorter the interval between plus edge points or minus edge points forming an edge line, the closer the reliability is to one. That is, when one detection area is detected in both the detailed mode and the simple mode, the reliability of the detailed mode is higher than the reliability of the simple mode. This is because the detailed mode selects target pixels more densely than the simple mode.

The section determination unit 212 generates section data specifying the detected parking section L (step S212). The parking space data includes the parking reference position S of the detected parking space L, the positions of the starting points of the two edge lines specifying the parking space L, and the reliability of the parking space L. FIG. The partition determination unit 212 outputs the generated partition data to the list management unit 213 .

The section determining unit 212 determines whether or not the identification of the pair formed by the two edge lines has ended (step S213). When the identification is finished (Yes in step S213), the partition determination unit 212 finishes the processing shown in FIG. If the identification has not ended (No in step S213), the partition determining unit 212 returns to step S206.

The second detection unit 22 performs the same processing as the first detection unit 21 to detect the parking lot L from the input frame, and outputs the block data specifying the detected parking lot L to the list management unit 213. do.

{3.3. Section list update (step S17)}
The section determination units 212 and 222 output section data indicating the detected parking section L to the list management section 213 . The list management section 213 updates the section list 18 based on the section data received from the section determination sections 212 and 222 (step S17 shown in FIG. 9).

FIG. 16 is a diagram showing an example of the partition list 18. As shown in FIG. Referring to FIG. 16, one line of data (record) is section data indicating each parking section L. In FIG. The section list 18 records an ID, detection time, parking reference position, first start point, second start point, and reliability. The ID is an identification number that uniquely identifies the parking space L recorded in the space list 18 . The detection time is the generation time of the frame in which the parking space L was detected. The parking reference position, first start point, second start point, and reliability are data recorded in the section data generated in step S212. The parking reference position, first start point, and second start point are described in the world coordinate system.

FIG. 17 is a flow chart showing the operation of the list management unit 213. As shown in FIG. When the list management unit 213 receives the division data from both the division determination units 212 and 222, it starts the processing shown in FIG. Hereinafter, step S17 will be described in detail, taking as an example the case where the list management unit 213 updates the section list 18 based on the section data of the parking section L1.

First, the list management unit 213 multiplies all reliability levels recorded in the partition list 18 by a coefficient smaller than 1 (step S251). As a result, the list management unit 213 reduces all reliability levels recorded in the partition list 18 . The reason for executing step S251 will be described later.

The list management unit 213 determines whether or not the parking reference position S1 recorded in the section data has been recorded in the section list 18 (step S252). For example, the list management unit 213 calculates the distance from the parking reference position of ID1 to the parking reference position S1 recorded in the section data. If the calculated distance is within the predetermined range indicating the same position, the list management unit 213 determines that the parking reference position S1 recorded in the section data has already been recorded in the section list 18 (Yes in step S252). ).

In this case, the list management unit 213 compares the reliability recorded in the partition data with the reliability of ID1 recorded in the partition list 18 (step S253).

If the reliability recorded in the section data is equal to or higher than the reliability of ID1 (Yes in step S254), list management unit 213 overwrites the record of ID1 with the section data of parking section L1 (step S255).

On the other hand, if the reliability recorded in the section data is lower than the reliability of ID1 (No in step S254), the list management unit 213 retains the record of ID1 as it is (step S258).

Returning to the description of step S252. If the parking reference position S1 is not recorded in the section list 18 (No in step S252), the list management unit 213 adds the section data of the parking section L1 to the section list 18 (step S257).

The list management unit 213 corrects the parking reference position S, the first start point, and the second start point of each parking section recorded in the section list 18 based on the running state of the vehicle 9 (step S256). The overwritten record and the newly added section data are corrected based on the amount of movement of the vehicle 9 during the period from frame generation to the current time. Records that have not been updated in the section list 18 are corrected based on the amount of movement of the vehicle 9 from the previous update time of the section list 18 to the current time. When the vehicle 9 moves, errors in the parking reference position S, the first start point, and the second start point recorded in the section list 18 increase. The list management unit 213 suppresses errors by correcting the parking reference position S, the first start point, and the second start point recorded in the section list 18 . The list management unit 213 outputs the corrected section list 18 to the vehicle control device 2 .

In addition, the list management unit 213 deletes the record having the parking reference position S at a predetermined distance from the origin Ow when executing step S256. This is because it is unlikely that the vehicle 9 will park in the parking section L located away from the vehicle 9 .

The reason for decreasing the reliability recorded in the partition list 18 in step S251 will be explained. If the record in the partition list 18 is not overwritten, the process of multiplying the confidence of this record by a factor less than one is repeated. As a result, the reliability of records that are not overwritten approaches zero. Since it is possible to prevent the parking lot L detected in the past from being continuously recorded in the block list, the accuracy of the parking lot L can be increased.

{3.4. Designation of detection area (step S15)}
(simple mode)
The detection setting unit 12 notifies the detection area designating unit 14 of the frame to be processed by the first detection unit 21 and the detection mode of the first detection unit 21 . Based on the notification from the detection setting unit 12, the detection area designation unit 14 designates the detection area of the frame processed by the first detection unit 21 (step S15 shown in FIG. 9). The detection area designation unit 14 also designates detection areas of frames processed by the second detection unit 22 in the same manner.

Designation of the detection area in the simple mode will be described below by taking as an example a case where the first detection unit 21 detects the parking space L from the left frame 31L in the simple mode. The same applies to designation of the detection area when the detection mode of the second detection unit 22 is the simple mode.

The area table 32 records in advance simple mode detection areas to be set for frames generated by each of the cameras 3 . The detection area in simple mode differs depending on the camera assigned to the first detection unit 21 . This is because the shooting range of each camera differs depending on the mounting position and optical axis direction of each camera.

When the left side camera 3L is assigned to the first detection unit 21, the detection area specifying unit 14 acquires the simple mode detection area associated with the left side camera 3L from the area table 32, and sets the acquired detection area to It is designated as the detection area of the left frame 31L. The detection area designation unit 14 notifies the boundary line detection unit 211 of the designated detection area of the left frame 31L. The boundary line detection unit 211 selects a pixel of interest from pixels within the notified detection area (step S201 shown in FIG. 13).

As a result, as shown in FIG. 7, the simple mode detection area 31K is specified in the left frame 31L. In the example shown in FIG. 7, an area where the vehicle body of the vehicle 9 does not appear in the left frame 31L is designated as the detection area. Thereby, the detection load of the parking space L in the first detection unit 21 can be reduced. Note that the detection area specifying unit 14 may specify, as the detection area, an area in which the road surface is assumed to appear in the left frame 31L.

The detection area specified in simple mode is specified in the image coordinate system. Therefore, the detection area specifying unit 14 outputs the detection area acquired from the area table 32 to the first detection unit 21 as it is.

(advanced mode)
The operation of the detection area designating section 14 will be described by taking as an example the case where the detailed mode is set as the detection mode of the first detecting section 21 . Assume that the detection setting unit 12 has assigned the left side camera 3L to the first detection unit 21 .

In this case, the detection area designating unit 14 designates the detection area based on the parking reference position S, the position of the first start point, and the position of the second start point of the parking section L recorded in the section list 18 . 18A and 18B are diagrams for explaining a method of setting the detection area in the detailed mode.

For example, when the parking section L1 is registered in the section list 18, the detection area specifying unit 14, based on the parking reference position S1, the position of the first start point, and the position of the second start point of the parking section L1, Two detection areas G1 and G2 are set. Specifically, the detection area specifying unit 14 specifies the midpoint M1 of the line segment connecting the starting point PE1 and the starting point ME1. The detection area specifying unit 14 specifies a straight line connecting the specified midpoint M1 and the parking reference position S1 as the center line K1 of the parking section L1.

The detection area specifying unit 14 sets a rectangular detection area G1 including the starting point PE1 of the edge line and having the long side in the direction in which the center line K1 extends. Further, the detection area specifying unit 14 sets a rectangular detection area G2 including the starting point ME1 of the edge line and having the long side in the direction in which the center line K1 extends. Thereby, two detection areas G1 and G2 based on the parking section L1 are set.

The lengths of the short sides of the detection areas G1 and G2 are set based on the general width of the white line W. The length of the long side of the detection areas G1 and G2 is set to 1.5 m, for example. That is, the detection area set in the detailed mode does not include the entire white line W. This is because the parking section L is specified based on the parking reference position S and the starting point of the boundary line B that forms the outline of the white line W, so it is not necessary to detect the entire white line W. As a result, it is possible to reduce the area of the detection area in the detailed mode. Even when all the pixels in the detection area are selected as pixels of interest, the parking section L can be detected quickly.

Although not shown in FIG. 18, the detection area designating unit 14 designates a detection area corresponding to the parking section L2. Specifically, the detection area corresponding to the parking section L2 is determined based on the parking reference position S2 of the parking section L2 and the positions of the two starting points (positive edge point PE5 and negative edge point ME6). As a result, the detection area G2 overlaps the detection area corresponding to the parking section L2. This is because the minus edge point ME1 and the plus edge point PE5 are the vertex of the same white line. In this case, the detection area designation unit 14 notifies the first detection unit 21 of the sum area of the two detection areas G1 and G2.

The detection area specified in detail mode is described in the world coordinate system. Therefore, the detection area specifying unit 14 transforms the coordinate system of the specified detection area from the world coordinate system to the image coordinate system of the left frame 31L. The detection area specifying unit 14 outputs the detection area converted into the image coordinate system to the first detection unit 21 .

In the example shown in FIG. 8, the hatched area 31N is the detection area designated in the detailed mode. The hatched area 31N is set so as to include end points near the vehicle 9 among the end points of the white line. The detection area specified in detailed mode is narrower than the detection area set in simple mode. The detailed mode contains fewer pixels in the detection area than the simple mode. When the detailed mode is set, all the pixels included in the detection area are set as pixels of interest, so the detection accuracy of the parking space L can be improved.

{4. First example of running state of vehicle 9}
FIG. 19 is a diagram showing a first example of a parking route for the vehicle 9 shown in FIG. Referring to FIG. 19, vehicle 9 is parked in parking section La by traveling along the route indicated by arrow C3 and then along the route indicated by arrow C4. Cameras and detection modes that are selected according to the running state of the vehicle 9 will be described using the case where the vehicle 9 is parked in the parking section La as an example.

{4.1. When the vehicle 9 is traveling straight at normal speed}
Assume that the vehicle 9 is traveling straight from position P11 to position P12 shown in FIG. 19 while decelerating. Specifically, the speed of the vehicle 9 is 15 km/h at position P11 and 8 km/h at position P12. The rotation angle of the steering wheel of the vehicle 9 is 0° from position P11 to position P12. In this case, the parking space recognition device 1 starts the processing shown in FIG. 9 at the position P11.

(camera selection)
Since the vehicle 9 is traveling straight at a speed of 8 km/h or more, the running state of the vehicle 9 corresponds to mode ID "1" of the setting table 19 shown in FIG. Therefore, the detection setting unit 12 selects the left side camera 3L and the right side camera 3R (step S13 shown in FIG. 9). The left side camera 3L is assigned to the first detection section 21 . The right side camera 3R is assigned to the second detection section 22 .

(Determination of detection mode)
Since the running state of the vehicle 9 corresponds to the mode ID "1", the detection setting unit 12 sets the detection mode of each of the first detection unit 21 and the second detection unit 22 to the simple mode (step shown in FIG. 9). S14).

The detection setting unit 12 notifies the image acquisition unit 11 of the camera selected in step S13. The image acquisition unit 11 outputs the frame generated by the camera selected in step S13 to the first detection unit 21 and the second detection unit 22 based on the notification from the detection setting unit 12 . Frames are output at regular intervals. The constant interval is determined based on the processing speeds of the first detection section 21 and the second detection section 22, and is, for example, 0.3 seconds.

Since the detection mode of the first detection unit 21 and the second detection unit 22 is the simple mode, the detection region specifying unit 14 detects the left frame 31L and the right frame 31R based on the region table 32 stored in the memory 17. is specified (step S15).

The first detection unit 21 and the second detection unit 22 detect the parking section L from the detection area of the left frame 31L and the detection area of the right frame 31R (step S16). The section list 18 is updated based on the parking section L detected based on step S16 (step S17).

The parking space recognition device 1 repeats the process of detecting the parking space L using the left and right cameras when the vehicle 9 is traveling straight ahead at a speed of 8 km/h or more.

When the vehicle 9 is traveling straight at a speed of 8 km/h or more, it is unlikely that the vehicle 9 will immediately stop for reverse parking. In this case, in order to detect as many parking spaces L as possible, the parking space recognition device 1 uses a left side camera 3L capable of capturing a wide leftward area of the vehicle 9 and a wide rightward area of the vehicle 9. Select the right side camera 3R that can shoot.

Also, the simple mode is set as the detection mode of the first detection section 21 and the second detection section 22 . Since the simple mode does not use the past detection result of the parking space L, a new parking space L can be detected. Therefore, many parking spaces L can be detected.

{4.2. When the vehicle 9 is traveling straight at an intermediate speed}
Assume that the vehicle 9 is traveling straight at 4 to 8 km/h from position P12 to position P13 shown in FIG. In this case, the rotation angle of the steering wheel is 0°.

The detection setting unit 12 determines that the running state in the section from the position P12 to the position P13 corresponds to the mode ID "3" of the setting table 19 shown in FIG. Therefore, the detection setting unit 12 assigns the left side camera 3L and the right side camera 3R to the first detection unit 21, and assigns the rear camera 3B to the second detection unit 22 (step S13). The first detection unit 21 alternately repeats the process of detecting the parking section L from the left frame 31L and the process of detecting the parking section L from the right frame 31R. By assigning two or more cameras to the first detection unit 21 in this manner, a larger number of cameras than the number of detection units included in the parking space detection unit 15 can be selected.

The detection setting unit 12 sets the detection modes of the first detection unit 21 and the second detection unit to the simple mode based on the running state in the section from the position P12 to the position P13 (step S14).

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera assigned to each of the first detection unit 21 and the second detection unit 22 (step S15). Specifically, the detection area specifying unit 14 determines the detection areas of the left frame 31L and the right frame 31R when the simple mode is set, and the determined two detection areas are specified by the first detection unit 21. to notify. The detection area specifying unit 14 determines the detection area of the rear frame 31B when the simple mode is set, and notifies the second detection unit 22 of the determined detection area.

The first detection unit 21 detects the parking section L from the detection area of the left frame 31L and the detection area of the right frame 31R (step S16). That is, the first detection unit 21 repeats the process of detecting the parking section L from the left frame 31L and the process of detecting the parking section L from the right frame 31R. The second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). The section list 18 is updated based on the parking section L detected based on step S16 (step S17).

If the vehicle 9 is traveling straight at 4-8 km/h, it is considered that the vehicle 9 is decelerating near an empty parking space L. Therefore, the parking space recognition device 1 detects the parking space L using not only the left side camera 3L and the right side camera 3R but also the rear camera 3B, assuming that the vehicle will be parked backwards.

{4.3. When the vehicle 9 makes a right turn at an intermediate speed}
Assume that the vehicle 9 is turning right at a speed of 4 to 8 km/h from position P13 to position P14 shown in FIG. In this case, the rotation angle of the steering wheel is 250°.

The detection setting unit 12 determines that the running state in the section from the position P13 to the position P14 corresponds to the mode ID "4" of the setting table 19 shown in FIG. Therefore, the detection setting unit 12 assigns the left side camera 3L to the first detection unit 21 and assigns the rear camera 3B to the second detection unit 22 (step S13). The detection setting unit 12 sets the detection modes of the first detection unit 21 and the second detection unit to the simple mode (step S14).

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera assigned to each of the first detection unit 21 and the second detection unit 22 (step S15). Specifically, based on the area table 32, the detection area specifying unit 14 determines the detection area of each of the left frame 31L and the rear frame 31B when the simple mode is set. The detection area specifying unit 14 notifies the first detection unit 21 of the detection area of the left frame 31L, and notifies the second detection unit 22 of the detection area of the rear frame 31B.

The first detection unit 21 detects the parking section L from the detection area of the left frame 31L (step S16). The second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). The section list 18 is updated based on the parking section L detected based on step S16 (step S17).

When the vehicle 9 is turning right at a speed of 4 to 8 km/h, it is considered that the driver of the vehicle 9 intends to park the vehicle 9 in the parking space L located to the left rear of the vehicle 9. be done. The parking space recognition device 1 selects the left side camera 3L that captures the area outside the turning locus of the vehicle 9 and the rear camera 3B that captures the scenery behind the vehicle 9 according to the above idea. The first detection unit 21 does not detect the parking section L from the right frame 31R, unlike when the vehicle 9 is traveling in the section P2-P3. Therefore, it is possible to continuously detect a parking section in which the vehicle 9 is highly likely to park.

{4.4. When the vehicle 9 turns right at the speed just before stopping}
Assume that the vehicle 9 makes a right turn at a speed of 4 km/h or less over the section from position P14 to position P15 shown in FIG. 19 and stops at position P15. The rotation angle of the steering wheel is 250°. The operation of the parking space recognition device 1 when the vehicle 9 travels in the section from the position P14 to the position P15 will be described.

The detection setting unit 12 determines that the running state in the section from the position P14 to the position P15 corresponds to the mode ID "7" of the setting table 19 shown in FIG. Therefore, the detection setting unit 12 assigns the left side camera 3L to the first detection unit 21 and assigns the rear camera 3B to the second detection unit 22 (step S13). The detection setting unit 12 sets the simple mode to the first detection unit 21, and sets the simple mode and the detailed mode to the second detection unit 22 (step S14).

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera assigned to each of the first detection unit 21 and the second detection unit 22 (step S15). Specifically, since the detection mode of the first detection unit 21 is the simple mode, the detection area specifying unit 14 determines the detection area of the left frame 31L based on the area table 32. FIG. The detection modes of the second detection unit 22 are the simple mode and the detailed mode. The detection area specifying unit 14 first determines the detection area of the rear frame 31B based on the area table 32. FIG.

The first detection unit 21 detects the parking section L from the detection area of the left frame 31L (step S16). The second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). The section list 18 is updated based on the parking section L detected based on step S15 (step S17).

The detection area specifying unit 14 determines the detection area of the rear frame 31B based on the updated partition list 18. FIG. That is, the detection area specifying unit 14 determines the detection area of the rear frame 31B in the detailed mode. In the detailed mode, the second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). After that, the second detection unit 22 alternately repeats the simple mode and the detailed mode. On the other hand, the first detection unit 21 repeatedly detects the parking section L in the simple mode.

If the vehicle 9 is turning to the right at a speed of 4 km/h or less, it is expected that the driver of the vehicle 9 will soon stop the vehicle 9 in order to turn the vehicle 9 back. The parking section La is located on the left rear side of the vehicle 9, and reverse parking will soon start. Therefore, the parking space recognizing device 1 executes processing for detecting the parking space L in the detailed mode from the rear frame 31B in addition to processing for detecting the parking space L in the simple mode from the left frame 31L and the rear frame 31B. . Thereby, the parking space recognition device 1 can detect the parking space La with high accuracy while detecting the parking space L in a wide range.

{4.5. When the vehicle 9 is reversing outside the parking section L}
When the parking instruction switch 6 is pressed after the vehicle 9 stops at the position P15, the vehicle control device 2 instructs the driver to select the parking section L. When the driver selects the parking section La, the vehicle control device 2 starts parking assistance. Specifically, the vehicle control device 2 steers the vehicle 9 by rotating the steering wheel. A driver operates an accelerator pedal or a brake pedal of the vehicle 9 .

Assume that the vehicle 9 is reversing outside the parking bay L. Specifically, the vehicle 9 is backing up in the section from position P15 to position P16 shown in FIG. A section from position P15 to position P16 is outside the parking section La. The parking space recognition device 1 determines whether or not the origin Ow is in the parking space La based on the parking reference position Sa of the parking space La in the world coordinate system, the position of the first starting point, and the position of the second starting point. to judge whether For example, when the origin Ow does not cross the line connecting the first start point and the second start point of the parking space La, the parking space recognition device 1 determines that the vehicle 9 is outside the parking space La.

The detection setting unit 12 determines that the running state in the section from the position P15 to the position P16 corresponds to the mode ID "8" of the setting table 19 shown in FIG. Therefore, the detection setting unit 12 selects only the rear camera 3B. That is, the first detection unit 21 and the second detection unit 22 detect the parking space L from the rear frame 31B. The detection setting unit 12 sets the detection mode of the first detection unit 21 to the simple mode, and sets the detection mode of the second detection unit 22 to the detailed mode (step S13).

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera assigned to each of the first detection unit 21 and the second detection unit 22 (step S15). Specifically, since the detection mode of the first detection unit 21 is the simple mode, the detection area specifying unit 14 determines the detection area of the rear frame 31B based on the area table 32. FIG. Since the detection mode of the second detection unit 22 is the detailed mode, the detection area specifying unit 14 determines the detection area of the rear frame 31B based on the partition list 18. FIG.

Since reverse parking has started, the parking section La in which the vehicle 9 is parked is positioned behind the vehicle 9 . Therefore, the parking space recognition device 1 can detect the parking space La located behind the vehicle 9 with high accuracy by executing the process of detecting the parking space L from the rear frame 31B in both the simple mode and the detailed mode. At the same time, detection omission of the parking space La located behind the vehicle 9 can be prevented.

{4.6. When the vehicle 9 is backing up in the parking section L}
Assume that the vehicle 9 is reversing in the parking bay L. Specifically, referring to FIG. 19, vehicle 9 is backing up in a section from position P16 to parking reference position Sa in parking section La. A section from the position P16 to the parking reference position Sa is within the parking section La.

The detection setting unit 12 determines that the running state in the section from the position P16 to the parking reference position Sa corresponds to the mode ID "9" of the setting table 19 shown in FIG. Therefore, the detection setting unit 12 selects the left side camera 3L and the right side camera 3R (step S13). The detection setting unit 12 determines the detection mode of both the first detection unit 21 and the second detection unit 22 to be the detailed mode (step S14).

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera assigned to each of the first detection unit 21 and the second detection unit 22 (step S15). Specifically, since the detection modes of the first detection unit 21 and the second detection unit 22 are the detailed mode, the detection areas of the left frame 31L and the right frame 31R are determined based on the partition list 18. FIG. At this time, it is desirable that the detection area determination unit 14 sets the length of the long side of the detection area to be larger than the general length of the parking space in the vehicle length direction. The entire white line in the left frame 31L can be included in the detection area. The same applies to the setting of the detection area of the right frame 31R.

The first detection unit 21 detects the parking section L from the detection area of the left frame 31L (step S16). The second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). The section list 18 is updated based on the parking section L detected based on step S16 (step S17).

When the vehicle 9 is backing up in the parking space La, the left side camera 3L and the right side camera 3R can photograph the entire white line that defines the parking space La. Even when the detailed mode is used, the parking space recognition device 1 can detect the entire white line that defines the parking space La and detect the parking reference position Sa of the parking space La.

When the vehicle 9 is backing up in the parking section La, one of the two boundary lines in the long side direction of the parking section La is captured by the left side camera 3L and the other is captured by the right side camera 3R. be done. In this case, the parking section determination unit 15 may determine the parking reference position Sa of the parking section La based on the edge line detected from the left frame 31L and the edge line detected from the right frame 31R. .

{5. Second example of running state of vehicle 9}
FIG. 20 is a diagram showing a second example of a parking route for the vehicle 9 shown in FIG. Referring to FIG. 20, vehicle 9 is parked in parking section Lb by traveling along the route indicated by arrow C5 and then along the route indicated by arrow C6. Cameras and detection modes that are selected according to the running state of the vehicle 9 will be described using the case where the vehicle 9 is parked in the parking section Lb as an example.

{5.1. When the vehicle 9 goes straight at normal speed}
Assume that the vehicle 9 is traveling straight from position P21 to position P22 shown in FIG. 20 while decelerating at a speed of 8 to 15 kmh. The rotation angle of the steering wheel of the vehicle 9 is 0°. The running state of the vehicle 9 corresponds to mode ID "1" shown in FIG. The parking space recognition device 1 operates in the same manner as the vehicle 9 moves from position P11 to position P12 shown in FIG.

{5.2. When the vehicle 9 runs straight at an intermediate speed}
Assume that the vehicle 9 is traveling straight from position P22 to position P23 shown in FIG. 20 at a speed of 4 to 8 km/h. The rotation angle of the steering wheel is 0°. The running state of the vehicle 9 corresponds to mode ID "3" shown in FIG. The parking space recognition device 1 operates in the same manner as when the vehicle 9 moves from position P12 to position P13 shown in FIG.

{5.3. When the vehicle 9 runs straight at the speed just before stopping}
Assume that the vehicle 9 is traveling straight from position P22 to position P23 shown in FIG. 20 at a speed of 4 km/h or less. The rotation angle of the steering wheel is 0°. The running state of the vehicle 9 corresponds to mode ID "6" shown in FIG. The detection setting unit 12 assigns the left side camera 3L and the right side camera 3R to the first detection unit 21, and assigns the rear camera 3B to the second detection unit 22 (step S13). The detection setting unit 12 sets the simple mode to the first detection unit 21, and sets the simple mode and the detailed mode to the second detection unit 22 (step S14).

The detection area specifying unit 14 specifies the detection area of the frame generated by the camera assigned to each of the first detection unit 21 and the second detection unit 22 (step S15). Specifically, the detection area specifying unit 14 determines the detection areas of the left frame 31L and the right frame 31R in the simple mode. The detection area specifying unit 14 determines the detection area of the rear frame 31B in the simple mode and the detection area of the rear frame 31B in the detailed mode.

When executing step S16, the first detection unit 21 alternately repeats the process of detecting the parking section L from the left frame 31L and the process of detecting the parking section L from the right frame 31R. When executing step S16, the second detection unit 22 alternately repeats the process of detecting the parking space L in the simple mode and the process of detecting the parking space L in the detailed mode. The list management section 17 updates the section list 18 based on the parking section L detected by the first detection section 21 and the second detection section 22 (step S17).

When the vehicle 9 moves straight ahead at the speed just before stopping, the parking space recognition device 1 allocates the left side camera 3L and the right side camera 3R to the first detection section 21 . This is because it is not possible to specify whether the vehicle 9 is parked in the rear right or rear left direction even though the vehicle 9 is about to start backward parking. In addition, the parking space recognition device 1 assigns the rear camera 3B to the second detection unit 22, and then sets the detection mode of the second detection unit 22 to the simple mode and the detailed mode. As a result, the parking space recognition device 1 can detect parking spaces over a wide range and improve the detection accuracy of the parking space located behind the vehicle 9 .

{5.4. When the vehicle 9 turns left at an intermediate speed}
Assume that the vehicle 9 turns right at a speed of 4 to 8 km/h from position P23 to position P24 shown in FIG. The rotation angle of the steering wheel is 250°. The running state of the vehicle 9 corresponds to mode ID "2" in the setting table 19 shown in FIG.

The detection setting unit 12 assigns the right side camera 3R to the first detection unit 21 and assigns the rear camera 3B to the second detection unit 22 (step S13). The detection setting unit 12 sets the detection modes of the first detection unit 21 and the second detection unit to the simple mode (step S14). Based on the area table 32, the detection area specifying unit 14 determines the detection areas of the left frame 31L and the rear frame 31B when the simple mode is set (step S15).

The first detection unit 21 detects the parking section L from the detection area of the left frame 31L (step S16). The second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). The section list 18 is updated based on the parking section L detected based on step S16 (step S17).

When the vehicle 9 turns left at a speed of 4 to 8 km/h, it is considered that the driver of the vehicle 9 intends to park the vehicle 9 in the parking space L located to the right rear of the vehicle 9 . The parking space recognition device 1 selects the right side camera 3L for photographing the area outside the turning locus of the vehicle 9 and the rear camera 3B for photographing the scenery behind the vehicle 9 according to the above idea. As a result, it is possible to continuously detect a parking section in which the vehicle 9 is highly likely to park.

{5.5. When the vehicle 9 turns left at the speed just before stopping}
Assume that the vehicle 9 turns left at a speed of 4 km/h or less from position P24 to position P25 shown in FIG. 20 and stops at position P25. The rotation angle of the steering wheel is 250°. In this case, the running state of the vehicle 9 corresponds to mode ID "5" in the setting table 19 shown in FIG.

The detection setting unit 12 assigns the right side camera 3R to the first detection unit 21 and assigns the rear camera 3B to the second detection unit 22 (step S13). The detection setting unit 12 sets the simple mode to the first detection unit 21, and sets the simple mode and the detailed mode to the second detection unit 22 (step S14).

Since the detection mode of the detection unit 21 is the simple mode, the detection area specifying unit 14 determines the detection area of the left frame 31L based on the area table 32 (step S15). The detection area specifying unit 14 first determines the detection area of the rear frame 31B based on the area table 32 (step S15).

The first detection unit 21 detects the parking section L from the detection area of the left frame 31L (step S16). The second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16). The section list 18 is updated based on the parking section L detected based on step S15 (step S17).

The detection area specifying unit 14 determines the detection area of the rear frame 31B based on the updated section list 18 (step S15). This is because the second detection unit 22 alternately repeats the simple mode and the detailed mode. In the detailed mode, the second detection unit 22 detects the parking section L from the detection area of the rear frame 31B (step S16).

If the vehicle 9 makes a left turn at a speed of 4 km/h or less, the driver is expected to soon stop the vehicle 9 in order to turn the vehicle 9 back. This is because the vehicle 9 is parked in the parking section Lb located on the right rear. The parking space recognition device 1 performs processing for detecting the parking space L in the detailed mode from the rear frame 31B in addition to processing for detecting the parking space L in the simple mode from the left frame 31L and the rear frame 31B. Thereby, the parking space recognition device 1 can detect the parking space Lb with high accuracy while detecting the parking space L in a wide range.

{5.6. When the vehicle 9 is reversing outside the parking section L}
After the vehicle 9 stops at the position P25 shown in FIG. 20, the driver depresses the parking instruction switch 6. FIG. The vehicle control device 2 instructs the driver to select the parking section L in response to the depression of the parking instruction switch 6 . When the driver selects the parking section Lb, the vehicle control device 2 starts parking assistance. Specifically, the vehicle control device 2 steers the vehicle 9 by rotating the steering wheel. A driver operates an accelerator pedal or a brake pedal of the vehicle 9 . The parking space recognition device 1 operates in the same manner as when the vehicle 9 moves from the position P15 to the position P16 shown in FIG. 19 in the section from the position P25 to the position P26.

{5.7. When the vehicle 9 is backing up in the parking section L}
Assume that the vehicle 9 is reversing in the parking lot Lb shown in FIG. Specifically, the vehicle 9 is backing up in the section from the position P26 to the parking reference position Sb in the parking section Lb. A section from the position P26 to the parking reference position Sa is within the parking section Lb. The parking space recognition device 1 operates from the position P26 to the parking reference position Sb in the same manner as when the vehicle 9 moves from the position P26 to the parking reference position Sb shown in FIG.

{summary}
As described above, the parking space recognition device 1 selects a camera to be used for detecting the parking space L from among the plurality of cameras 3 based on the running state of the vehicle 9 . Thereby, the parking space recognition device 1 can reduce the detection load of the parking space L. FIG. Further, the parking space recognition device 1 changes the content of processing for detecting the parking space L based on the running state of the vehicle 9 . Thereby, the parking space recognition device 1 can reduce the detection load of the parking space L without lowering the detection accuracy of the parking space L.

{Modification}
(When parking spaces are marked with U-shaped white lines)
In the above-described embodiment, an example in which the parking section L is divided by a single white line has been described. However, the parking space recognition device 1 may detect a parking space L set by a white line drawn in a U shape.

FIG. 21 is a diagram for explaining detection of a parking space L set by a white line drawn in a U shape. In FIG. 21, the size of the white line is exaggerated and enlarged. Referring to FIG. 21, parking space L5 is defined by U-shaped white lines W5 and W6.

The white line W5 includes white lines W5a and W5b. The white line W5a is parallel to the white line W5b. The dark area W5c is an area between the white lines W5a and W5b, where the asphalt is exposed. The white line W6 includes white lines W6a and W6b. The white line W6a is parallel to the white line W6b. The dark area W6c is an area between the white lines W6a and W6b, where the asphalt is exposed.

The boundary line detection unit 211 detects the positive edge lines CP5a, CP5b, CP6a and CP6b and the negative edge lines CM5a, CM5b, CM6a and CM6b as described above. The plus edge line CP5a and the minus edge line CM5a form part of the outline of the white line W5a. The plus edge line CP5b and the minus edge line CM5b form part of the outline of the white line W5b. The plus edge line CP6a and the minus edge line CM6a form part of the outline of the white line W6a. The plus edge line CP6b and the minus edge line CM6b form part of the outline of the white line W6b.

The division determination unit 212 generates a pair by the method described above, and calculates the interval between two edge lines forming the generated pair. Specifically, section determination unit 212 identifies positive edge line CP5a and negative edge line CM5b as a candidate pair. The area between the plus edge line CP5a and the minus edge line CM5b is the dark area W5c.

However, the width of dark area W5c is much smaller than the width of parking space L5. Therefore, the distance between the plus edge line CP5a and the minus edge line CM5b is outside the predetermined range corresponding to the width of the parking space. Therefore, the section determination unit 212 does not detect the dark region R5c as the parking section L. FIG. Similarly, the dark area W6c of the white line W6 is not detected as the parking section L either.

The section determination unit 212 generates a pair including the positive edge line CP5b and the negative edge line CM6a, and determines that the interval between the two edge lines forming this pair is within a predetermined range corresponding to the width of the parking section. do. That is, the section determination unit 212 detects the parking section L5 based on the plus edge line CP5b and the minus edge line CM6a.

(When the parking space is diagonal to the runway)
22A and 22B are diagrams for explaining the detection of a parking space L set diagonally with respect to the road A. FIG. In FIG. 22, the size of the white line is exaggerated and enlarged. Referring to FIG. 22, parking section L7 is defined by white lines W7 and W8. The angle between the track A and the long side of the parking section L7 is, for example, 45°. That is, the parking section L7 is a parallelogram. Even if the parking space L7 is set diagonally with respect to the track A, the parking space detection unit 15 can detect the parking space L7 by the method described above. A specific description will be given below.

The section determination unit 212 detects the parking section L7 based on the plus edge line CP7 and the minus edge line CM8 using the same method as described above.

The section determination unit 212 determines a parking reference position S7 for the parking section L7. Specifically, the section determining unit 212 identifies a line segment E7 connecting the starting point of the positive edge line CP7 and the starting point of the negative edge line CM8. The division determination unit 212 determines a straight line passing through the midpoint M7 of the specified line segment E7 and parallel to either one of the plus edge line CP1 and the minus edge line CM2 as the center line K7. The section determination unit 212 determines a point on the center line K7 located in the parking section L7 and at a position of a length Lg from the midpoint M7 as the parking reference position S7.

FIG. 23 is a diagram showing another method of determining the parking reference position in the parking section L7. In FIG. 23, the size of the white line is exaggerated and enlarged. The section determining unit 212 generates a straight line VL that is orthogonal to the extension line CE extending from the minus edge line CM8 and that passes through the starting point of the plus edge line CP7. The virtual starting point VP is the intersection of the straight line VL and the extension line CE. The section determination unit 212 identifies the middle point M7' of the line segment E7' connecting the virtual start point VP and the start point of the plus edge line CP7.

The division determination unit 212 generates a center line K7 passing through the midpoint M7' and parallel to either the plus edge line or the minus edge line. The section determination unit 212 determines a point on the center line K7' in the parking section L7 and at a distance Lg from the midpoint M7' as the parking reference position S7 of the parking section L7.

Note that, when specifying the midpoint M7', the section determining unit 212 may generate a straight line VL that passes through the starting point of the negative edge line CM8 and that is orthogonal to the negative edge line CM8. In this case, the midpoint M7' is within the parking space L7.

(Other modifications)
In the above-described embodiment, an example in which the driver of the vehicle 9 instructs to start parking and selects the parking section L is described, but the present invention is not limited to this. The vehicle control device 2 uses the detection result of the parking space L by the parking space recognition device 1 even when selecting a parking space for parking the vehicle 9 and controlling the parking of the vehicle 9 in the selected parking space. You may

In the above-described embodiment, an example in which the detection setting unit 12 acquires the rotation angle of the steering wheel as the running state of the vehicle 9 has been described, but the present invention is not limited to this. The first detector 21 may determine whether the vehicle 9 is turning based on the rotation angle of the steered wheels of the vehicle 9 . That is, the detection setting unit 12 may select the camera and determine the detection mode based on the steering angle of the vehicle 9 .

In the above embodiment, the detection setting unit 12 selects the camera used for detecting the parking space L and determines the detection mode based on the speed of the vehicle 9, the rotation angle of the steering wheel, and the position of the vehicle 9. Although the example of performing the above has been described, the present invention is not limited to this. The detection setting unit 12 may select a camera to be used for detecting the parking space L and determine a detection mode based on the running state of the vehicle 9 . For example, acceleration of the vehicle 9 may be used as the running state of the vehicle 9 .

In the above-described embodiment, the detection setting unit 12 performs both camera selection and detection mode determination, but the present invention is not limited to this. The detection setting unit 12 may execute either one of camera selection and detection mode determination.

Although the detection setting unit 12 selects two cameras from the four cameras 3 in the above embodiment, the present invention is not limited to this. The detection setting unit 12 may select a smaller number of cameras than the number of cameras mounted on the vehicle 9 . For example, when the parking space detection unit 15 includes a third detection unit (not shown) in addition to the first detection unit 21 and the second detection unit, the parking space detection unit 15 detects three parking spaces from the image. Processing can be performed in parallel. In this case, the detection setting unit 12 may select three cameras.

In the above-described embodiment, an example was described in which the detection region specifying unit 14 specifies the detection region based on the detection mode determined by the detection setting unit 12, but the present invention is not limited to this. The parking space recognition device 1 does not have to include the detection area specifying unit 14 . In this case, the detection area is common to both the simple mode and the detailed mode.

In the above-described embodiment, an example in which the parking space recognition device 1 performs both camera selection and detection mode determination has been described, but the present invention is not limited to this. The parking space recognition device 1 may perform either one of camera selection and detection mode determination. When the parking space recognition device 1 determines the detection mode and does not select a camera, at least one camera may be mounted on the vehicle 9 .

Further, in the above-described embodiment, an example in which the parking section determination unit 15 does not specify the depth of the detected parking section L has been described, but the present invention is not limited to this. When the parking section determination unit 15 detects two boundary lines B that form the parking section L, the parking section determination unit 15 may determine four vertices that define the parking section L, assuming that the parking section L is a rectangle. . For example, the parking section determination unit 15 determines the starting points of the two boundary lines B forming the parking section L to be two vertices among the four vertices forming the quadrangle. The parking section determination unit 15 determines two points at a predetermined distance from each of the starting points of the two boundary lines B as the remaining vertices of the four vertices. In this case, the parking section determination unit 15 may record the four vertices defining the parking section L in the section list.

Further, in the above embodiment, the parking section determination unit 15 may determine whether or not another vehicle is parked in the parking section L detected. For example, the parking section determination unit 15 may detect whether or not another vehicle is parked in the detected parking section L based on the detection result of the distance measuring sensor. Alternatively, the parking section determining unit 15 may store vehicle features in advance and perform image recognition processing based on the stored vehicle features. In either case, whether or not another vehicle is parked in the detected parking space L may be determined based on the positional relationship between the detected other vehicle and the detected parking space L.

Moreover, in the above-described embodiment, the parking section determination unit 15 may determine whether or not another vehicle is parked. For example, the parking section determination unit 15 may detect whether or not another vehicle is parked in the detected parking section L based on the detection result of the distance measuring sensor. Alternatively, the parking section determining unit 15 may store vehicle features in advance and perform image recognition processing based on the stored vehicle features. In either case, whether or not another vehicle is parked in the detected parking space L may be determined based on the positional relationship between the detected other vehicle and the detected parking space L.

Further, in the above-described embodiment, a case has been described in which, when the parking section determination unit 15 detects the parking section L, the parking reference position and the two starting points of the two boundary lines B are recorded in the section list 18. , but not limited to this. The parking section determination unit 15 does not have to record the two starting points of the two boundary lines B in the section list 18 . In other words, the parking section determination unit 15 may detect the parking reference position S that serves as a reference for determining whether the parking of the vehicle 9 has been completed when parking the vehicle 9 from the image acquired by the camera 3. . That is, the parking space recognition device 1 operates as a parking position recognition device that detects the parking reference position S from the frame captured by the selected camera. The parking space detector 15 operates as a reference position detector. In this case, the description of the parking space in each functional block of the parking space recognition device 1 described in the above embodiment should be replaced with the parking reference position S.

In the above-described embodiment, an example in which the detection setting unit 12 determines the detailed mode and the simple mode has been described, but the present invention is not limited to this. The detection setting unit 12 may determine the first mode and the second mode, which have different parking space detection conditions, based on the running state of the vehicle 9 . When the first mode is selected, the parking space detection unit 15 detects the parking space from the frame generated by the selected camera under a preset first condition. A parking space may be detected from a frame generated by the selected camera under a second condition different from the first condition. Even in this case, the content of the parking space detection process can be changed based on the running state of the vehicle 9, so the load of parking space detection can be reduced.

Further, in the image processing apparatus described in the above embodiments, each functional block may be individually integrated into one chip by a semiconductor device such as an LSI, or may be integrated into one chip so as to include part or all of them. good. Although LSI is used here, it may also be called IC, system LSI, super LSI, or ultra LSI depending on the degree of integration.

Further, the method of circuit integration is not limited to LSI, and may be implemented by a dedicated circuit or a general-purpose processor. An FPGA or a configurable processor may be used after the LSI is manufactured.

Also, part or all of the processing of each functional block in each of the above embodiments may be implemented by a program. Part or all of the processing of each functional block in each of the above embodiments is performed by a central processing unit (CPU) in a computer. A program for performing each process is stored in a storage device such as a hard disk or ROM, and is read from the ROM or RAM and executed.

Further, each process of the above embodiment may be realized by hardware, or may be realized by software (including cases where it is realized together with an OS (operating system), middleware, or a predetermined library). . Furthermore, it may be realized by mixed processing of software and hardware.

For example, when each functional block of the above embodiment (including modifications) is realized by software, each function is You may make it implement|achieve a part by software processing.

Also, the execution order of the processing methods in the above embodiment is not necessarily limited to the description of the above embodiment, and the execution order can be changed without departing from the gist of the invention.

A computer program that causes a computer to execute the method described above and a computer-readable recording medium that records the program are included in the scope of the present invention. Examples of computer-readable recording media include flexible disks, hard disks, CD-ROMs, MOs, DVDs, DVD-ROMs, DVD-RAMs, large-capacity DVDs, next-generation DVDs, and semiconductor memories. .

The computer program is not limited to being recorded on the recording medium, and may be transmitted via an electric communication line, a wireless or wired communication line, a network represented by the Internet, or the like.

Although the embodiments of the present invention have been described above, the above-described embodiments are merely examples for carrying out the present invention. Therefore, the present invention is not limited to the above-described embodiment, and can be implemented by appropriately modifying the above-described embodiment without departing from the spirit of the present invention.

{Note}
The parking space recognition device may have the following first to tenth configurations.

A first configuration is a parking space recognition device that includes an image acquisition unit, a detection setting unit, and a parking space detection unit. The image acquisition unit acquires a plurality of images from a plurality of cameras mounted on the vehicle. The detection setting unit selects a camera to be used for detecting the parking space from among the plurality of cameras based on the running state of the vehicle. The parking space detection unit acquires the image generated by the selected camera from the image acquisition unit, and detects the parking space from the acquired image.

According to the first configuration, the parking space detection target image is changed according to the running state of the vehicle, so that the parking space detection load can be reduced.

A second configuration is the first configuration where the number of cameras selected is less than the plurality of cameras.

According to the second configuration, the parking space can be detected even when the parking space detection unit cannot simultaneously detect the parking space from the plurality of images generated by the plurality of cameras.

A third configuration is the second configuration, wherein the detection setting unit selects a camera to be used based on the speed of the vehicle.

According to the third configuration, it is possible to change the image from which the parking space is to be detected according to the speed of the vehicle.

A fourth configuration is the first or second configuration, wherein the detection setting unit selects the camera to be used based on the rotation angle of the steering wheel of the vehicle or the angle of the steered wheels of the vehicle.

According to the fourth configuration, it is possible to change the image to be detected of the parking space based on the angle of the steering wheel or steered wheels.

A fifth configuration is any one of the second to fourth configurations, wherein the detection setting unit selects the first to third cameras, assigns the first camera and the second camera to the first detection unit, A third camera is assigned to the second detector. The parking space detector includes a first detector and a second detector. The first detection unit alternately repeats a process of detecting a parking space from the image generated by the first camera and a process of detecting a parking space from the image generated by the second camera. The second detector detects the parking area from the image generated by the third camera.

According to the fifth configuration, the first detection unit alternately switches the detection target of the parking space. Accordingly, it is possible to select more cameras than the number of detection units.

A sixth configuration is the fifth configuration, wherein the plurality of cameras includes a left side camera and a right side camera. The left side camera captures the scenery on the left side of the vehicle. The right side camera captures the scenery to the right of the vehicle. The detection setting unit selects the left side camera and the right side camera when the speed of the vehicle is greater than a predetermined first reference value and the vehicle is moving forward.

This is because, according to the sixth configuration, the left side camera and the right side camera can detect more parking spaces L than the front camera and the rear camera when the vehicle is traveling on the lane of the parking lot. By the detection setting unit selecting the left side camera and the right side camera when the speed of the vehicle is greater than the first reference value, many parking spaces can be detected.

A seventh configuration is the fifth or sixth configuration, wherein the parking space detector includes a first detector and a second detector. The multiple cameras include a left side camera, a right side camera, and a rear camera. The left side camera captures the scenery on the left side of the vehicle. The right side camera captures the scenery to the right of the vehicle. The rear camera captures the scenery behind the vehicle. The detection setting unit assigns the left side camera and the right side camera to the first detection unit when the speed of the vehicle is within a predetermined speed range and the steering angle of the vehicle is smaller than the predetermined angle, and the rear camera is set to the first detection unit. Allocate to the second detector. When the speed of the vehicle is within a predetermined speed range and the steering angle is equal to or greater than a predetermined angle, the detection setting unit selects a camera positioned outside the turning trajectory of the vehicle, out of the left side camera and the right side camera. The first detector is assigned, and the rear camera is assigned to the second detector. The first detection unit and the second detection unit detect parking spaces from the frames generated by the cameras assigned by the detection setting unit.

According to the seventh configuration, it is possible to detect a parking section in which the vehicle 9 is highly likely to park.

An eighth configuration is any one of the fifth to seventh configurations, and the plurality of cameras includes a left side camera, a right side camera, and a rear camera. The left side camera captures the scenery on the left side of the vehicle. The right side camera captures the scenery to the right of the vehicle. The rear camera captures the scenery behind the vehicle. The detection setting unit selects the rear camera when the vehicle is backing up and the vehicle is located outside the parking space. The detection setting unit selects the left side camera and the right side camera when the vehicle is backing up and the vehicle is positioned within the parking space.

According to the eighth configuration, the rear camera is selected when the vehicle is backing up and the vehicle is located outside the parking space. Thereby, the parking section in which the vehicle is parked can be detected with high accuracy. The left side camera and the right side camera are selected when the vehicle is backing up and the vehicle is located in a parking space. As a result, it is possible to photograph the entire white line that defines the parking space in which the vehicle is to be parked, so that the parking space can be detected.

A ninth configuration is a parking position recognition device, which includes an image acquisition section, a detection setting section, and a reference position detection section. The image acquisition unit acquires a plurality of images from a plurality of cameras mounted on the vehicle. The detection setting unit selects, from among the plurality of cameras, a camera to be used for detecting the parking reference position based on the running state of the vehicle. The reference position detection section acquires the image generated by the selected camera from the image acquisition section, and detects the parking reference position from the acquired image.

According to the ninth configuration, since the image to be detected for the parking reference position is changed according to the running state of the vehicle, it is possible to reduce the detection load of the parking reference position.

The parking space recognition method used in the parking space recognition device 1 may include the following steps a), b), and c). The a) step acquires a plurality of images from a plurality of cameras mounted on the vehicle. The step b) selects one of the plurality of cameras to be used for parking space detection based on the running state of the vehicle. The step c) selects an image generated by the selected camera from among the acquired images and detects parking spaces from the selected image.

The parking space recognition device 1 may have the tenth to eighteenth configurations described below.

In a tenth configuration, the parking space recognition device recognizes parking spaces using images generated by at least one camera mounted on the vehicle. The parking space recognition device includes an image acquisition unit, a detection setting unit, and a parking space detection unit. The image acquisition unit acquires an image from at least one camera. The detection setting unit selects a detection mode for detecting a parking space from the acquired image from among a first mode and a second mode based on the running state of the vehicle. When the detection setting unit selects the first mode, the parking space detection unit detects the parking space from the acquired image under a predetermined first condition, and when the detection setting unit selects the second mode, acquires the parking space. A parking space is detected from the obtained image under a second condition different from the first condition.

According to the tenth configuration, the detection mode is changed based on the running state of the vehicle. Therefore, the load of the parking space detection process can be reduced without lowering the parking space detection accuracy.

An eleventh configuration is the tenth configuration, wherein the parking space detector includes a detection area designator. The detection area designating unit designates a preset detection area in the acquired image when the detection setting unit selects the first mode, and designates the already detected area when the detection setting unit selects the second selection mode. Specify detection regions based on parking spaces. The parking space detection unit detects a parking space from the detection area specified in the acquired image.

According to the eleventh configuration, when the second mode is selected, the detection area is detected based on the already detected parking space, so the parking space can be quickly detected.

A twelfth configuration is the tenth or eleventh configuration, wherein the parking section detection section includes a boundary line detection section and a section determination section. The boundary line detection unit selects a pixel of interest in the acquired image and determines whether or not an edge is formed by the pixel of interest, thereby obtaining at least a part of the outline of the lane line indicating the range of the parking space. Detect boundaries. The section determination unit determines a parking section based on the detected boundary line. The selection interval of the pixels of interest when the second mode is selected is narrower than the selection interval of the pixels of interest when the first mode is selected.

According to the twelfth configuration, when the second mode is selected, the parking space can be detected with higher accuracy than when the first mode is selected.

A thirteenth configuration is the twelfth configuration, wherein the detection area specified when the first mode is selected is wider than the detection area specified when the second mode is selected.

According to the thirteenth configuration, the detection area specified when the first mode is selected is wider than the detection area specified when the second mode is selected. Therefore, parking spaces can be detected from a wide area. Further, unlike the case where the second mode is selected, when the first mode is selected, it becomes possible to detect a new parking space.

A fourteenth configuration is the eleventh configuration, wherein the boundary line detection unit detects two boundary lines extending in the long side direction of the lane marking. When the two boundary lines are parallel and the interval between the two boundary lines is within a predetermined range, the section determination unit determines that the parking section is detected by the two boundary lines.

According to the fourteenth configuration, the parking space is detected based on the positional relationship between the detected two boundary lines. Since it is not necessary to use the length of the two boundary lines to detect the parking space, the parking space can be detected even if part of the white line is hidden by an obstacle. For example, parking spaces where other vehicles are already parked can be detected.

A fifteenth configuration is the tenth configuration, wherein when the first mode is selected, the parking space detector detects a boundary line, which is at least a part of an outline of a line that defines the parking space, with a predetermined accuracy. A parking reference position is determined based on the boundary line detected from the image. When the second mode is selected, the parking section detection unit specifies a detection area based on the parking reference position that has already been detected, and detects and detects the boundary line from the detection area with an accuracy higher than a predetermined accuracy. A parking reference position is determined based on the boundary line. The parking reference position is a reference position for stopping the vehicle within the parking space.

According to the fifteenth configuration, when the second mode is selected, the parking section detection unit designates the detection area based on the already detected parking reference position, and based on the boundary line detected from the detection area. Determine the parking reference position. Thereby, the parking reference position of the parking space detected in the first mode can be determined with higher accuracy.

A sixteenth configuration is the tenth configuration, wherein the detection setting unit selects the detection mode based on at least one of the speed of the vehicle, the steering angle of the vehicle, and the position of the vehicle.

According to the sixteenth configuration, it is possible to quickly change the detection mode according to changes in the running state of the vehicle.

A seventeenth configuration is a parking position recognition device that recognizes a parking reference position using an image generated by at least one camera mounted on a vehicle, comprising: an image acquisition unit; a detection setting unit; and The image acquisition unit acquires an image from at least one camera. The detection setting unit selects a detection mode for detecting the parking reference position from the acquired image from among a first mode and a second mode based on the running state of the vehicle. When the detection setting unit selects the first mode, the reference position detection unit detects the parking reference position under a predetermined first condition from the acquired image, and when the detection setting unit selects the second mode. , the parking reference position is detected from the acquired image under a second condition different from the first condition.

According to the seventeenth configuration, the detection mode is changed based on the running state of the vehicle. Therefore, the load of the detection processing of the parking reference position can be reduced without lowering the detection accuracy of the parking reference position.

A parking space recognition device according to a tenth configuration may execute a parking space recognition method comprising a) steps, b) steps, and c) steps. The step a) acquires images from at least one camera. The b) step selects a detection mode for detecting a parking space from the obtained image from among a first mode and a second mode, based on the running state of the vehicle. The c) step detects a parking space from an image acquired under a predetermined first condition when the first mode is selected, and detects a second condition different from the first condition when the second mode is selected. Detect parking spaces from the image acquired in .

100 parking support system 1 parking space recognition device 2 vehicle control device 11 image acquisition unit 12 detection setting unit 14 detection area designation unit 15 parking space detection unit 21 first detection unit 22 second detection units 211, 221 boundary line detection unit 221, 222 section determination unit 213 list update unit 214 position correction unit

Claims (7)

generated by a plurality of cameras including a left side camera capturing a view to the left of the vehicle, a right side camera capturing a view to the right of the vehicle, and a rear camera capturing a view behind the vehicle an image acquisition unit that acquires a plurality of images;
A detection mode for selecting a camera to be used for detecting a parking space from among the plurality of cameras based on the running state of the vehicle, and detecting the parking space from an image generated by the selected camera, a detection setting unit that selects from a first mode and a second mode based on the running state of the vehicle;
When the detection setting unit selects the first mode, a predesignated first detection area is designated in the image generated by the selected camera among the plurality of images acquired by the image acquisition unit. a detection area specifying unit that specifies a second detection area narrower than the first detection area in the image generated by the selected camera when the detection setting unit selects the second mode;
When the detection setting unit selects the first mode, the parking space is detected with a first accuracy from the first detection area specified by the detection area specifying unit, and the detection setting unit selects the second mode. a parking space detection unit that detects the parking space from the second detection area designated by the detection area designating unit with a second accuracy when the detection area designating unit detects the parking space. The parking space recognition device of claim 1, comprising:
The parking space detection unit is
including a first detection unit and a second detection unit,
The detection setting unit assigns one or more cameras from among the plurality of cameras to each of the first detection unit and the second detection unit;
The first detection unit and the second detection unit are parking space recognition devices that detect parking spaces in parallel in terms of time. A parking space recognition device according to claim 2, comprising:
The detection setting unit assigns the left side camera and the right side camera to the first detection unit, assigns the rear camera to the second detection unit,
The first detection unit alternately repeats a process of detecting a parking space from the image generated by the left side camera and a process of detecting a parking space from the image generated by the right side camera,
The parking space recognition device, wherein the second detection unit detects the parking space from the image generated by the rear camera. A parking space recognition device according to claim 2 or 3,
When the speed of the vehicle is within a predetermined speed range and the steering angle of the vehicle is smaller than a predetermined angle, the detection setting unit sets the left side camera and the right side camera to the first detection unit. and assigning the rear camera to the second detection unit,
When the speed of the vehicle is within the predetermined speed range and the steering angle is equal to or greater than the predetermined angle, the detection setting unit detects the turning trajectory of the vehicle from the left side camera and the right side camera. assigning a camera positioned outside the first detection unit to the first detection unit, assigning the rear camera to the second detection unit;
The parking space recognition device, wherein the first detection unit and the second detection unit detect the parking space from frames generated by the cameras assigned by the detection setting unit. The parking space recognition device according to any one of claims 1 to 4,
The parking section detection unit
A range of a parking space is determined by selecting target pixels in the specified first detection area or the specified second detection area and determining whether or not an edge is formed by the selected target pixels. a boundary line detection unit that detects a boundary line that is at least part of the contour of the marking line;
a section determination unit that determines a parking section based on the detected boundary line;
The parking space recognition device, wherein the selection interval of the pixels of interest when the second mode is selected is narrower than the selection interval of the pixels of interest when the first mode is selected. The parking space recognition device according to any one of claims 1 to 5,
When the first mode is selected, the parking space detection unit detects a boundary line, which is at least a part of an outline of a space line indicating the range of the parking space, from the image with the first accuracy. determine the parking reference position based on the boundary,
When the second mode is selected, the detection area designating unit designates the second detection area based on the end points of already detected boundary lines,
When the second mode is selected, the parking space detection unit detects a boundary line with the second accuracy from the second detection area designated by the detection area designation unit, and detects the boundary line based on the detected boundary line. to determine the parking reference position,
The parking space recognition device, wherein the parking reference position is a reference position for stopping the vehicle in the parking space. generated by a plurality of cameras including a left side camera capturing a view to the left of the vehicle, a right side camera capturing a view to the right of the vehicle, and a rear camera capturing a view behind the vehicle obtaining a plurality of images;
A camera used for detecting a parking space is selected from among the plurality of acquired images based on the running state of the vehicle, and detection for detecting a parking space from the image generated by the selected camera. selecting a mode from a first mode and a second mode based on the running state of the vehicle;
When the first mode is selected, a predesignated first detection area is designated in the image generated by the selected camera among the plurality of acquired images, and the second mode is selected. if so, designating in the image generated by the selected camera a second detection area that is narrower than the first detection area;
When the first mode is selected, a parking space is detected from the designated first detection area with a first precision, and when the second mode is selected, the parking space is parked from the designated second detection area. detecting a parking space with a second accuracy.

Images