CN112162550B - Three-dimensional target tracking method for active safety collision avoidance of automobile - Google Patents
Three-dimensional target tracking method for active safety collision avoidance of automobile Download PDFInfo
- Publication number
- CN112162550B CN112162550B CN202010909302.6A CN202010909302A CN112162550B CN 112162550 B CN112162550 B CN 112162550B CN 202010909302 A CN202010909302 A CN 202010909302A CN 112162550 B CN112162550 B CN 112162550B
- Authority
- CN
- China
- Prior art keywords
- target
- automobile
- dimensional target
- dimensional
- collision avoidance
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0257—Control of position or course in two dimensions specially adapted to land vehicles using a radar
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/12—Target-seeking control
Abstract
The invention provides a three-dimensional space target tracking technology for active safe collision avoidance of an automobile, which belongs to the technical field of three-dimensional space target tracking, and comprises the following steps of firstly, acquiring detection results of all three-dimensional targets by a sensor and a perception algorithm which are installed on an automatic collision avoidance system of the automobile; expanding the detection results and the speed as target states to a three-dimensional space to obtain a motion model of each target at the current t moment, and further building a corresponding observation model; then, aiming at each target, processing a motion model and an observation model of the target through a UFIR batch processing filter by using all data of the target before the time t to obtain a prediction result of the target at the time t + 1; and comparing the prediction result at the time of t +1 with the judgment condition of the safety collision avoidance system, and actively controlling the automobile to avoid collision when the collision avoidance condition is triggered. The invention effectively links the previous frame and the next frame, enhances the robustness of the target state and the safety redundancy of the ADAS system.
Description
Technical Field
The invention belongs to the technical field of three-dimensional space target tracking, and particularly relates to a three-dimensional target tracking method for automobile active safety collision avoidance.
Background
As the ADAS System (Advanced Driving Assistance System) is more and more widely applied, the active safety collision avoidance System which is focused on becomes more and more important, and the safety redundancy thereof is particularly important as a safety function. However, the conventional automobile anti-collision system mainly utilizes ultrasonic ranging and infrared ranging, and the sensors have the defects of short measuring distance and large environmental influence, and moreover, the millimeter wave radar used in the current ADAS system has the defects of weak detection on pedestrians and partial non-metal obstacles, small visual field, low angular resolution and no elevation information; these drawbacks may all expose the drawbacks of active bumper systems.
And the three-dimensional target detection result obtained by calculation is high in precision and wide in visual field by utilizing the fusion of the camera and the laser radar and other sensors, and the defects can be overcome. However, when the simple three-dimensional target detection encounters an occluded or discontinuous target, target information is easily lost. Therefore, the continuity of the target can be enhanced by using a three-dimensional target tracking method, and the motion prediction of the shielded target can be carried out.
The current commonly used state estimation target tracking method usually uses a Kalman filter for prediction and update, but because the specific situation of the target is unknown, the noise in the filter is very difficult to determine, and the prediction effect is further influenced. Moreover, the method is an optimal estimation method under a linear gaussian condition, and the Infinite Impulse Response (IIR) characteristic of the method is easy to saturate, so that the filtering effect is reduced.
And the finite impulse response Filter (FIR) can avoid the statistical characteristics of noise, and the unbiased finite impulse response filter (UFIR) has stronger robustness to model errors and noise, so that the Kalman filter can be popularized in the state estimation problem.
Disclosure of Invention
Aiming at the existing problems, the invention provides a three-dimensional target tracking method for automobile active safety collision avoidance, which is based on three-dimensional target tracking of an unbiased finite impulse response filter (UFIR) to enhance the redundancy of automobile active safety collision avoidance perception.
The method comprises the following specific steps:
firstly, aiming at an automatic automobile collision avoidance system, acquiring detection results (x, y, z, ry, l, w, h and s) of all three-dimensional targets by using an installed sensor and a perception algorithm;
the sensor comprises a monocular camera, a binocular camera, a laser radar, or the fusion of the laser radar and the camera;
the perception algorithm comprises pure image three-dimensional target detection, pure laser radar three-dimensional target detection, laser radar and camera fused three-dimensional target detection and the like;
(x, y and z) are coordinates of each three-dimensional target under the automobile coordinate system, ry is an included angle between the orientation of each three-dimensional target and an x axis of an origin coordinate system, (l, w and h) are respectively the length, width and height corresponding to each three-dimensional target, and s is a classification confidence score.
And step two, aiming at the current time t, the detection result and the speed of each three-dimensional target are used as target states, and the target states are expanded to a three-dimensional space to obtain a motion model of each three-dimensional target.
The equation of the motion model is:
xt=Fxt-1+wt
wherein xtIs the state of a certain three-dimensional object at the current time t,
i.e. xt=[x y z ry l w h vx vy vz]T(ii) a Wherein v isx,vy,vzRepresenting the speed of the three-dimensional target along the xyz direction under the coordinate system of the automobile.
xt-1Is the state of the target at time t-1, F is the state transition matrix, wtThe mean value at time t is 0, the covariance matrix is Q, and the process noise follows a normal distribution.
Thirdly, building an observation model corresponding to each three-dimensional target by using the motion model of the target at the time t;
the observation model equation is:
zt=Hxt+vt
wherein z istFor the state measurement of each target at time t, H is the measurement matrix, vtThe mean value at the time t is 0, the covariance matrix is R, and the measurement noise obeys normal distribution;
step four, aiming at each three-dimensional target, processing a motion model and an observation model of the target through a UFIR batch processing filter by using all data of the target before the time t to obtain a prediction result of the target at the time t + 1;
the prediction equation is:
where m is t-N +1, and N is all the historical data before the adopted time t, that is, the latest N pieces of historical observation data.Coefficient matrix A for an unbiased state estimate at time t +1tI.e. the state transition matrix F.
Ct,mTo map the matrix, Ct,m=[Hm Hm-1 … Ht]T;
Yt,mTo expand the observation vector, Yt,m=[Zm T Zm-1 T … Zt T]T;
By adjusting the value of the historical data amount N, the influence on the prediction result at the moment t +1 is observed: the smaller N, the smaller the historical data memorized, the heavier the new data, the faster the new data response, but the less the noise filtering effect. On the contrary, the larger N is, the more historical data is, the smaller noise influence is, the smaller new data weight is, the larger model inertia is, and the slower response is.
After the prediction equation, an observation equation is executed, and model parameters are updated using the observation data.
And fifthly, comparing the prediction result of each three-dimensional target at the moment of t +1 with the judgment condition of the safety anti-collision system, and actively controlling the automobile to avoid collision when the anti-collision condition is triggered.
The discrimination conditions comprise three aspects of speed, distance and speed direction:
when the relative velocity vector of the three-dimensional target is opposite to the velocity vector of the automobile in direction and the same in magnitude, the three-dimensional target is in a static state; or the relative speed direction of the three-dimensional target points to the automobile, and the relative speed is greater than the speed threshold value of the safety anti-collision system, indicating that the target possibly points to the automobile at high speed, and needing to take safety anti-collision measures;
when the distance between the three-dimensional target and the automobile is within the distance threshold value of the safety collision avoidance system, the target is close to the automobile, and safety collision avoidance measures are necessary to be taken.
The invention has the advantages that:
(1) according to the three-dimensional target tracking method for the active safety collision avoidance of the automobile, the t +1 moment is predicted and tracked through all historical data before the t moment, the problem of frame dispersion before and after a target is solved, the continuity on the target time sequence is improved, and robustness is provided for small amount of target shielding. And the three-dimensional speed of the target can be predicted, the problem that the previous and the next frames of pure target detection are not associated is solved, and the three-dimensional space speed of the target is predicted.
(2) The three-dimensional target tracking method for the active safety collision avoidance of the automobile avoids the noise characteristic which is difficult to obtain, and has stronger robustness for target tracking. The predicted position and speed accuracy is higher, and the method can be used for judging conditions of an automobile active safety anti-collision system and comprehensively considering the target position and speed.
Drawings
Fig. 1 is a flowchart of a three-dimensional target tracking method for active safety collision avoidance of an automobile according to the present invention.
Fig. 2 is a tracking result of the three-dimensional target tracking method of the present invention on the KITTI data set.
Detailed Description
In order to make the technical method, objects and advantages of the invention more apparent, the invention will be described in further detail.
According to the three-dimensional target tracking method for the active safety collision avoidance of the automobile, the fragmentation problem of pure three-dimensional target detection is solved by predicting and tracking the result of the three-dimensional target detection, the state estimation and prediction are performed on the transient shielding of the target in connection with the previous frame and the next frame, the target detection performance is effectively improved, the estimated speed can be used as the active safety collision avoidance judgment condition of the other dimension, and the safety redundancy is higher.
The method comprises the steps of firstly establishing a motion state space model of a target, then setting parameters such as the length of a UFIR window, inputting a current frame and a historical frame into a UFIR filter, estimating to obtain the target state of the next frame, updating the filter when observation data are obtained through an observation equation, and simultaneously comparing a predicted result with a judgment condition to realize the active safe collision avoidance of the automobile.
As shown in fig. 1, the specific algorithm is:
firstly, aiming at an automatic automobile collision avoidance system, acquiring detection results (x, y, z, ry, l, w, h and s) of all three-dimensional targets by using an installed sensor and a perception algorithm;
the sensor comprises a monocular camera, a binocular camera, a laser radar, or the fusion of the laser radar and the camera; the embodiment uses laser radar and camera fusion as sensors.
The perception algorithm comprises pure image three-dimensional target detection, pure laser radar three-dimensional target detection, laser radar and camera fused three-dimensional target detection and the like; in this embodiment, a three-dimensional target detection algorithm in which a laser radar and a camera are fused is used to obtain a target result (x, y, z, ry, l, w, h, s); (x, y and z) are coordinates of each three-dimensional target under the automobile coordinate system, ry is an included angle between the orientation of each three-dimensional target and an x axis of an origin coordinate system, (l, w and h) are respectively the length, width and height corresponding to each three-dimensional target, and s is a classification confidence score.
And step two, aiming at the current time t, the detection result and the speed of each three-dimensional target are used as target states, and the target states are expanded to a three-dimensional space to obtain a motion model of each three-dimensional target.
The state space model state prediction equation of the discrete system model is as follows:
xt=Fxt-1+wt
wherein xtThe state of a certain three-dimensional target at the current time t is formed by the detection result and the speed of the target.
I.e. xt=[x y z ry l w h vx vy vz]T(ii) a Wherein v isx,vy,vzRepresenting the speed of the three-dimensional target along the xyz direction under the coordinate system of the automobile.
xt-1Is the state of the target at time t-1, F is the state transition matrix, wtThe mean value at time t is 0, the covariance matrix is Q, and the process noise follows a normal distribution.
In the object tracking problem here, the motion parameters of the object are generally taken as: such as speed, position, etc. as state variables of the system. Since the acquired target state is a three-dimensional target detection result, the target state is expanded to a three-dimensional space for filtering.
Target State x in the present exampletThe modification can be made according to the actual detection result, and the motion model used includes but is not limited to a constant speed model, namely xt=xt-1+ v Δ t. Taking a state transition matrix F according to the motion model as:
the process noise covariance matrix Q is:
the prior estimated covariance matrix P is:
thirdly, building an observation model corresponding to each three-dimensional target by using the motion model of the target at the time t;
the observation model equation is:
zt=Hxt+vt
wherein z istFor the state measurement of each target at time t, H is the measurement matrix, vtThe mean value at the time t is 0, the covariance matrix is R, and the measurement noise obeys normal distribution; suppose here that the noise v is measuredtAnd process noise wtAre independent of each other.
The measurement matrix in this embodiment is a state variable. Taking a measurement matrix covariance matrix R as:
step four, aiming at each three-dimensional target, processing a motion model and an observation model of the target through a UFIR batch processing filter by using all data of the target before the time t to obtain a prediction result of the target at the time t + 1;
the prediction equation is:
where m is t-N +1, and N is all the historical data before the adopted time t, that is, the latest N pieces of historical observation data.Coefficient matrix A for an unbiased state estimate at time t +1tI.e. the state transition matrix F.
Ct,mTo map the matrix, C is used due to the use of N frames of datat,m=[Hm Hm-1 … Ht]T;
Yt,mTo expand the observation vector, Yt,m=[Zm T Zm-1 T … Zt T]T;
By adjusting the value of the historical data amount N, the influence on the prediction result at the moment t +1 is observed: the smaller N, the smaller the historical data memorized, the heavier the new data, the faster the new data response, but the less the noise filtering effect. On the contrary, the larger N is, the more historical data is, the smaller noise influence is, the smaller new data weight is, the larger model inertia is, and the slower response is.
In this embodiment, the window length is obtained by using the minimum mean square error method, that is: vt=E{etet T};
Where e is the error between the actual and estimated values of the state variables,i.e. the optimal window length N is estimated such that the mean square error is minimized. In this embodiment, the UFIR filter performs iterative prediction on the historical N data to obtain an estimation result of the state variable.
After the prediction equation, an observation equation is executed, and model parameters are updated using the observation data.
The result of the tracking method is shown in fig. 2, in which information of ID, speed, distance, orientation, etc. of the target can be seen.
And fifthly, comparing the prediction result of each three-dimensional target at the moment of t +1 with the judgment condition of the safety anti-collision system, and actively controlling the automobile to avoid collision when the anti-collision condition is triggered.
The discrimination conditions comprise three aspects of speed, distance and speed direction:
when the relative velocity vector of the three-dimensional target is opposite to the velocity vector of the automobile in direction and the same in magnitude, the three-dimensional target is in a static state; or the relative speed direction of the three-dimensional target points to the automobile, and the relative speed is greater than the speed threshold value of the safety anti-collision system, indicating that the target possibly points to the automobile at high speed, and needing to take safety anti-collision measures;
when the distance between the three-dimensional target and the automobile is within the distance threshold value of the safety collision avoidance system, the target is close to the automobile, and safety collision avoidance measures are necessary to be taken.
The above description is only for the specific implementation of the present invention, and the present invention is not limited to the embodiments described herein, and includes other equivalent methods within the scope of the present invention, or methods with similar purpose.
Claims (5)
1. A three-dimensional target tracking method for automobile active safety collision avoidance is characterized by comprising the following specific steps:
firstly, aiming at an automatic automobile collision avoidance system, acquiring detection results (x, y, z, ry, l, w, h and s) of all three-dimensional targets by using an installed sensor and a perception algorithm;
(x, y and z) are coordinates of each three-dimensional target under the automobile coordinate system, ry is an included angle between the orientation of each three-dimensional target and an x axis of an origin coordinate system, (l, w and h) are respectively the length, width and height corresponding to each three-dimensional target, and s is a classification confidence score;
step two, aiming at the current time t, the detection result and the speed of each three-dimensional target are used as target states, and the target states are expanded to a three-dimensional space to obtain a motion model of each three-dimensional target;
the equation of the motion model is:
xt=Fxt-1+wt
wherein xtIs the state of a certain three-dimensional object at the current time t,
i.e. xt=[x y z ry l w h vx vy vz]T(ii) a Wherein v isx,vy,vzRepresenting the speed of the three-dimensional target along the xyz direction under the automobile coordinate system;
xt-1is the state of the target at time t-1, F is the state transition matrix, wtProcess noise at time t;
thirdly, building an observation model corresponding to each three-dimensional target by using the motion model of the target at the time t;
the observation model equation is:
zt=Hxt+vt
wherein z istFor the state measurement of each target at time t, H is the measurement matrix, vtIs the measurement noise at time t;
step four, aiming at each three-dimensional target, processing a motion model and an observation model of the target through a UFIR batch processing filter by using all data of the target before the time t to obtain a prediction result of the target at the time t + 1;
the prediction equation is:
wherein m is t-N +1, and N is all historical data before the adopted t moment, namely the latest N pieces of historical observation data;coefficient matrix A for an unbiased state estimate at time t +1tNamely a state transition matrix F;
Ct,mto map the matrix, Ct,m=[Hm Hm-1…Ht]T;
Yt,mTo expand the observation vector, Yt,m=[Zm T Zm-1 T]Zt T]T;
And fifthly, comparing the prediction result of each three-dimensional target at the moment of t +1 with the judgment condition of the safety anti-collision system, and actively controlling the automobile to avoid collision when the anti-collision condition is triggered.
2. The method for tracking the three-dimensional target for the active safety collision avoidance of the automobile according to claim 1, wherein the sensor in the first step comprises a monocular camera, a binocular camera, a laser radar or a fusion of the laser radar and the camera;
the perception algorithm comprises pure image three-dimensional target detection, pure laser radar three-dimensional target detection and laser radar and camera fused three-dimensional target detection.
3. The three-dimensional target tracking method for active safety collision avoidance of the automobile as claimed in claim 1, wherein in the fourth step, after the prediction equation, the observation equation is executed, and the model parameters are updated by using the observation data.
4. The three-dimensional target tracking method for the active safety collision avoidance of the automobile according to claim 1, wherein in the fourth step, the influence on the predicted result at the t +1 moment is observed by adjusting the value of the historical data quantity N: the smaller N is, the smaller the memorized historical data is, the larger the weight of new data is, the faster the new data response is, but the noise filtering effect is weakened; on the contrary, the larger N is, the more historical data is, the smaller noise influence is, the smaller new data weight is, the larger model inertia is, and the slower response is.
5. The three-dimensional target tracking method for the active safety collision avoidance of the automobile according to claim 1, wherein the distinguishing conditions in the fifth step include three aspects of speed magnitude, distance and speed direction:
when the relative velocity vector of the three-dimensional target is opposite to the velocity vector of the automobile in direction and the same in magnitude, the three-dimensional target is in a static state; or the relative speed direction of the three-dimensional target points to the automobile, and the relative speed is greater than the speed threshold value of the safety anti-collision system, indicating that the target possibly points to the automobile at high speed, and needing to take safety anti-collision measures;
when the distance between the three-dimensional target and the automobile is within the distance threshold value of the safety collision avoidance system, the target is close to the automobile, and safety collision avoidance measures are necessary to be taken.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010909302.6A CN112162550B (en) | 2020-09-02 | 2020-09-02 | Three-dimensional target tracking method for active safety collision avoidance of automobile |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010909302.6A CN112162550B (en) | 2020-09-02 | 2020-09-02 | Three-dimensional target tracking method for active safety collision avoidance of automobile |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112162550A CN112162550A (en) | 2021-01-01 |
CN112162550B true CN112162550B (en) | 2021-07-16 |
Family
ID=73858639
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010909302.6A Active CN112162550B (en) | 2020-09-02 | 2020-09-02 | Three-dimensional target tracking method for active safety collision avoidance of automobile |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112162550B (en) |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103294907A (en) * | 2013-05-21 | 2013-09-11 | 北京航空航天大学 | Trace model based abnormal degradation data inspection and analysis method |
CN107992052A (en) * | 2017-12-27 | 2018-05-04 | 纳恩博(北京)科技有限公司 | Method for tracking target and device, mobile equipment and storage medium |
CN108919837A (en) * | 2018-07-12 | 2018-11-30 | 北京航空航天大学 | A kind of dynamic (dynamical) automatic driving vehicle Second Order Sliding Mode Control method of view-based access control model |
CN109034018A (en) * | 2018-07-12 | 2018-12-18 | 北京航空航天大学 | A kind of low latitude small drone method for barrier perception based on binocular vision |
CN109808695A (en) * | 2019-02-28 | 2019-05-28 | 北京航空航天大学 | A kind of car travel mode method of adjustment and device |
CN109961461A (en) * | 2019-03-18 | 2019-07-02 | 湖南工程学院 | A kind of multiple mobile object tracking based on three-dimensional layered graph model |
CN110301864A (en) * | 2019-05-05 | 2019-10-08 | 珠海格力电器股份有限公司 | A kind of intellective dust collector system and its follower method based on millimetre-wave radar |
EP3579174A1 (en) * | 2018-06-08 | 2019-12-11 | Hexagon Technology Center GmbH | Mobile vehicles in manufacturing |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US10539669B2 (en) * | 2014-10-08 | 2020-01-21 | Texas Instruments Incorporated | Three dimensional (3D) tracking of objects in a radar system |
CN107315413B (en) * | 2017-07-12 | 2020-07-21 | 北京航空航天大学 | Multi-vehicle cooperative positioning algorithm considering relative positions between vehicles in vehicle-vehicle communication environment |
US10909377B2 (en) * | 2018-04-18 | 2021-02-02 | Baidu Usa Llc | Tracking objects with multiple cues |
US11807227B2 (en) * | 2018-11-02 | 2023-11-07 | Intel Corporation | Methods and apparatus to generate vehicle warnings |
CN111198567B (en) * | 2020-01-17 | 2021-06-01 | 北京大学 | Multi-AGV collaborative dynamic tracking method and device |
-
2020
- 2020-09-02 CN CN202010909302.6A patent/CN112162550B/en active Active
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103294907A (en) * | 2013-05-21 | 2013-09-11 | 北京航空航天大学 | Trace model based abnormal degradation data inspection and analysis method |
CN107992052A (en) * | 2017-12-27 | 2018-05-04 | 纳恩博(北京)科技有限公司 | Method for tracking target and device, mobile equipment and storage medium |
EP3579174A1 (en) * | 2018-06-08 | 2019-12-11 | Hexagon Technology Center GmbH | Mobile vehicles in manufacturing |
CN108919837A (en) * | 2018-07-12 | 2018-11-30 | 北京航空航天大学 | A kind of dynamic (dynamical) automatic driving vehicle Second Order Sliding Mode Control method of view-based access control model |
CN109034018A (en) * | 2018-07-12 | 2018-12-18 | 北京航空航天大学 | A kind of low latitude small drone method for barrier perception based on binocular vision |
CN109808695A (en) * | 2019-02-28 | 2019-05-28 | 北京航空航天大学 | A kind of car travel mode method of adjustment and device |
CN109961461A (en) * | 2019-03-18 | 2019-07-02 | 湖南工程学院 | A kind of multiple mobile object tracking based on three-dimensional layered graph model |
CN110301864A (en) * | 2019-05-05 | 2019-10-08 | 珠海格力电器股份有限公司 | A kind of intellective dust collector system and its follower method based on millimetre-wave radar |
Non-Patent Citations (2)
Title |
---|
Object detection and object tracking in front of the vehicle using front view camera;Juraj Ciberlin,等;《IEEE》;20191231;第27-32页 * |
深度学习的目标跟踪算法综述;李玺,等;《中国图像图形学报》;20191231;第2057-2080页 * |
Also Published As
Publication number | Publication date |
---|---|
CN112162550A (en) | 2021-01-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9199643B1 (en) | Sensor odometry and application in crash avoidance vehicle | |
CN112285714A (en) | Obstacle speed fusion method and device based on multiple sensors | |
CN111521195B (en) | Intelligent robot | |
CN111666891B (en) | Method and device for estimating movement state of obstacle | |
Schneiderman et al. | A discriminating feature tracker for vision-based autonomous driving | |
CN111578894B (en) | Method and device for determining heading angle of obstacle | |
Kim et al. | Detection and tracking of road barrier based on radar and vision sensor fusion | |
CN112462381A (en) | Multi-laser radar fusion method based on vehicle-road cooperation | |
CN114495064A (en) | Monocular depth estimation-based vehicle surrounding obstacle early warning method | |
CN113189975A (en) | Method for determining position data and/or movement data of a vehicle | |
CN110889862A (en) | Combined measurement method for multi-target tracking in network transmission attack environment | |
US6956959B2 (en) | Apparatus for recognizing environment | |
CN110426714B (en) | Obstacle identification method | |
WO2021102676A1 (en) | Object state acquisition method, mobile platform and storage medium | |
CN112162550B (en) | Three-dimensional target tracking method for active safety collision avoidance of automobile | |
Domhof et al. | Multi-sensor object tracking performance limits by the cramer-rao lower bound | |
CN115792894A (en) | Multi-sensor fusion vehicle target tracking method, system and storage medium | |
US11613272B2 (en) | Lane uncertainty modeling and tracking in a vehicle | |
US11756312B2 (en) | Orientation-agnostic lane tracking in a vehicle | |
CN114861725A (en) | Post-processing method, device, equipment and medium for perception and tracking of target | |
CN115082562A (en) | External parameter calibration method, device, equipment, server and vehicle-mounted computing equipment | |
CN113511194A (en) | Longitudinal collision avoidance early warning method and related device | |
CN115440094B (en) | Obstacle detection method, device and storage medium for helicopter ground proximity warning | |
Alturas et al. | Modeling and Optimization of an Obstacle Detection System for Small Fixed-wing UAV | |
CN117250595B (en) | False alarm suppression method for vehicle-mounted millimeter wave radar metal well lid target |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |