Summary of the invention
The technical problem to be solved by the invention is to provide a kind of intelligent driving vehicles under extreme weather to perceive systemSystem and method, make it is unmanned can adapt to more scenes, for the normal of automatic driving vehicle and be safely operated guarantor is providedBarrier.
In order to solve the above technical problems, the technical solution used in the present invention is as follows.
Vehicle sensory perceptual system under misty rain scene, including laser radar sensor on the car body is arranged, laser radar is passedThe output end of sensor is connect with the input terminal of Vehicular intelligent controller；It is additionally provided on the car body defeated with Vehicular intelligent controllerThe millimetre-wave radar sensor and visual sensor group, the visual sensor group for entering end connection include binocular camera, rainQuantity sensor, temperature sensor, haze dust sensor, fog amount sensor and optical sensor.
A kind of vehicle cognitive method under misty rain scene, specifically includes the following steps:
A. after confirmation vehicle enters intelligent driving mode, Vehicular intelligent controller acquires environmental data by each sensor；
B. analyzing and determining vehicle periphery according to the data of sensor acquisition, whether there are obstacles and obstacle information；
C. decision running velocity and travelling route, and it is sent to vehicle correlation execution unit, driving vehicle is normalOperation.
Vehicle cognitive method under above-mentioned misty rain scene judges the method for vehicle-surroundings ambient conditions in step B are as follows:
B1. weather condition is judged according to p (z) value that optical sensor acquires first, as p (z) >=D, then vehicle is runCurrent environment is daytime；
Snowy day is judged whether it is according to p (m) value that temperature sensor acquires, and as p (m) < T0, is illustrated for snowy day, intoRow step B2；As p (m) >=T, the weather that snows is excluded, is further continued under being judged whether it is according to p (x) value that precipitation rain fall sensor acquiresRainy day illustrates as p (x) >=R for rainy day, progress step B2；As p (x) < R, rainy weather is excluded, is further continued for according to mist amountP (y) value of sensor acquisition judges whether there is mist, as F1 >=p (y) >=F0, then it represents that has certain mist amount, carries out step B2；As p (y) < F0, illustrating that current weather does not have mist, continue the pn acquired according to haze dust sensor) value judges whether it is haze dirt dayGas indicates certain haze dirt as M1 >=p (y) >=M0, carries out step B2；As p (y) < M0, then explanation does not have haze dirt, is walkedRapid B2；
B2. according to laser radar sensor and millimetre-wave radar sensor detected vehicle peripheral obstacle information, if there isBarrier then calculates sensor confidence P, and sensor confidence P is by laser radar confidence level P1, vision confidence level P2(xyzmn) and millimetre-wave radar confidence level P3 combination is constituted；
B3. it is calculated according to step B2 and obtains confidence level, the vehicle-surroundings information and step acquired in conjunction with binocular cameraThe judging result of B1 provides obstacle information.
Vehicle cognitive method under above-mentioned misty rain scene, the obstacle information include barrier size, orientation and fortuneScanning frequency degree.
Vehicle cognitive method under above-mentioned misty rain scene, the decision exported in step C are as follows: when weather is snowy day, rootSnowfall is judged according to the value that sensor confidence calculates, and Vehicle Speed is provided according to snowfall；When weather is rainy day,Rainfall is judged according to the value that sensor confidence calculates, and Vehicle Speed is provided according to rainfall；When weather is haze skyWhen, haze size is judged according to the value that sensor confidence calculates, Vehicle Speed is provided according to haze size；Root is gone back simultaneouslyAccording to the data that laser radar sensor, millimetre-wave radar sensor and binocular camera acquire, it is big to judge whether there is otherType barrier, such as there are other large obstacles, then the travel speed and travelling route of vehicle are provided according to obstacle information.
Due to using above technical scheme, the invention technological progress is as follows.
The present invention is applied in automatic driving vehicle, can accurately perceive the vehicle-surroundings environment under various extreme weathers,Make it is unmanned can adapt to more scenes, provided safeguard for automatic driving vehicle normal and safe operation.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, the present invention will be described in further detail.
Vehicle sensory perceptual system under a kind of misty rain scene, including laser radar sensor on the car body, laser thunder is arrangedOutput end up to sensor is connect with the input terminal of Vehicular intelligent controller；It is additionally provided on car body defeated with Vehicular intelligent controllerThe millimetre-wave radar sensor and visual sensor group, the visual sensor group for entering end connection include binocular camera, rainQuantity sensor, temperature sensor, haze dust sensor, fog amount sensor and optical sensor.
Laser radar sensor is for detecting obstacles around the vehicle, and 360 degree of detection range, 100 meters of detecting distance or so,Detection accuracy < 5cm has the characteristics that precision is high, range is wide, but penetrability is not strong, is easy to detect haze sleet etc. at barrierHinder object.
Millimetre-wave radar sensor be the millimere-wave band radar equipment based on Doppler effect, have it is at low cost, dynamically withThe remote feature of track, detecting distance, but noise is more, resolution ratio is low, and field angle is small, is easy erroneous detection, missing inspection static-obstacle thing.
Binocular camera based on dual camera drift angle principle be arranged, can cognitive disorders object attribute, calculate obstacle distance,Have the characteristics that at low cost, resolution is high, detecting distance is remote, but height is required to illumination, night is not available.
Precipitation rain fall sensor is used to measure the raindrop density in front windshield of vehicle.When there is no raindrop on glass, by sending outMost of light out all reflects windshield, and it is seldom to reflect the light intensity being received；When raindrop are more on glass, kept offWind glass-reflected is increased by received light intensity back, and then sensor output changes.For detect it is current whether rain andRainy size sets pre-value R, if current rainfall is greater than R, then it is assumed that rainfall can cause erroneous detection, phase to laser radar sensorGuan Xingwei 1,；If being less than R, then it is assumed that do not influence, correlation 0.
Temperature sensor is installed on vehicle outer wall, for detecting current environmental temperature, judges whether there is snowy possibilityProperty.Pre-value T0, T1 and probability M are set, wherein (0,1) M ∈, T0 is lower limit temperature pre-value initially set, is dropped at such a temperatureA possibility that snow, is big, and T1 is ceiling temperature, very small a possibility that snowfall at such a temperature, and between T0 to T1, possibility is placed in the middle,Probability M.
Haze dust sensor is installed on vehicle outer wall, and for detecting current vehicle surrounding enviroment dust particles, judgement is currentWhether there is haze dirt, set predetermined value M0, M1 and probability N, wherein (0,1) N ∈, M0 is lower limit grain number initially set, InOn laser radar without influence under the value, M1 is that upper grain value can generate bigger influence, M0 more than the value to radarTo might have influence, probability N between M1.
For optical sensor for perceiving current light, can relatively clearly go out respectively be currently daytime or night, ifPre-value D is determined, when current light p (z) is greater than D value, it is believed that be daytime, camera is strong correlation at this time, and detection data, which can be done, joinsIt examines, and higher with reference to weight；When current light p (z) is less than D value, it is believed that be night, camera is weak correlation, data at this timeWeak correlation should not.
Fog amount sensor is installed on vehicle outer wall, for detecting the mist amount size of vehicle-surroundings, it is judged whether or notMist sets predetermined value F0, F1 and probability Y, wherein (0,1) Y ∈, and F0 is lower numerical limit initially set, at this value to laserFor radar without influence, F1 is upper limit value, more than the value, bigger influence can be generated to radar, might have between F0 to F1It influences, probability Y.
Based on the vehicle cognitive method of vehicle sensory perceptual system under above-mentioned misty rain scene, specifically includes the following steps:
A. after confirmation vehicle enters intelligent driving mode, Vehicular intelligent controller acquires environmental data by each sensor.
B. analyzing and determining vehicle periphery according to the data of sensor acquisition, whether there are obstacles and obstacle information；BarrierHindering object information includes barrier size, orientation and the speed of service.
B1. weather condition is judged according to p (z) value that optical sensor acquires first, as p (z) >=D, then vehicle is runCurrent environment is daytime；
Snowy day is judged whether it is according to p (m) value that temperature sensor acquires, and as p (m) < T0, is illustrated for snowy day, intoRow step B2；As p (m) >=T, the weather that snows is excluded, is further continued under being judged whether it is according to p (x) value that precipitation rain fall sensor acquiresRainy day illustrates as p (x) >=R for rainy day, progress step B2；As p (x) < R, rainy weather is excluded, is further continued for according to mist amountP (y) value of sensor acquisition judges whether there is mist, as F1 >=p (y) >=F0, then it represents that has certain mist amount, carries out step B2；As p (y) < F0, illustrating that current weather does not have mist, continue the pn acquired according to haze dust sensor) value judges whether it is haze dirt dayGas indicates certain haze dirt as M1 >=p (y) >=M0, carries out step B2；As p (y) < M0, then explanation does not have haze dirt, is walkedRapid B2.
B2. according to laser radar sensor and millimetre-wave radar sensor detected vehicle peripheral obstacle information, if there isBarrier then calculates sensor confidence P, and sensor confidence P is by laser radar confidence level P1, vision confidence level P2(xyzmn) and millimetre-wave radar confidence level P3 combination is constituted；Confidence level judgement principle as shown in Figure 1, each sensor powerWeight ratio is as shown in Figure 2.
B3. it is calculated according to step B2 and obtains confidence level, the vehicle-surroundings information and step acquired in conjunction with binocular cameraThe judging result of B1 provides obstacle information.
C. decision running velocity and travelling route, and it is sent to vehicle correlation execution unit, driving vehicle is normalOperation.
Specific strategy are as follows: when weather is snowy day, snowfall is judged according to the value that sensor confidence calculates, according to dropSnowfall provides Vehicle Speed；When weather is rainy day, rainfall is judged according to the value that sensor confidence calculates, according toRainfall provides Vehicle Speed；When weather is haze sky, haze size is judged according to the value that sensor confidence calculates,Vehicle Speed is provided according to haze size；Simultaneously also according to laser radar sensor, millimetre-wave radar sensor and doubleThe data of mesh camera acquisition, judge whether there is other large obstacles, and such as there are other large obstacles, then according to obstacleObject information provides the travel speed Vp and travelling route of vehicle.