CN113221638A

CN113221638A - Vehicle target sensing method and system

Info

Publication number: CN113221638A
Application number: CN202110349952.4A
Authority: CN
Inventors: 丛伟伦; 王宽; 汤兆丰; 李涛; 任凡; 邓皓匀; 陈剑斌; 熊新立
Original assignee: Chongqing Changan Automobile Co Ltd
Current assignee: Chongqing Changan Automobile Co Ltd
Priority date: 2021-03-31
Filing date: 2021-03-31
Publication date: 2021-08-06
Anticipated expiration: 2041-03-31
Also published as: CN113221638B

Abstract

The invention provides a vehicle target sensing method and system, which are based on the detection capability of a monocular camera, detect lane lines in the driving process of a vehicle, clarify the driving track of the vehicle, compensate blind areas of the camera by additional ultrasonic waves, detect the transverse and longitudinal distances of left and right vehicles of the vehicle, and confirm whether the left and right vehicles invade the lane through lane line data to be dangerous targets, so that the vehicle can give early warning in advance, traffic accidents are reduced, and the safety of vehicle personnel is ensured.

Description

Vehicle target sensing method and system

Technical Field

The invention belongs to the field of intelligent driving of automobiles, and particularly relates to a perception technology for a target vehicle IN a CUT-IN scene.

Background

Chinese patent document CN201920607552.7 discloses an ultrasonic ranging obstacle avoidance automobile safety early warning system, which comprises a single chip microcomputer, an ultrasonic receiving and detecting circuit, an ultrasonic transmitting circuit and an alarm circuit, wherein the ultrasonic ranging system is installed in an automobile system, and the automobile automatic obstacle avoidance is realized by effectively and rapidly measuring the distances between the front and the rear of the automobile and between the left and the right of the automobile, so that the driver is prevented from errors caused by obstacle avoidance through human eye observation. The technology has the problems that only the distance between an obstacle and a vehicle is considered, and whether the obstacle invades a lane under a normal driving scene or not is not considered, so that the condition of rear-end collision is possibly caused.

In addition, chinese patent document CN201920351678.2 proposes a vehicle radar warning device and a vehicle, which includes at least one millimeter wave radar installed on a side of the vehicle, at least one ultrasonic radar installed around the vehicle, and a warning host computer for acquiring monitoring signals of the millimeter wave radar and the ultrasonic radar in real time and generating warning information, and performing obstacle monitoring and warning on the vehicle in high-speed and low-speed driving states respectively. Although the technology makes up the disadvantages of ultrasonic waves in a high-speed scene through the millimeter wave radar, the transverse and longitudinal distances of the target can be still detected in the high-speed scene, whether the target invades the lane is not verified, early warning cannot be given in advance, and potential safety hazards exist

Disclosure of Invention

The invention provides a vehicle target sensing method and a vehicle target sensing system, which solve the problem of fine sensing when a left vehicle and a right vehicle invade a lane under a low-speed CUT-IN scene, namely accurately sensing the transverse and longitudinal position and the course angle of a vehicle target, realizing the sensing and tracking of the targets at the two sides under the CUT-IN scene by the detection capability of a monocular camera on a lane line and the accurate detection capability of an ultrasonic radar on the transverse and longitudinal distances of the targets at the left side and the right side of the vehicle under the low-speed road condition, and ensuring the early warning capability of the vehicle and the personal safety of personnel on the vehicle.

The technical scheme of the invention is as follows:

a vehicle object perception method, comprising the steps of:

step 1, receiving lane line data acquired by a monocular camera and data acquired by an ultrasonic radar for vehicle targets of two adjacent lanes; the data acquired by the ultrasonic radar comprises target data in adjacent lanes on the left side and the right side, the transverse and longitudinal distance of the target and a sector of a detection target. The sector of the detection target is a sector of an area divided when the ultrasonic sensor mounted on the vehicle detects the ultrasonic sensor. The data acquired by the monocular camera comprise four lane line data of a left lane, a right lane and a right lane.

And 2, judging the reasonability of target data acquired by the ultrasonic radar and tracking the target.

And 3, judging whether the target has a tendency of entering the lane to become a dangerous target.

Further, the step 2 comprises:

step 2.1, detecting whether the transverse and longitudinal distances of the target indicate that the target exists in a side lane or not, and detecting whether the transverse and longitudinal distance jumping is reasonable or not;

and 2.2, tracking a reasonable side lane target, adding a tracking ID to the target, and if the ID changes, changing the target, wherein the target is not the previous target.

Further, the step 3 comprises: and fitting the course of the target by the left lane line and the right lane line data of the lane lines and tracking the transverse and longitudinal distance of the target, and then calculating whether the transverse distance of the target has a tendency of invading the lane. Usually, according to the target lateral distance and the heading angle, the distance from the head position of the target vehicle to the lane can be calculated, and if the distance value steadily and continuously decreases or the distance value is negative, the target vehicle has a tendency of invading the lane.

The calculation of whether the lateral distance of the target has a tendency of invading the lane specifically includes:

and further judging whether the target invades the lane according to lane line data, if the target wheel is pressed, the target is considered to have the tendency of invading the lane, otherwise, the vehicle is considered to have no tendency of invading the lane.

Further, the method also comprises a step 4 of transmitting all target data to a vehicle target output module, converting the target data into a CAN data format and then transmitting the CAN data format to a CAN bus.

Further, the method also comprises the step 1 of preprocessing the lane line data, wherein the preprocessing comprises the completion and the prediction of the lane line.

The invention also provides a vehicle target sensing system, which realizes the method and comprises the monocular camera, the ultrasonic radar and the sensor controller.

The monocular camera acquires lane line data, the ultrasonic radar acquires data of vehicle targets of adjacent lanes on two sides, and the data comprises target data in the adjacent lanes on the left side and the right side, the transverse and longitudinal distance of the targets and a sector of a detection target; the monocular camera and the ultrasonic radar are connected with the sensor controller through CAN lines, the sensor controller receives and processes data, the reasonability of target data acquired by the ultrasonic radar is judged, the vehicle target is tracked, and whether the vehicle target has a tendency of invading the road to become a dangerous target or not is judged.

Further, the system also comprises a vehicle target output module which receives the target data processed by the sensor controller, converts the target data into a CAN data format and then sends the CAN data format to the CAN bus.

The invention is based on a sensor system controller, a monocular camera and an ultrasonic radar, realizes the perception and tracking of targets on two sides of a vehicle IN a CUT-IN scene, provides a method for solving the problem that target data cannot be predicted IN the scene, and has the following specific advantages:

1. the invention detects the lane line in the driving process of the vehicle based on the detection capability of the monocular camera, defines the driving track of the vehicle, compensates the blind area of the camera by the additional ultrasonic wave, detects the transverse and longitudinal distances of the left vehicle and the right vehicle of the vehicle, and confirms whether the left vehicle and the right vehicle invade the lane to be dangerous targets through the lane line data, thereby leading the vehicle to early warn in advance.

2. The invention uses ultrasonic waves to detect the transverse and longitudinal distances of close-distance targets in a low-speed scene with high precision, tracks the transverse and longitudinal distances of the targets on the left side and the right side, senses whether the targets invade the lane to become dangerous targets, makes a timely response, reduces traffic accidents and ensures the safety of vehicle personnel.

Drawings

FIG. 1 is a schematic diagram of a vehicle target sensing system based on ultrasonic and monocular cameras;

FIG. 2 is a flow chart of a vehicle object sensing method of the present invention;

FIG. 3 left vehicle CUT-IN scenario.

Detailed Description

The technology of the present invention is further explained below with reference to the drawings.

As shown in fig. 1, in the vehicle target sensing system architecture based on the ultrasonic and the monocular camera, the monocular camera and the ultrasonic radar are connected with the sensor controller through the CAN line, and the controller receives and processes data and sends target sensing information on the left side and the right side of the vehicle to the CAN bus for subsequent control.

The monocular camera is used for acquiring lane line data, the ultrasonic radar is used for acquiring data of vehicle targets in adjacent lanes on two sides, and the data comprises target data in the adjacent lanes on the left side and the right side, the transverse and longitudinal distance of the targets and sectors for detecting the targets. The monocular camera and the ultrasonic radar are connected with the sensor controller through CAN lines, the sensor controller receives and processes data, the reasonability of target data acquired by the ultrasonic radar is judged, the vehicle target is tracked, and whether the vehicle target has a tendency of invading the road to become a dangerous target or not is judged. And the vehicle target output module receives the target data processed by the sensor controller, converts the target data into a CAN data format and then sends the CAN data format to the CAN bus.

The system is adopted to realize vehicle target perception, and mainly comprises three steps:

step 1, receiving monocular camera lane line data and data of ultrasonic waves for adjacent lane targets on two sides.

And 2, judging the reasonability of ultrasonic target data and tracking the target.

The above three steps will be described in detail with reference to the accompanying drawings:

as shown in fig. 2, a flow chart of a vehicle target sensing method is shown.

The method comprises the steps of firstly, respectively obtaining data through a monocular camera and an ultrasonic radar, obtaining four lane line data of a left lane, a right lane and a right lane from the monocular camera, and obtaining target data of the left lane, the right lane, the left lane, the right lane and the right lane from the ultrasonic radar, wherein the target data comprises the transverse distance and the longitudinal distance of a target and a sector of a detection target.

And then preprocessing the lane line data, wherein the preprocessing comprises the completion and prediction of the lane line. The prediction may be performed by fitting the lane line by the least square method based on a predetermined number of frames of data, for example, 20 frames of data, before the lane line is lost, and thereby predicting the lane line.

And then, carrying out target data reasonability detection, including detecting whether the transverse and longitudinal distances of the target indicate that the target exists in a side lane, detecting whether the transverse and longitudinal distances are reasonable in jumping, and the like. For example, according to the speed and distance of the received data, the horizontal and vertical distances of the next frame of target can be respectively calculated, and the uniform motion is performed through the speed value; or the distance of the next frame is calculated according to the change value of the distance, and whether the distance jump is reasonable or not is judged.

Then, a reasonable side-lane target is tracked, a tracking ID is added to the target, and if the ID changes, the target changes and is not the previous target.

And then fitting the course of the target by the left lane line and the right lane line data of the lane lines and tracking the transverse and longitudinal distance of the target, and then calculating whether the transverse distance of the target has the tendency of invading the lane.

And finally, all the target data are transmitted to a vehicle target output module, converted into a CAN data format and then transmitted to a CAN bus.

Here, a left vehicle CUT-IN scene is taken as an example, and as shown IN fig. 3, red is a left lane line and a right lane line detected by the monocular camera, a green region is an ultrasonic detection range, and a black module is an installation position of the ultrasonic radar on the vehicle. The left vehicle invades the lane from a left side lane CUT-IN, during the period, the corner radar and the side angle radar IN the front left side have the capability of detecting the target vehicle, stably output the transverse and longitudinal distance of the target vehicle, and track the change of the ultrasonic distance of the two sectors to know the motion trend and the course of the target vehicle; as shown IN the figure, the left vehicle CUT-IN is a dangerous target, is converted into a CAN data format through an output module and is sent to a CAN bus, and then contributes to subsequent control.

In a further embodiment, the logic for determining whether the vehicle is a dangerous target specifically comprises: after lane line data and data of target vehicles on the left side and the right side are obtained, tracking the transverse distance of the target, if the transverse distance of the target is reduced for 10 frames (namely 500ms), considering that the vehicle target has a approaching trend, otherwise, the vehicle is normally driven; and further judging whether the target invades the lane according to lane line data, if the target wheel is pressed, the target is considered as a dangerous target, otherwise, the vehicle is considered to run only in the side lane in an inclined manner.

Claims

1. A vehicle object perception method, comprising the steps of:

step 1, receiving lane line data acquired by a monocular camera and data acquired by an ultrasonic radar for vehicle targets of two adjacent lanes; the data acquired by the ultrasonic radar comprises target data in adjacent lanes on the left side and the right side, the transverse and longitudinal distance of the target and a sector of a detection target;

step 2, judging the reasonability of target data acquired by the ultrasonic radar and tracking the target;

2. The vehicle object perception method according to claim 1, wherein the data acquired by the monocular camera in step 1 includes four lane line data, left, right, and right.

3. The vehicle object perception method according to claim 1 or 2, wherein the step 2 includes:

4. The vehicle object perception method according to claim 1 or 2, wherein the step 3 includes: and fitting the course of the target by the left lane line and the right lane line data of the lane lines and tracking the transverse and longitudinal distance of the target, and then calculating whether the transverse distance of the target has a tendency of invading the lane.

5. The vehicle object perception method according to claim 4, wherein the calculating whether the lateral distance of the object has a tendency to invade the own lane is specifically:

if the transverse distance of the target is reduced for n frames, the vehicle target is considered to have a approaching trend, otherwise, the vehicle is normally driven; and further judging whether the target invades the lane according to lane line data, if the target wheel is pressed, considering that the target has the tendency of invading the lane, and if not, considering that the vehicle has no tendency of invading the lane.

6. The vehicle object sensing method according to claim 1, 2, 3 or 4, further comprising a step 4 of transmitting all object data to a vehicle object output module, converting the object data into a CAN data format and transmitting the CAN data format to a CAN bus.

7. The vehicle object perception method according to claim 1, 2, 3 or 4, further comprising performing lane line data preprocessing in step 1, wherein the preprocessing includes lane line completion and prediction.

8. A vehicle target perception system implementing the method of claims 1-7, comprising a monocular camera, an ultrasonic radar, and a sensor controller;

9. The vehicle object sensing system of claim 8, further comprising a vehicle object output module for receiving object data processed by the sensor controller, converting the object data into a CAN data format and transmitting the CAN data format to the CAN bus.