Specific embodiment
To make purpose, technical scheme and the advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present inventionAccompanying drawing, the technical scheme of the embodiment of the present invention is clearly and completely described.Obviously, described embodiment is thisBright a part of embodiment, rather than the embodiment of whole.Based on described embodiments of the invention, ordinary skillThe every other embodiment that personnel are obtained on the premise of without the need for creative work, belongs to the scope of protection of the invention.
Unless otherwise defined, technical term used herein or scientific terminology should be in art of the present invention and haveThe ordinary meaning that the personage of general technical ability is understood." first ", " second " and similar word used in the disclosure is notAny order, quantity or importance are represented, and is used only to distinguish different parts.Equally, " one ", " one " orThe similar word such as " being somebody's turn to do " does not indicate that quantity is limited yet, but represents and have at least one.What " including " or "comprising" etc. were similar toWord means to occur element before the word, and either object is covered the element or object that occur in the word presented hereinafter and its is waitedTogether, other elements or object are not excluded for." connection " either the similar word such as " connected " be not limited to physics orThe connection of machinery, and can be including electrical connection, either directly still indirectly." on ", D score, "left", "right"Etc. being only used for representing relative position relation, after the absolute position for being described object changes, then the relative position relation is likely toCorrespondingly change.
Embodiments of the invention provide a kind of determination methods of vehicle lane change trend, judgment means and Computer Storage and are situated betweenMatter, the determination methods of the vehicle lane change trend include：S1, chooses a frame road image as present frame road image；S2, identificationAnd the identification point of selection target vehicle；S3, recognizes lane line；S4, calculates the identification point to the vertical range of the lane lineAnd store；S5, according to the least one frame of road image before the present frame road image and the present frame road imageThe vertical range judge whether the target vehicle has lane change trend, if lane change trend, then carry out step S6, ifWithout lane change trend or lane change trend can not be judged, then carry out step S1；S6, exports lane change trend result.The present invention is implementedExample provide vehicle lane change trend determination methods in existing image acquiring device, for example, on the basis of camera, by drawingEnter a kind of mobile target image identification and movement tendency computational methods, recognize the identification point of target vehicle, calculate the identification point and arriveThe vertical range of lane line, is identified only recognizing the mark of target vehicle without the intact form to target vehiclePoint, is just capable of achieving the judgement of the lane change trend to target vehicle, so just reduces the lane change Trend judgement to target vehiclePostpone, realize the horizontal lane change trend to target vehicle and timely and effectively predict, effectively evaded target vehicle lane change toThe danger that Current vehicle brings.
First embodiment
Fig. 1 shows the exemplary process diagram of the determination methods of vehicle lane change trend according to a first embodiment of the present invention,As shown in figure 1, the determination methods of the vehicle lane change trend of first embodiment of the present invention offer include：S1, chooses a frame roadImage is used as present frame road image；The identification point of S2, identification and selection target vehicle；S3, recognizes lane line；S4, calculates markPoint is known to the vertical range of nearest lane line and store；S5, according to present frame road image before least one frame of roadThe vertical range of image and present frame road image judges whether target vehicle has lane change trend, if lane change trend, then entersRow step S6, if without lane change trend or can not judge lane change trend, carries out step S1；S6, output lane change trend knotReally.
Exemplarily, in step S5, judge whether target vehicle can be according to present frame road image with lane change trendWith the frame road image before present frame road image, two frame road images, three frame road images or more multiframe mileage chartTarget vehicle is carried out to the vertical range of nearest lane line as in, and for example, present frame road image is the i-th frame road image,The then judgement to target vehicle lane change trend can be carried out according to the i-th two field picture and the i-th -1 frame road image, or according toI frame road images, the i-th -1 frame road image and the i-th -2 frame road image and carry out, or according to the i-th frame road image, i-th -1 frame road image, the i-th -2 frame road image and the i-th -3 frame road image and carry out, or according to the i-th frame road image, i-th -1 frame road image, i-th -2 frame road image ... the i-th-n frames road image and carry out, wherein n is less than i and whole more than 0Number.
For example, if present frame road image is the first frame road image, at least frame before present frame road imageThe vertical range of road image will be obviously not present, then the step of determination methods S5 can not judge whether target vehicle hasLane change trend, S1-S4 the step of before would be repeated for.
Exemplarily, the least one frame of road image before present frame road image can be with present frame road imageThe road image of successive frame, or, the road image of successive frame is can not be, embodiments of the invention are not limited thisIt is fixed, as long as can interpolate that the lane change trend of target vehicle.
Exemplarily, in the first embodiment of the present invention, the identification point of target vehicle can be vehicle hub centre,Angle summit that angle summit that the moulding point of the covering of vehicle, vehicle are formed with ground area shading, the edge of vehicle are formed orThe characteristic point for carrying out feature extraction by convolutional neural networks and obtaining.
Correspondingly, the identification point of identification and selection target vehicle can include：In the wheel hub of identification and selection target vehicleThe angle top that the moulding point of the covering of the heart, identification and selection target vehicle, identification and selection target vehicle are formed with ground area shadingAngle summit or recognized and selection target car by convolutional neural networks that the edge of point, identification and selection target vehicle is formedCharacteristic point.
It should be noted that in order that the selection of target vehicle identification point is needed to mark with higher accuracy rateThe association attributes of point is limited, it is, needing to add more restriction in the identification simultaneously identification point of selection target vehicle.
For example, when selecting the angle summit that vehicle is formed with ground area shading as identification point, in addition it is also necessary to consider target vehicleThe length of edge line and the length of ground area shading and the corner dimension that the two is formed, choose to non-car so as to exclude as far as possibleIdentification point possibility.For example, for other buildings on road, for example, street lamp, it may have formed on the groundShade, but the length of ground area shading that formed of street lamp differs markedly from the length of the ground area shading that vehicle is formed, and itsThe length of edge line is also different from vehicle, and the angle that it is formed with ground area shading is cloudy also different from target vehicle and its groundThe angle that shadow is formed, the restriction more than, it is clear that can exclude and choose non-vehicle, for example, other on the road such as street lampThe possibility of object.
Correspondingly, in the determination methods of vehicle lane change trend according to a first embodiment of the present invention, recognize and select meshThe angle summit that mark vehicle is formed with ground area shading can include：Calculate and obtain the target vehicle edge line length withAnd the length of described ground area shading；Calculate and obtain the angle that the target vehicle is formed with described ground area shading；And according toThe length of the edge line, the length of described ground area shading and the angle are recognized from the road image and choose describedThe angle summit of target vehicle and ground area shading.
Exemplarily, when the hub centre of target vehicle is chosen as identification point, recognize and select the target vehicleHub centre the step of can include：The circular pattern or circular arc pattern in road image is recognized first；It is then determined that shouldHub centre of the center of circular pattern or circular arc pattern as target vehicle.Here circular arc pattern is referred to and can determineThe circular arc pattern at its center, it is, circular arc pattern needs the part for being circular pattern.
Further, in order to the circular pattern or circular arc pattern that can ensure that identification are the wheel hubs of target vehicle, in identificationDuring circular pattern or circular arc pattern in road image, the radial dimension for considering circular pattern or circular arc pattern is needed, it is rightFor road, the object for being rendered as circular pattern or circular arc pattern is exactly mainly the tire of vehicle, and with the wheel hub of vehicleThe object for circular pattern or circular arc pattern of size matching is few, so, is considering circular pattern or circular arcThe radial dimension of shape pattern and identify after circular pattern or circular arc pattern, improve the accuracy of identification of vehicle hub.ExampleProperty, the radial dimension of selection of circular pattern or circular arc pattern may be referred to the wheel hub of Vehicles Collected from Market all vehicles on saleParameter, for example, a diameter of 14 inches -18 inches of usual wheel hub, then identification road image in circular pattern or circular arcThe circular or circular arc figure of the picture size with the hub diameter corresponding to 14 inches -18 inches can be selected during patternCase.Correspondingly, recognize that circular pattern or circular arc pattern in road image can include：There is correspondence in identification road imageIn the circular or circular arc pattern of the picture size of 14 inches -18 inches of diameter.
Further, the precision identified in order to further improve wheel hub, in addition to considering size, it is also contemplated that addThe restriction of boss shape, by the internal circular or circular arc pattern with radial pattern in road image target vehicle is regarded asWheel hub, and then hub centre is regarded as at its center.Correspondingly, the circular pattern or circular arc pattern in road image is recognizedCan include：The circular or circular arc of the picture size with the diameter corresponding to 14 inches -18 inches in identification road imagePattern；Or, recognize the circular or circular arc of the picture size in road image with the diameter corresponding to 14 inches -18 inchesPattern and the internal circle with radial pattern and circular arc pattern are selected from the circle and circular arc pattern；Or,The internal circular pattern with radial pattern or circular arc pattern in identification road image.
Further, identification road image in circular pattern or circular arc pattern after and it is determined that circular pattern orBefore hub centre of the center of circular arc pattern as the target vehicle, the hub centre of identification and selection target vehicle is alsoIncluding：Select the region that the circular pattern or circular arc pattern are located；The circular pattern or circular arc pattern are locatedThe circular pattern or the fringe region in circular arc pattern region are screened and rejected in region, for example, rejects single framesAbout the three of the top of the left side of image and the fringe region on a right side and single-frame images/part, so as to remove some and meshThe unrelated region of mark vehicle image, for example, road edge or sky etc..
Further, the determination methods in the vehicle lane change trend of first embodiment of the invention choose a frame road describedImage as after present frame road image and identification and selection target vehicle identification point before can also include to describedPresent frame road image carries out image procossing.
Further, in order to lift judging efficiency, it is described carry out image procossing to the present frame road image afterAnd can also include selecting the region at the target vehicle place in the identification and before the identification point of selection target vehicle.
Exemplarily, in the first embodiment of the present invention, image procossing can include Image semantic classification, image conversion, sameState filtering, mask de-noising and edge extracting etc., Image semantic classification can include image interception, adjustment picture size etc., imageConversion can be that cromogram is converted to into gray-scale map, and in the first embodiment of the present invention, image procossing is not limited to the above and is limitedFixed, other image processing methods commonly used in the art can also be included.
Below to basis by taking the two continuous frames road image before present frame road image and present frame road image as an exampleThe step of determination methods of the vehicle lane change trend of first embodiment of the invention, S5 was illustrative.
Exemplarily, according to the least one frame of road image before the present frame road image and the present frame roadThe vertical range of road image judges whether the target vehicle has lane change trend to include：Choose the present frame figurePicture, for example, the road image of two continuous frames before the i-th two field picture, and the current frame image, for example, the i-th -1 frame road imageWith the i-th -2 frame road image, vertical range D of the identification point of middle target vehicle to the nearest lane linei、Di-1And Di-2,Wherein DiFor the current frame image, it is, the i-th two field picture, vertical range, Di-1Before the current frame imageFirst frame road image, it is, the i-th -1 frame road image, vertical range, Di-2Before for the current frame imageTwo frame road images, it is, the i-th -2 frame road image, vertical range；According to the identification point of the target vehicle of continuous three frameVertical range to the nearest lane line calculates and obtains any two difference value A and B；Judged according to lane change Trend rulesWhether the target vehicle has lane change trend, wherein the lane change Trend rules symbol that is included in two difference values is identical and twoIndividual difference value be all higher than first threshold or two difference values symbol is identical and the absolute value of one of described two difference values is bigIn another difference value and Second Threshold and in the case of, then judge that target vehicle has lane change trend, the first thresholdLateral velocity threshold is with the Second Threshold.
It should be noted that in the first embodiment of the present invention, identification point the hanging down to nearest lane line of target vehicleStraight distance can be defined as：When target vehicle be located at lane line left side when, vertical range be on the occasion of, accordingly, target vehicle positionWhen lane line right side, then vertical range is negative value.For example, the lane line A for being specified, the vehicle B on the left of itDistance to lane line A is on the occasion of the distance of the vehicle C on the right side of it to lane line A is negative value.
Further, in the determination methods of the vehicle lane change trend of first embodiment of the invention, according to described continuous threeThe vertical range of frame is calculated and obtains any two difference value A and B to be included：Calculate and obtain the described vertical of the frame of arbitrary neighborhood twoThe difference △ D1=Di-1-Di-2 and △ D2=Di-Di-1 of distance；According to formula：A=△ D1/T and B=△ D2/T are calculated simultaneouslyObtain described two difference values, wherein T be choose a frame road image as time of present frame road image, described image atThe time of reason, the time for recognizing and selecting the identification point, the time of identification lane line and the identification point is calculated to describedThe time sum of the vertical range of lane line.
Here, it should be noted that when the difference of vertical range of consecutive frame road image is calculated, be according to sameComputation rule is calculated, and can be that the vertical range of a later frame road image deducts hanging down for former frame road image for exampleStraight distance, or can be that the vertical range of former frame road image deducts the vertical range of a later frame road image, abilityIt is more than the selection that the technical staff in domain can be appropriate regular, as long as ensureing the difference of the vertical range of all consecutive frame road imagesAll adopt identical computation rule.
It is more than carrying out as a example by the lane change trend for judging vehicle by the vertical range according to continuous three frames road imageIllustrate, it is described above for being judged according to the vertical range of two continuous frames, four frames or more multiframe road imageIt is applicable.
For example, for being judged according to the vertical range of two continuous frames road image, using a later frame road imageVertical range deducts the vertical range of former frame road image or the vertical range of former frame road image deducts a later frame roadThe vertical range of road image obtains a difference value, then judges whether the target vehicle has lane change according to lane change Trend rulesTrend, here lane change Trend rules can be：More than a lateral velocity threshold, then target vehicle has lane change trend to the difference value,The acquisition of difference value is identical with the method for continuous three frame of the above, it is, calculating and obtaining the vertical range of this two frameDifference △ D=Di-Di-1；According to formula：A=△ D/T obtain difference value, and wherein T is to choose a frame road image as currentTime of frame road image, the time of described image process, the time for recognizing and selecting the identification point, identification lane line whenBetween and calculate the identification point to the time sum of the vertical range of the lane line.
For example, for being judged according to the vertical range of continuous four frames road image, it is three that difference is obtainDifference value A1, B1 and C1 (sequencing arrangement temporally), wherein A1 is that the second two field picture and the first two field picture enter in four framesRow computing is obtained, and B1 is that the 3rd two field picture and the second two field picture carry out computing and obtain in four frames, and C1 is the 4th frame in four framesImage and the 3rd two field picture carry out what computing was obtained), lane change Trend rules can be identical and three in the symbol of three difference valuesIndividual difference value is all higher than that the symbol of first threshold or three difference values is identical and arranges in chronological order in three difference valuesThe absolute value of posterior any one difference value of row in three difference values more than being sequentially arranged before the difference valueIt is adjacent differential value and Second Threshold and, for example, the absolute value of B1 more than A1 and Second Threshold and or C1 absolute valueMore than B1 and Second Threshold and in the case of, judge that target vehicle has lane change trend, the first threshold and described secondThreshold value is lateral velocity threshold, and remaining lane change for judging vehicle with the vertical range according to continuous three frames road image becomesThe method of gesture is identical.For continuous multiple frames road image or discontinuous frame road image in the case of, can be according to the aboveAnalogized, for the simplicity of description, will not be repeated here.
It should be noted that for a person skilled in the art, lateral velocity threshold here, for example, the first thresholdValue and Second Threshold, in the scope of 0.2m/s～0.3m/s, can be chosen according to actual conditions from the scope, but sameFirst threshold and Second Threshold are different values in a kind of lane change Trend judgement rule, and embodiments of the invention will not be carried out to thisLimit.
Further, in the determination methods of vehicle lane change trend according to a first embodiment of the present invention, the mark is calculatedKnowing point can include calculating the corresponding pixel of the identification point to the track to the vertical range of the nearest lane lineThe vertical range of line.
Here, it should be noted that the corresponding pixel of the identification point selected may have multiple, but select wherein manyAny one corresponding pixel in individual pixel, then calculate its vertical range for arriving nearest lane line all allows in errorIn the range of, can realize the purpose of the present invention.
Alternatively, calculating the identification point can also include to the vertical range of the nearest lane line：By the roadRoad image is converted to top plan view；Calculate vertical range of the identification point to the lane line described in the top plan view.
In embodiments of the invention, how to calculate the identification point and do not limit to the vertical range of the nearest lane lineImplementation method more than, as long as can realize the scheme of the purpose of the present invention in the protection domain of the embodiment of the present invention.
Exemplarily, before one frame road image of the selection is as present frame road image, according to present invention enforcementThe determination methods of the vehicle lane change trend of example, can also include obtaining road image using image acquiring device.
Alternatively, before one frame road image of the selection is as present frame road image, according to embodiments of the present inventionVehicle lane change trend determination methods can also include using image acquiring device obtain road video image.
Here, image acquiring device can be mounted in the camera of Current vehicle, and camera can be arranged on vehicleFront portion, as shown in Fig. 2 camera 101 is arranged on the front portion of vehicle.Alternatively, camera can also be arranged at the middle part of vehicleOr rear portion, as long as road image or video image can be obtained just can be with.
It should be noted that obtained using image acquiring device be the video image of road when, can be from video figureAs in choose a frame road image as present frame road image, and obtain be road image when, can be with control based on direct control chart pictureAcquisition device shoots real-time road image as present frame road image, or can be from the road captured by image acquiring deviceSelect a sub-picture as present frame road image in the image of road, present frame road image here means to show that Current vehicle is worked asThe road image at front moment.
It should be noted there that in an embodiment of the present invention, target vehicle refer in front of Current vehicle and according toAlong at least 6 vehicles that distance value of the lane line direction away from the Current vehicle sorts from small to large.
In the determination methods of vehicle lane change trend according to embodiments of the present invention, in front of the Current vehicle extremelyFew 6 vehicles, at each moment, to each vehicle the judgement of vehicle lane change trend are carried out, if finding one car of any of whichHave lane change trend, then export vehicle lane change judged result, Current vehicle can be reminded in the form of voice or imageDriver, or for automatic driving vehicle, the controller of Current vehicle can directly receive the judgement knot of vehicle lane changeReally, slow down or the corresponding behavior such as avoid so as to make.
It should be noted that the vehicle in front of Current vehicle is probably what is changed, it is, 6 vehicles at a upper momentAt least one is different with the 6 of subsequent time vehicles, and for determination methods according to embodiments of the present invention, it is sentenced all the timeDisconnected be current time be located at Current vehicle in front of vehicle lane change trend, if the A cars at a upper moment in subsequent timeJing then just no longer pays close attention to the traveling of A cars, if B cars are constantly at least not in the range of at least 6 vehicles in subsequent timeIn the range of 6 vehicles, then the determination methods will persistently track the traveling of B cars, it is, obtaining mileage chart at each momentAs and carry out the judgement of vehicle lane change trend, till its lane change.
It should be noted there that positioned at Current vehicle front refer to the headstock of the vehicle Current vehicle headstock itBefore.And, at least 6 vehicles can further be limited according to specific traffic rules, for example, for can only from work asVehicle in front left-hand lane could lane change traffic rules, at least 6 vehicles can be further defined in front of Current vehicle,At least 6 vehicles that distance value positioned at Current vehicle left-hand lane and away from the Current vehicle sorts from small to large.
The determination methods of vehicle lane change trend according to embodiments of the present invention are carried out in detail with reference to specific exampleExplanation.
Example 1
Using the angle summit of target vehicle and ground area shading formation as the identification point of target vehicle, Fig. 3 A- in the exampleFig. 3 F show identification point for target vehicle and the example of the judgement vehicle lane change on the angle summit of ground area shading formation.
Fig. 3 A show the frame road image obtained by image acquiring device, and what the cross spider in Fig. 3 B was identified is to knowOther lane line, in Fig. 3 C, the crosspoint of two lines mark be target vehicle identification point：Target vehicle and ground area shading shapeInto angle summit, so a target vehicle in front of Current vehicle is identified by identification point, rather than by carIntact form recognize the target vehicle, so as to when vehicle do not fully appear in Current vehicle front when just can recognize the car, so just can timely judge the lane change trend of the vehicle.Fig. 3 D are shown in the continuous multiple frames road image of acquisitionThe diagram of vertical range of the identification point of the target vehicle away from the left-hand lane line of Current vehicle, wherein ordinate is target vehicleVertical range of the identification point away from the lane line, its unit is 1, here 1 be the distance between neighbor pixel of image, exampleIf vertical range is the distance between 15 15 pixels of expression, abscissa is road image time series, from Fig. 3 D, meshThe vertical range of the identification point away from the lane line of mark vehicle all the time diminishing, and in the more than tenth frame road image, target vehicleVertical range of the identification point away from the lane line have been changed to negative value, it means that target vehicle has been positioned at the lane lineRight side, and in 20-30 frame road images, the vertical range of the identification point of target vehicle away from the lane line is stablyIt is between 15-30, it can be seen that, the vehicle has been completed lane change, and the track from the left of Current vehicle is incorporated to Current vehicleThe track at place.In the determination methods of the vehicle lane change trend of the embodiment of the present invention, according at least two frames of the 10th frame or soImage, just may determine that the obvious lane change trend of the vehicle, so as to export the result.
This point can be confirmed by the road image shown in Fig. 3 E and 3F, and as seen from the figure, the vehicle is finally from left sideTrack enters the track of Current vehicle.
Example 2
Using the angle summit of the edge formation of vehicle as the identification point of target vehicle in the example, Fig. 4 A- Fig. 4 F are illustratedThe example of the judgement vehicle lane change on the angle summit that identification point is formed for the edge of vehicle.
Fig. 4 A show the frame road image obtained by image acquiring device, and what the cross spider in Fig. 4 B was identified is to knowOther lane line, in Fig. 4 C, the crosspoint of two lines mark be target vehicle identification point：What the edge of target vehicle was formedAngle summit, so identifies a target vehicle in front of Current vehicle by identification point.Fig. 4 D show acquisitionThe diagram of vertical range of the identification point of the target vehicle in continuous multiple frames road away from the left-hand lane line of Current vehicle, wherein indulgingCoordinate is the vertical range of the identification point away from the lane line of target vehicle, and its unit is 1, here 1 be image neighborPoint the distance between, such as vertical range be the distance between 15 expression, 15 pixels abscissa be road image time sequenceRow, from Fig. 4 D, the vertical range of the identification point of target vehicle away from the lane line is kept approximately constant, it can be seen that, the carThere is no lane change trend.
This point can be confirmed by the road image shown in Fig. 4 E and 4F, and as seen from the figure, the vehicle is travelled all the time is working asThe left-hand lane of vehicle in front is without lane change.
From above example, determination methods of vehicle lane change trend according to an embodiment of the invention, by identification simultaneouslyThe identification point of selection target vehicle, and calculate the identification point to the vertical range of lane line, by track the vertical range withThe change of time, can accurately judge target vehicle whether lane change, it is, accurately judging the lane change of target vehicleTrend, the lane change trend without just can interpolate that vehicle in the case of the intact form of identification vehicle, so just canEnough accomplish the timely judgement to vehicle lane change trend, reduce the delay of the lane change Trend judgement to target vehicle, it is right to realizeThe horizontal lane change trend of target vehicle timely and effectively predicts, has effectively evaded what target vehicle lane change brought to Current vehicleIt is dangerous.
Second embodiment
The second embodiment of the present invention provides a kind of judgment means of vehicle lane change trend, including processor and storageDevice, be stored with instruction in the memory, when the reason device performs those instructions, just performs described in first embodiment of the inventionVehicle lane change trend determination methods, for simplicity, the determination methods of performed vehicle lane change trend are instructed here no longerRepeated.
Exemplarily, the judgment means of vehicle lane change trend according to a second embodiment of the present invention also include：Road imageAcquiring unit, it obtains the video image of road image or road, and Fig. 5 shows that vehicle according to a second embodiment of the present invention becomesThe block diagram of the judgment means of road trend.
Exemplarily, the road image acquiring unit can be mounted in the camera of Current vehicle, and camera can setPut in the front portion of vehicle, as shown in Fig. 2 camera 101 is arranged on the front portion of vehicle.Alternatively, camera can also be arranged atThe middle part or rear portion of vehicle, as long as road image or video image can be obtained just can be with.
Exemplarily, in the judgment means of vehicle lane change trend according to embodiments of the present invention, processor, memory withAnd road image acquiring unit can be arranged in Current vehicle.Wherein processor and memory obtain single with road imageFirst signal connection, after processor judges whether vehicle has lane change trend, just will determine that result is supplied to the control of vehicleDevice.
It is alternatively possible to be that road image acquiring unit is arranged in Current vehicle, and processor and memory are arranged onIt is connected in remote server and with road image acquiring unit signal, it is, processor obtains road image acquiring unit obtainingThe road image for taking or video image, and process and stored in memory with result of calculation, the processor has obtained vehicle lane changeThe control device of vehicle is input to after the judged result of trend, so that Current vehicle carries out corresponding behavior, for example, is slowed downOr avoid etc..
Alternatively, the judged result of vehicle lane change trend can also be supplied to remote vehicle controller by processor so thatThe controller can cause it to make corresponding behavior with remote control Current vehicle.
It should be noted that in an embodiment of the present invention, Current vehicle refers to need the lane change for knowing other vehiclesThe vehicle of trend, and target vehicle refers to the vehicle for needing to judge its lane change trend.
3rd embodiment
The third embodiment of the present invention provides a kind of computer-readable storage medium, is stored thereon with the executable finger of computerOrder, when the instruction is executed by a computing apparatus, performs sentencing for the vehicle lane change trend described in first embodiment of the inventionDisconnected method, for simplicity, the determination methods that performed vehicle lane change trend is instructed here are no longer repeated.
Embodiments of the invention provide a kind of determination methods of vehicle lane change trend, judgment means and Computer Storage and are situated betweenMatter, the determination methods of the vehicle lane change trend include：A frame road image is chosen as present frame road image；Recognize and selectThe identification point of target vehicle；Identification lane line；The identification point is calculated to the vertical range of the lane line and is stored；According to instituteThe vertical range for stating the least one frame of road image before present frame road image and the present frame road image is sentencedWhether the target vehicle that breaks has lane change trend, if lane change trend, then lane change trend result is exported, if become without lane changeGesture can not judge lane change trend, then carry out repeating above step.Vehicle lane change trend provided in an embodiment of the present inventionDetermination methods, by being identified to target vehicle identification point, and calculate the identification point of target vehicle to the vertical of lane lineJudging the lane change trend of target vehicle, the intact form without recognizing target vehicle reduces and even eliminates distanceThe judgement of vehicle lane change trend postpones, and avoids the weaker detection accuracy of overall vehicle visual identity and visual identity and resolutionPower, embodiments of the invention can effectively lift vision sensor in the degree of accuracy for judging target vehicle transverse shifting and in timeDegree, so as to improve automatic driving vehicle performance when target vehicle is cut transversely into this track scene is being tackled, and is effectively evadedThe danger that target vehicle lane change brings to Current vehicle.A kind of determination methods of vehicle lane change trend, including：
(1) a kind of determination methods of vehicle lane change trend, including：
S1, chooses a frame road image as present frame road image；
The identification point of S2, identification and selection target vehicle；
S3, recognizes lane line；
S4, calculates identification point to the vertical range of nearest lane line and stores；
S5, according to present frame road image before least one frame of road image and present frame road image it is vertical away fromFrom judging whether target vehicle has lane change trend, if lane change trend, then carry out step S6, if without lane change trend orLane change trend can not be judged, then carry out step S1；
S6, exports lane change trend result.
(2) determination methods of the vehicle lane change trend of basis (1), road image and the present frame road of the frame of wherein at least oneImage is the road image of successive frame.
(3) determination methods of the vehicle lane change trend of basis (1), wherein the identification point bag of identification and selection target vehicleInclude：Identification and selection target vehicle hub centre, identification and selection target vehicle covering moulding point, recognize and select meshAngle summit or pass through that the edge on angle summit, identification and selection target vehicle that mark vehicle is formed with ground area shading is formedConvolutional neural networks identification and the characteristic point of selection target vehicle.
(4) determination methods of the vehicle lane change trend of basis (3), wherein identification and selection target vehicle and ground area shading shapeInto angle summit include：
Calculate and obtain the length of the edge line of target vehicle and the length of ground area shading；
Calculate and obtain the angle that target vehicle is formed with ground area shading；
Length according to edge line, area shading length and angle target vehicle is recognized and chosen from road imageWith the angle summit of ground area shading.
(5) determination methods of the vehicle lane change trend of basis (3), wherein the hub centre bag of identification and selection target vehicleInclude：
Circular or circular arc pattern in identification road image；
It is determined that hub centre of the center of circular or circular arc pattern as target vehicle.
(6) determination methods of the vehicle lane change trend of basis (5), wherein the hub centre of identification and selection target vehicle existsAfter circular or circular arc pattern in identification road image and it is determined that the center of circular or circular arc pattern is used as target carriageHub centre before also include：
The region for selecting circular or circular arc pattern to be located；
In circle or circular arc pattern region being screened and being rejected to the region that circular or circular arc pattern is locatedFringe region.
(7) determination methods of the vehicle lane change trend of basis (1), are choosing a frame road image as present frame mileage chartAlso include as after and in identification and before the identification point of selection target vehicle：
Image procossing is carried out to present frame road image.
(8) determination methods of the vehicle lane change trend of basis (7), wherein at least before according to present frame road imageThe road image of frame and the vertical range of present frame road image judge whether target vehicle has lane change trend to include：
The identification point of the road image of two continuous frames before current frame image and current frame image is chosen to nearest trackVertical range D of linei、Di-1And Di-2, wherein DiFor the vertical range of current frame image, Di-1Before for current frame image firstThe vertical range of frame road image, Di-2The vertical range of the second frame road image before for current frame image,；
Any two difference value A and B are calculated and obtained according to the vertical range of continuous three frame；
Judge whether target vehicle has lane change trend according to lane change Trend rules,
Wherein lane change Trend rules be included in two difference values symbol it is identical and two difference values are all higher than first thresholdOr two difference values symbol is identical and the absolute value of one of two difference values is more than another difference value and Second ThresholdWith in the case of, judge that target vehicle has lane change trend
First threshold and Second Threshold are lateral velocity threshold.
(9) determination methods of the vehicle lane change trend of basis (8), wherein calculating and obtaining according to the vertical range of continuous three frameTaking any two difference value A and B includes：
Calculate and obtain the difference △ D1=D of the vertical range of the frame of arbitrary neighborhood twoi-1-Di-2With △ D2=Di-Di-1；
According to formula：A=△ D1/T and B=△ D2/T calculate and obtain two difference values,
Wherein T is to choose a frame road image as time of present frame road image, the time of image procossing, identification simultaneouslySelect the time sum of time, the time of identification lane line and the calculating identification point of identification point to the vertical range of lane line.
(10) determination methods of the vehicle lane change trend of basis (1), wherein calculating identification point hanging down to nearest lane lineStraight distance includes：
The corresponding pixel of identification point is calculated to the vertical range of nearest lane line.
(11) determination methods of the vehicle lane change trend of basis (1), wherein calculating identification point hanging down to nearest lane lineStraight distance includes：
Road image is converted to into top plan view；
Vertical range of the identification point to nearest lane line in Calculation Plane top view.
(12) determination methods of the vehicle lane change trend of basis (1), are choosing a frame road image as present frame roadAlso include before image：
Road image is obtained using image acquiring device.
(13) determination methods of the vehicle lane change trend of basis (1), are choosing a frame road image as present frame roadAlso include before image：
The video image of road is obtained using image acquiring device.
(14) determination methods of the vehicle lane change trend of basis (7), wherein carrying out at image to present frame road imageAlso include after reason and in identification and before the identification point of selection target vehicle：
The region that selection target vehicle is located.
(15) determination methods of the vehicle lane change trend of basis (1), wherein target vehicle include being located at Current vehicle frontAnd at least 6 vehicles sorted from small to large according to the distance value along lane line direction away from Current vehicle.
(16) a kind of judgment means of vehicle lane change trend, including processor and memory, be stored with instruction in memory,When computing device is instructed, the determination methods of any one in (1)-(15) are performed.
(17) according to the judgment means of (16), also include：
Road image acquiring unit, obtains the video image of road image or road.
(18) according to the judgment means of (17), wherein processor, memory and road image acquiring unit are arranged on and work asIn vehicle in front.
(19) according to the judgment means of (17), wherein road image acquiring unit is arranged in Current vehicle, processor andMemory is arranged in remote server and is connected with road image acquiring unit signal.
(20) a kind of computer-readable storage medium, is stored thereon with computer executable instructions, when instruction is held by computing deviceDuring row, the determination methods of any one in (1)-(15) are performed.
The above, the only specific embodiment of the present invention, but the protection domain of the embodiment of the present invention is not limited toThis, any those familiar with the art can readily occur in change in the technical scope that the embodiment of the present invention is disclosedOr replace, all should cover within the protection domain of the embodiment of the present invention.