Rear white line estimation device, target recognition device and method
The device and method improve rear white line estimation accuracy by integrating GNSS, yaw rate, and front white line data to reduce errors, enhancing lane identification and safety predictions.
Patent Information
- Application Number
- JP2023525374
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2021-06-04
- Filing Date
- 2022-02-04
- Publication Date
- 2024-11-27
- Estimated Expiration
- 2042-02-04
AI Technical Summary
Existing technologies struggle to accurately estimate the rear white line of a vehicle, which is crucial for predicting lane changes and ensuring safety, as methods like GNSS fail in tunnels and built-up areas, and yaw rate integration leads to error accumulation.
A device and method that utilize a white line detection unit, relative heading estimation, and a target object recognition system to accurately estimate the rear white line by combining GNSS, yaw rate, and front white line data, reducing errors through relative bearing calculations.
Enhances the accuracy of rear white line estimation, improving lane identification and reducing errors associated with GNSS and yaw rate integration, especially in environments where these methods are unreliable.
Smart Images

Figure 0007591141000001 
Figure 0007591141000002 
Figure 0007591141000003
Abstract
Description
[Technical field]
[0001] The present invention relates to a rear white line estimation device, a target recognition device and a method for grasping and recognizing the position of a target behind a vehicle in relation to a rear white line. [Background technology]
[0002] In vehicle operation, it is necessary to ensure safety behind the vehicle, and it is necessary to detect, for example, a vehicle as a target using a radar device and confirm its position. However, for example, when the vehicle is traveling on a road with multiple lanes, it may be difficult for the radar device to determine which lane the following vehicle is traveling in. For example, on a curved road that turns left, the following vehicle traveling in a lane to the right of the lane the vehicle is traveling in may be located directly behind the vehicle, making it impossible to determine the lane the following vehicle is traveling in using only the radar device.
[0003] In this regard, Patent Document 1 is known as a driving support device that uses a radar device to recognize a following vehicle and determine the lane in which the following vehicle is traveling in order to enable lane determination. The driving support device described in Patent Document 1 "calculates the shape of the lane in a section from the current position of the host vehicle to a predetermined distance behind, based on a group of lane information consisting of a plurality of lane information pieces indicating the relative positions between the host vehicle and the lane detected from a plurality of images continuously captured by an imaging device during a predetermined data acquisition period from the present to the past, and a travel trajectory of the host vehicle during the data acquisition period calculated based on the detection results of a vehicle speed sensor and a yaw rate sensor during the data acquisition period." [Prior art documents] [Patent documents]
[0004] [Patent Document 1] JP 2019-046150 A Summary of the Invention [Problem to be solved by the invention]
[0005] In Patent Document 1, detection results of a vehicle speed sensor and a yaw rate sensor are used to calculate the lane shape in a section from the current vehicle position to a certain distance behind. It is also known that the same thing can be done by using GNSS.
[0006] Incidentally, it is important to take into account rear white lines when estimating the lane of the vehicle behind, which is necessary for risk prediction in lane change assistance, etc. Vehicles equipped with ADAS (Advanced Drive Assistance System) generally have sensors that detect white lines ahead, but do not have sensors that detect white lines behind, so it is necessary to estimate the rear white lines using a map or previously detected white lines ahead.
[0007] However, there is a problem that the vehicle's heading, which is necessary for estimating the rear white line, cannot be obtained with high accuracy. For example, GNSS, which relies on maps, cannot be used in tunnels or built-up areas, and errors in yaw rate calculations accumulate, and the errors become larger when changing lanes.
[0008] SUMMARY OF THE PRESENT DISCLOSURE In view of the above, an object of the present invention is to provide a rear white line estimation device, a target object recognition device, and a method thereof that enable highly accurate estimation of a rear white line. [Means for solving the problem]
[0009] In view of the above, the present invention is a "rear white line estimation device characterized by having a white line detection unit that detects a white line ahead of the vehicle, a white line orientation generation unit that determines the orientation of the vehicle relative to the white line at each specified time based on the detection result of the white line detection unit, and a relative orientation estimation unit that estimates a relative orientation from the orientation of the vehicle at a certain point based on the orientation of the vehicle determined by the white line orientation generation unit."
[0010] Further, in the present invention, "a white line detection unit that detects a white line ahead of the vehicle, a white line direction generation unit that determines the direction of the vehicle relative to the white line at each predetermined time based on the detection result of the white line detection unit, a relative direction estimation unit that estimates a relative direction from the direction of the vehicle at a certain point based on the direction of the vehicle determined by the white line direction generation unit, a target detection unit that detects a target behind the vehicle, a movement distance estimation unit that estimates a movement distance of the vehicle, and a relative direction at each predetermined time estimated by the relative direction estimation unit and a relative direction at each predetermined time estimated by the movement distance estimation unit. a vehicle trajectory estimation unit that estimates a movement trajectory of the vehicle relative to the white line based on a movement distance per time of the vehicle, a white line position estimation unit that estimates a position of a white line that is outside a current detection range of the white line detection unit based on detection results of past detections by the white line detection unit and the movement trajectory, and a target position identification unit that identifies a positional relationship between a white line outside the detection range and a target based on the position of the white line outside the detection range estimated by the white line position estimation unit and the position of the target detected by the target detection unit.
[0011] Furthermore, the present invention is defined as "a rear white line estimation method using a computer, which detects a white line in front of the vehicle, determines the orientation of the vehicle relative to the white line at each specified time based on the detection result of the white line, and estimates a relative orientation from the orientation of the vehicle at a certain point based on the orientation of the vehicle."
[0012] The present invention is further described as "a rear white line estimation method using a computer, which detects a white line ahead of the vehicle, determines the orientation of the vehicle relative to the white line at each prescribed time based on the detection results of the white line, estimates a relative orientation from the orientation of the vehicle at a certain point based on the orientation of the vehicle, detects targets behind the vehicle, estimates the distance traveled by the vehicle, estimates a movement trajectory of the vehicle relative to the white line based on the estimated relative orientation at each prescribed time and the estimated movement distance at each prescribed time, estimates the position of a white line that is outside the detection range of a current white line detection based on the detection results and movement trajectory of previously detected white lines, and identifies the positional relationship between a white line outside the detection range and the target based on the estimated position of the white line outside the detection range and the detected position of the target." Effect of the Invention
[0013] According to the present invention, it is possible to improve the accuracy of identifying a lane in which a rear vehicle is present. [Brief description of the drawings]
[0014] [Figure 1] 1 is a diagram showing an example of the overall configuration of a target recognition device according to the present invention; [Diagram 2] 4 is a flowchart showing the processing contents in a trajectory point sequence derivation unit 30. FIG. [Diagram 3] FIG. 4 is a diagram showing the characteristics of each direction detection method. [Figure 4] 4 is a flowchart showing the processing contents in an object position identification unit 40. FIG. [Diagram 5] FIG. 11 is a flowchart showing a process of estimating a rear white line according to the second embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0015] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the first embodiment, an example of the overall configuration of a target recognition device will be described, and in the second embodiment, a specific example of rear white line estimation, which is a main element in the target recognition device, will be described. EXAMPLES
[0016] FIG. 1 is a diagram showing an example of the overall configuration of a target recognition device according to the present invention. The target recognition device 10 in FIG. 1 obtains outputs from a plurality of sensors S, and finally outputs the object position and the lane to which it belongs. These sensors S and their outputs are latitude, longitude, and positioning state from the GNSS (S1), vehicle speed from the speed sensor S2, yaw rate from the yaw rate sensor S3, a dot sequence group of forward white lines from the front camera S4, and object (target) position from the rear radar. These sensor inputs are time-series information sampled and input at a constant period, and each data includes time information at the time of sampling. The amount of data may be reduced by expressing the time information by treating the transmission timing by the sensor as time. The amount of data may also be reduced by inputting the sensor only when the sensor detects an object to be detected.
[0017] The target recognition device 10 is a part of a function of a driving support device mounted on a vehicle capable of automatic driving, and is a function of outputting lane change availability information of the driving support device. The target recognition device 10 also includes a control unit configured by a computer having a CPU, ROM, RAM, input / output devices, etc. connected to a bus, and the control unit controls the target recognition device 10 based on a predetermined program.
[0018] The processing functions of the target object recognition device 10 in FIG. 1 can be broadly divided into a distance / direction conversion unit 20 that converts an input signal into a distance and direction signal, a trajectory point sequence derivation unit 30 that determines a vehicle movement trajectory as a trajectory point sequence from the output of the distance / direction conversion unit 20, and an object position identification unit 40 that determines the object position behind the vehicle and the lane to which it belongs.
[0019] A trajectory point sequence derivation unit 30 that obtains the vehicle movement trajectory as a trajectory point sequence obtains distance information and direction information from the output of the distance / direction conversion unit 20. Of these, distance information can be obtained from latitude, longitude, and positioning state from the GNSS (S1) and vehicle speed information from the speed sensor S2, and direction information can be obtained from latitude, longitude, and positioning state from the GNSS (S1), the yaw rate from the yaw rate sensor S3, and a point sequence group of the forward white line from the front camera S4.
[0020] For this purpose, in the distance / direction conversion unit 20, the latitude, longitude and positioning state from the GNSS (S1) are stored in the storage unit 21a, and GNSS distance information 21b, GNSS direction accuracy estimation information 21c and GNSS direction information 21d included in the GNSS information are extracted. Of these, the GNSS distance information 21b is provided to a travel distance estimation unit 31 of the trajectory point sequence derivation unit 30, and the GNSS direction accuracy estimation information 21c and GNSS direction information 21d are provided to a direction selection unit 32 of the trajectory point sequence derivation unit 30.
[0021] Moreover, in the distance / direction conversion unit 20 , accumulated distance information 22 b obtained by integrating the vehicle speed from the speed sensor S 2 is provided to a travel distance estimation unit 31 of the locus point sequence derivation unit 30 .
[0022] In the distance / orientation conversion unit 20, the yaw rate from the yaw rate sensor S3 is integrated to obtain accumulated orientation information 23d, and the orientation accuracy is estimated to obtain orientation accuracy estimation information 23c. Note that the orientation accuracy in this case is high when the distance is short and the displacement is small, but decreases when the distance is long and the displacement is large, so the accuracy can be calculated by quantifying it. The accumulated orientation information 23d and the orientation accuracy estimation information 23c are provided to the orientation selection unit 32 of the trajectory point sequence derivation unit 30.
[0023] Furthermore, in the present invention, the distance / direction conversion unit 20 newly obtains white line direction generation information 24d from the point sequence group of the white line ahead captured by the front camera S4, and also obtains direction accuracy estimation information 24c by estimating the direction accuracy at that time, and the white line history is stored in the white line history storage unit 24e. The white line direction generation information 24d and direction accuracy estimation information 23c are provided to the direction selection unit 32 of the trajectory point sequence derivation unit 30. The direction accuracy in this case is high when the white line change is small, but decreases when the white line change is large, so the accuracy can be obtained by quantifying it. The method of calculating the white line direction generation information 24d will be described in detail in the second embodiment.
[0024] In this way, in the present invention, a group of points of forward white lines is newly used in addition to the GNSS and yaw, which are conventionally known as requirements for rearward estimation. Based on this, the trajectory point sequence derivation unit 30 estimates the travel distance in the travel distance estimation unit 31 from the GNSS distance information 21b included in the GNSS information and the accumulated distance information 22b obtained by integrating the vehicle speed.
[0025] The idea behind estimating travel distance is to use the difference in GNSS positions if GNSS conditions are better, but when the positioning conditions deteriorate due to the terrain, obstacles such as tunnels and buildings, radio wave conditions, etc. and it is inappropriate to use GNSS, it is better to estimate and output the cumulative distance information 22b obtained by integrating the vehicle speed over time as the distance.
[0026] In addition, the trajectory point sequence derivation unit 30 preferably selects and outputs one of the directions of the GNSS, yaw, and forward white line point sequence groups selected by the direction selection unit 32. In this case, the selection is based on the information on the accuracy of each direction (21c, 23c, 24c) obtained in association with each of the GNSS, yaw, and forward white line point sequence groups.
[0027] A flow showing the processing contents in the trajectory point sequence derivation unit 30 is shown in Fig. 2. This flow is repeatedly executed for each predetermined control period as shown in processing step St0. When it reaches the time determined by the control period, in processing step St1 in Fig. 2, GNSS, yaw, and direction information based on the point sequence group of the white line ahead (21d, 23d, 24d) and information on the accuracy of the direction (21c, 23c, 24c) are input as inputs required for the processing here.
[0028] In processing step St2, the conditions are determined as follows: if the GNSS is good and the speed is fast, or if the GNSS is good and there is little change in the accumulated orientation 23d, and if either of these is met (yes), the GNSS orientation 21d is selected as the orientation information in processing step St3.
[0029] If the condition is not satisfied (no) in the judgment of processing step St2, the judgment of processing step St4 or processing step St5 is performed. For example, in processing step St4, if there are many changes in the cumulative orientation 23d and the orientation information 24d based on the point sequence group of the white line ahead is usable, the orientation 24d relative to the white line is selected as the orientation information in processing step St6.
[0030] Similarly, in process step St5, if there is little change in the cumulative heading 23d and the heading information 24d based on the dot sequence group of the forward white lines cannot be used, the heading 23d based on the yaw rate is selected as the heading information in process step St7.
[0031] If none of the judgments in the process steps St2, St4, and St5 are met (no), then in the process step St8, it is advisable to take a measure to maintain the selection result in the previous control cycle as a second-best measure. The various conditions for the above judgment may be detected directly, or may use the information on each azimuth accuracy (21c, 23c, 24c).
[0032] The result of the selection of the direction information is passed to the vehicle trajectory estimation unit 33, which is the next processing unit, together with the distance information from the travel distance estimation unit 31. After this selection, the process returns to step St0 and waits until the start of activation in the next control cycle.
[0033] The selection in Fig. 2 reflects the fact that the characteristics of each direction detection method are as shown in Fig. 3. According to the comparison table in Fig. 3, the vertical axis lists the direction estimation methods, such as GNSS, yaw, and direction detection using the dot sequence group of the white line ahead, and the horizontal axis lists the situations in which it is impossible or undesirable to use, and the conditions under which it is highly accurate. According to this, GNSS is not suitable for use in tunnels or built-up areas, and the conditions under which it can be detected with high accuracy are when driving at high speeds and when positioning is good. There are no particular situations in which it cannot be used with yaw rate, and it can be used at any time, but the conditions for high accuracy are when the distance and displacement are small. Conversely, this means that when the distance or displacement is large, the accuracy decreases and it becomes unsuitable for use. The direction using the dot sequence group of the white line ahead cannot be used in an environment where there are no white lines, but it can be detected with high accuracy when the change in the white line is small.
[0034] The vehicle's own trajectory estimation unit 33 in the trajectory point sequence derivation unit 30 performs processing using distance information and displacement information at a control cycle (hereinafter sometimes referred to as a frame) of, for example, 50 ms intervals, and stores three types of orientations detected at this time using the GNSS, yaw, and forward white line point sequence group. In addition, the GNSS is basically relied upon, but when the GNSS cannot be used, the current position and orientation are accumulated from the orientation and movement distance. Note that the orientation here means that the current position and orientation are accumulated using the difference from the orientation <previous reference orientation> stored in the same manner in the previous control cycle. The accumulated current position and orientation are obtained as information on the trajectory point sequence.
[0035] Based on the above processing, the object position identification unit 40 receives input of information on a sequence of trajectory points regarding the current position and direction from the vehicle trajectory estimation unit 33, information 24e of the white line history that holds the sequence of points of the white line ahead captured by the front camera S4, and information on the object position grasped by the rear radar S5.
[0036] Using these pieces of information, the object position identification unit 40 inputs, as shown in the processing flow of FIG. 4, the following data required for processing in processing step St11: the position of the target (radar target) currently grasped by the rear radar S5 in the vehicle center coordinates, the difference (relative direction) between the current vehicle direction and the direction saved in the same manner as last time, and the current position of the vehicle in the ground fixed coordinates (vehicle trajectory). Here, the vehicle center coordinates are a coordinate system based on the position and direction of the vehicle, and for example, correspond to an orthogonal coordinate system with the rear wheel axle center as the origin, the front of the vehicle as the positive x-axis, the left of the vehicle as the positive y-axis, and the upward direction as the positive z-axis. The reference of the vehicle center coordinates changes when the vehicle moves or changes direction. The ground fixed coordinates are a coordinate system based on a specific position outside the vehicle, such as on a road, and correspond to, for example, UTM (Universal Transverse Mercator) orthogonal coordinates converted from latitude and longitude into an orthogonal coordinate system. Ground fixed coordinates are basically a coordinate system that remains constant regardless of the movement of the vehicle. However, as is done in the use of UTM, multiple ground fixed coordinates may be switched and used according to the movement of the vehicle to reduce errors caused by the Earth's surface not being flat. In addition, the method of taking the position and direction that serves as the reference for the ground fixed coordinates is arbitrary, and for example, the ground fixed coordinates may be obtained by using the vehicle center coordinates at a specific time without changing the reference with the movement of the vehicle. Since the ground fixed coordinates in this example do not require the vehicle position on the Earth like UTM, they can be used even in situations where GNSS is never available. In addition, as in the following example, the z-axis may be omitted and the coordinates may be expressed in two dimensions assuming that the road surface is flat, thereby reducing the calculation load.
[0037] Next, using these data, in processing step St12, the position of the radar target is projected onto ground-fixed coordinates, and in processing step St13, a frame in the vehicle trajectory in which a white line is detected at a position closest to the ground-fixed coordinates of the radar target is extracted. In processing step St14, the radar target is compared with the white line at the vehicle center coordinates in that frame, and the lane is identified.
[0038] According to the present invention described above, even when the use of GNSS is not suitable, lane identification can be performed by reducing the azimuth error using the relative azimuth from a specific point of the lane line direction instead of the cumulative azimuth (yaw rate integral) which has a large error. In addition, the error caused by the azimuth error when converting the radar target from the vehicle center coordinate to the ground fixed coordinate can be suppressed, and the accuracy of comparison with the historical lane line can be improved.
[0039] The processing in this embodiment may be performed only when the object position and the lane to which the object belongs are required, such as when there is a following vehicle that poses a collision risk, thereby saving necessary computer resources. The present invention may also be performed when a sensor capable of detecting rear white lines, such as a rear camera, is mounted, and the detection result by the sensor capable of detecting the rear white lines and the rear white lines estimated by this processing may be compared to improve accuracy and redundancy. EXAMPLES
[0040] In the second embodiment, a specific example of rear white line estimation, which is a main element in the target recognition device, will be described. FIG. 5 shows a specific processing flow example of rear white line estimation. The processing flow in FIG. 5 can be roughly divided into processes of a vehicle direction θ calculation unit St30, a moving distance and moving trajectory estimation unit St40, and a lane identification unit St50. The vehicle direction θ calculation unit St30 corresponds to the process of white line direction generation 24d in FIG. 2, the moving distance and moving trajectory estimation unit St40 corresponds to the process of moving distance estimation 31 and vehicle trajectory estimation 33 in FIG. 2, and the lane identification unit St50 corresponds to the process of the object position identification unit 40 in FIG. 2.
[0041] In the vehicle direction θ calculation unit St30 of the processing flow in Figure 5, first, in processing step St31, a time-series sequence of multiple white line points (array of x, y array) detected by sensor S4 (front camera) is acquired in each frame (control period: 50 ms period).
[0042] Next, in processing step St32, parameters are calculated for each white line point sequence. For example, a plurality of white line point sequences obtained in time series are expressed in each control cycle as an approximate straight line ax+by+c=0, and linear approximation is performed using the least squares method, for example. In the case of this formula, a, b, and c are calculated as parameters, where b>0, a 2 +b 2 = 1, and the major axis of the error ellipse obtained from the distribution of points is taken as the line segment length L.
[0043] In processing step St33, it is determined whether or not a white line has been detected. If a white line has been detected, the process proceeds to processing step St34, where the vehicle's orientation θ relative to the white line in each frame is calculated. For example, a weighted average of the coefficients of an approximation line, weighted by the line segment length, is taken for a sequence of multiple points to calculate the angle. For example, θ=atan2(Σ(L*a),Σ(L*b)) is calculated. Note that atan2 is a standard function of the C language that receives the y and x coordinate values of a point as arguments in that order, and returns the angle in radians that the straight line connecting the origin and the argument point makes with the x-axis.
[0044] If it is determined in step St33 that the white line has not been detected, the process proceeds to step St35, where it is estimated that the azimuth from the previous frame will be continued since the white line is not visible. An integral value based on the yaw rate when a curve or white line is not detected may be used.
[0045] In the processing step St32, an example of linear approximation is shown, but this may be polynomial approximation as an approximate curve. In this case, a curve can also be expressed, and the relative angle from the current value to the forward position can be obtained from the derivative of the polynomial and integrated. For example, y=ax 2 If it is expressed as +bx+c, the relative angle atan2(2ax, a) at the forward position x can be found using the derivative y'=ax+b. Also, if the approximation accuracy is improved by expressing it as a non-polynomial such as a circular arc, the relative angle can be found by differentiation in the same way.
[0046] In the processing flow of Fig. 5, in the movement distance and movement trajectory estimation unit St40, in processing step St41, the movement (dx, dy) from the last GNSS position and white line direction is calculated and integrated in order to estimate the movement trajectory along with the movement distance. As for the calculation method, the movement distance is calculated as the product of the speed and the elapsed time from the previous frame, and the movement direction is calculated by adding the product of the vehicle direction θ to the yaw rate and the elapsed time from the previous frame. This calculation method is simple and includes errors, but the errors can be reduced by dividing the elapsed time from the previous frame, for example.
[0047] In the lane identification unit St50 of the processing flow in FIG. 5, in processing step St51, the position of the white line at the target position is obtained and the lane is identified. Specifically, for example, the approximate curve parameters (quadratic approximation + range) for each frame are calculated and recorded, and the position at which the white line near the target can be detected with the highest quality is determined. For example, if the position 20 m ahead is the easiest to detect, then the position 20 m behind is determined. Using this, the frame of the self-position is obtained, and the lane is identified by comparing it with the position of the white line at that time.
[0048] According to the method for estimating rear white line in the second embodiment, a third method other than the rear white line estimation using GNSS and yaw rate can be provided, and the best method for estimating rear white line can be obtained by combining these methods while taking into account the advantages and disadvantages of each method. [Explanation of symbols]
[0049] 20: Distance / direction conversion unit, 30: Trajectory point sequence derivation unit, 31: Movement distance estimation unit, 32: Direction selection unit, 33: Vehicle trajectory estimation unit, 40: Object position identification unit, S1: GNSS, S2: Speed sensor, S3: Yaw rate sensor, S4: Front camera
Claims
1. a white line detection unit that detects a white line ahead of the vehicle; a white line direction generation unit that determines a direction of the vehicle relative to the white line for each predetermined time based on a detection result of the white line detection unit; a relative direction estimation unit that estimates a relative direction from the direction of the vehicle at a certain point based on the direction of the vehicle determined by the white line direction generation unit; a target detection unit that detects a target behind the vehicle; a movement distance estimation unit that estimates a movement distance of the vehicle; a vehicle path estimation unit that estimates a path of movement of the vehicle relative to the white line based on the movement distance at each of the predetermined times detected by the white line detection unit; a white line position estimation unit that estimates a position of a white line that is currently outside a detection range of the white line detection unit based on a detection result of the white line detection unit previously detected and the movement path; and a target position identification unit that identifies a positional relationship between the white line outside the detection range and the target based on the position of the white line outside the detection range estimated by the white line position estimation unit and the position of the target detected by the target detection unit, the relative direction estimation unit includes a direction selection unit that selects, when estimating a relative direction from the direction of the vehicle at a certain point based on the direction of the vehicle, the direction calculated by the white line direction generation unit, the direction calculated from the GNSS, or the direction calculated from a yaw rate; The direction selection unit normally selects the direction determined from the GNSS, and in a situation where the use of the GNSS is inappropriate, selects either the direction determined by the direction to white line generation unit or the direction determined from the yaw rate.
2. A target recognition device as claimed in claim 1, the direction selection unit selects the direction determined from the GNSS when the GNSS is good and the speed is fast, or when a first condition that the GNSS is good and there is little change in cumulative direction is met, selects the direction based on the group of dots of the white line ahead when the first condition is not met, and if there is much change in cumulative direction and the direction based on the group of dots of the white line ahead can be used, and selects the direction based on the yaw rate when the first condition is not met, and there is little change in cumulative direction and the direction based on the group of dots of the white line ahead cannot be used.
3. A target recognition device as claimed in claim 1, The detection result of the white line is a sequence of white line points that is periodically output, the sequence of white line points is expressed by a function, and a direction is calculated from the position of the sequence of white line points at different times.
4. A target recognition device as claimed in claim 1, A target recognition device, wherein the host vehicle trajectory estimation unit uses a change in relative orientation to estimate a movement trajectory of the host vehicle relative to the white line.
5. A target recognition device according to any one of claims 1 to 4, A target recognition device characterized in that the position and direction of a vehicle are associated and stored.
6. A target recognition device according to any one of claims 1 to 5. The white line position estimation unit records the positions of white lines detected in the past, and uses approximate parameters for the recording.
7. A target recognition device according to any one of claims 1 to 6. The white line position estimation unit uses the vehicle's position on the movement trajectory at the time when the white line closest to the target is captured and the recorded white line.
8. A method for target recognition using a computer, comprising: detects a white line ahead of the host vehicle, determines an orientation of the host vehicle relative to the white line at each predetermined time based on a result of the detection of the white line, estimates a relative orientation from the orientation of the host vehicle at a certain point based on the orientation of the host vehicle, detects a target object behind the host vehicle, estimates a movement distance of the host vehicle, estimates a movement trajectory of the host vehicle relative to the white line based on the estimated relative orientation at each predetermined time and the estimated movement distance at each predetermined time, estimates a position of a white line that is outside a detection range of a current white line detection based on detection results of the white line detected in the past and the movement trajectory, and identifies a positional relationship between the white line outside the detection range and the target object based on the estimated position of the white line outside the detection range and the detected position of the target object, When estimating a relative direction from the orientation of the vehicle at a certain point based on the orientation of the vehicle, selecting one of the orientation of the vehicle with respect to the white line calculated at each predetermined time based on the detection result of the white line, the orientation calculated from the GNSS, and the orientation calculated from the yaw rate, A target recognition method characterized in that a direction obtained from GNSS is normally selected, and in a situation where the use of GNSS is inappropriate, either the direction of the vehicle relative to the white line obtained at each specified time or a direction obtained from the yaw rate is selected based on the detection result of the white line.
Citation Information
Patent Citations
Travel support device
JP2019046150A
Compartment line recognition device
JP2019194037A
Lane fusion system using forward-view and rear-view cameras
US20120062747A1
Object recognition apparatus and vehicle travel controller using same
WO2015194371A1
