CN105825185A - Early warning method and device against collision of vehicles - Google Patents
Early warning method and device against collision of vehicles Download PDFInfo
- Publication number
- CN105825185A CN105825185A CN201610148003.9A CN201610148003A CN105825185A CN 105825185 A CN105825185 A CN 105825185A CN 201610148003 A CN201610148003 A CN 201610148003A CN 105825185 A CN105825185 A CN 105825185A
- Authority
- CN
- China
- Prior art keywords
- front truck
- car
- marking area
- video image
- picture frame
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/16—Anti-collision systems
- G08G1/166—Anti-collision systems for active traffic, e.g. moving vehicles, pedestrians, bikes
Abstract
The invention discloses an early warning method against collision of vehicles. The method comprises that a frame of a road-ahead video image of the vehicle are collected; the image frame is used to determine whether there is a close-range vehicle, satisfying a predefined work environment, is ahead of the vehicle in the road; when it is determined that the close-range vehicle ahead exists, a trail feature substantial area of the vehicle ahead is determined in the image frame; the substantial area is tracked, and the position and/or the size of the substantial area in video image frames are/is recorded; and the vehicle distance state corresponding to change of the recorded position and/or size of the substantial area in the video image frames is determined, and early-warning against collision of the vehicles is emitted according to the corresponding vehicle distance state. The invention also discloses an early warning device against collision of the vehicles. The early warning method and device can be used to reduce the hardware cost, and improve the application range and accuracy of early warning against collision.
Description
Technical field
The present invention relates to target identification technology field, particularly relate to vehicle collision avoidance method for early warning and device.
Background technology
At present, prior art generally utilize radar equipment detect front vehicles, and by the depth information of environment before acquisition vehicle for the potential trend judging that this car and front vehicles collide, but owing to detection angles when radar equipment utilizes electromagnetic wave to judge preceding object thing is less, thus can there is the problem at detection dead angle, thus cause Detection results not ideal enough and simultaneously radar equipment relatively costly.
Summary of the invention
Present invention is primarily targeted at a kind of vehicle collision avoidance method for early warning of offer and device, it is intended to solve that the hardware cost employed in existing vehicle collision avoidance technology is higher and the dissatisfactory technical problem of Detection results.
For achieving the above object, the present invention provides a kind of vehicle collision avoidance method for early warning, and described vehicle collision avoidance method for early warning includes:
Gather the picture frame of this car road ahead video image；
Determine whether this car road ahead exists predefined closely front truck by described picture frame；
When determine there is closely front truck time, from described picture frame, determine described front truck tail feature marking area；
Follow the tracks of described marking area, and record described marking area position in the picture frame of described video image and/or size；
Judge described marking area position in the picture frame of described video image and/or the spacing state corresponding to change in size recorded, and send vehicle collision avoidance early warning according to corresponding spacing state.
Preferably, using machine learning algorithm to carry out off-line training grader, wherein, the off-line training sample of described grader at least includes the closely front truck sample of this car road ahead, remote front truck sample and without car sample；Described determine whether this car road ahead exists predefined closely front truck and include by described picture frame:
The described grader obtained according to training in advance, analyzes the characteristics of image of described picture frame, obtains the classification results corresponding to described picture frame；
Carrying out comprehensive descision according to the described classification results obtained with the vehicle characteristics preset, to determine whether this car road ahead exists closely front truck, described vehicle characteristics at least includes the correlated characteristic of car plate and/or car light and/or bumper.
Preferably, described when determine there is closely front truck time, from described picture frame, determine that described front truck tail feature marking area includes:
When determine there is closely front truck time, from described picture frame, extract some image outlines of described front truck tail feature, and generate the external bounding box of some described image outlines；
According to car plate and/or car light and/or the correlated characteristic of bumper, the some specific external bounding box of the correlated characteristic meeting car plate and/or car light and/or bumper is filtered out from some described external bounding boxes, and determine that described specific external bounding box region corresponds to described front truck tail feature marking area, wherein, described correlated characteristic at least includes that car plate, car light, bumper are distinguished one or more in the symmetry characteristic feature of corresponding shape facility, size characteristic, color characteristic and car light.
Preferably, the described marking area that described judgement is recorded position in the picture frame of described video image and/or the spacing state corresponding to change in size, and send vehicle collision avoidance early warning according to corresponding spacing state and include:
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or size constancy, it is determined that Ben Che and the spacing of described front truck correspond to keep constant；
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or change in size, if the position that described marking area is in the current image frame of described video image moves up relative to the position in previous image frame and/or described marking area size in the current image frame of described video image diminishes relative to the size in previous image frame, then judge that this car corresponds to increase with the spacing of described front truck；
If the position that described marking area is in the current image frame of described video image moves down relative to the position in previous image frame and/or described marking area size in the current image frame of described video image is relative to becoming large-sized in previous image frame, then judge that this car corresponds to reduce and send vehicle collision avoidance early warning with the spacing of described front truck.
Further, for achieving the above object, the present invention also provides for a kind of vehicle collision avoidance prior-warning device, and described vehicle collision avoidance prior-warning device includes:
Image capture module, for gathering the picture frame of this car road ahead video image；
By described picture frame, closely front truck judge module, for determining whether this car road ahead exists predefined closely front truck；
Marking area determines module, for when determine there is closely front truck time, from described picture frame, determine described front truck tail feature marking area；
Tracking module, is used for following the tracks of described marking area, and records described marking area position in the picture frame of described video image and/or size；
Early-warning judgment module, for judging the described marking area position in the picture frame of described video image and/or the spacing state corresponding to change in size that are recorded, and sends vehicle collision avoidance early warning according to corresponding spacing state.
Preferably, described closely front truck judge module uses machine learning algorithm to carry out off-line training grader, and wherein, the off-line training sample of described grader at least includes the closely front truck sample of this car road ahead, remote front truck sample and without car sample；
Described closely front truck judge module includes:
Taxon, for the grader obtained according to training in advance, analyzes the characteristics of image of described picture frame, obtains the classification results corresponding to described picture frame；
Comprehensive descision unit, for carrying out comprehensive descision according to the described classification results obtained with the vehicle characteristics preset, to determine whether this car road ahead exists closely front truck, described vehicle characteristics at least includes the correlated characteristic of car plate and/or car light and/or bumper.
Preferably, described marking area determines that module includes:
Contours extract unit, for when determine there is closely front truck time, from described picture frame, extract some image outlines of described front truck tail feature, and generate the external bounding box of some described image outlines；
Marking area determines unit, for the correlated characteristic according to car plate and/or car light and/or bumper, the some specific external bounding box of the correlated characteristic meeting car plate and/or car light and/or bumper is filtered out from some described external bounding boxes, and determine that described specific external bounding box region corresponds to described front truck tail feature marking area, wherein, described correlated characteristic at least includes that car plate, car light, bumper are distinguished one or more in the symmetry characteristic feature of corresponding shape facility, size characteristic, color characteristic and car light.
Preferably, described early-warning judgment module specifically for:
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or size constancy, it is determined that Ben Che and the spacing of described front truck correspond to keep constant；
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or change in size, if the position that described marking area is in the current image frame of described video image moves up relative to the position in previous image frame and/or described marking area size in the current image frame of described video image diminishes relative to the size in previous image frame, then judge that this car corresponds to increase with the spacing of described front truck；
If the position that described marking area is in the current image frame of described video image moves down relative to the position in previous image frame and/or described marking area size in the current image frame of described video image is relative to becoming large-sized in previous image frame, then judge that this car corresponds to reduce and send vehicle collision avoidance early warning with the spacing of described front truck.
In the present invention, only need to use picture pick-up device (such as photographic head) that vehicle collision avoidance early warning process can be carried out, thus save hardware cost.The present invention is especially by gathering this car road ahead video image, and then by analysis video image to determine whether this car road ahead exists closely front truck and front truck tail feature marking area, and by following the tracks of front truck tail feature marking area and judging front truck tail feature marking area position in the picture frame of video image and/or the spacing state corresponding to change in size, send vehicle collision avoidance early warning under corresponding spacing state.The present invention is by obtaining this car road ahead video image, thus can obtain the detection angles that this front side is bigger, promotes the effect of detection.Additionally, the present invention is by following the tracks of front truck tail feature marking area to determine the spacing state between this car and front truck, thus expand the multiple applicable situation of vehicle collision avoidance early warning, such as parking anti-slip, front truck lane change etc..
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of vehicle collision avoidance method for early warning one embodiment of the present invention；
Fig. 2 is closely front truck one embodiment schematic diagram in vehicle collision avoidance method for early warning of the present invention；
Fig. 3 is the refinement schematic flow sheet of step S20 in Fig. 1；
Fig. 4 is the refinement schematic flow sheet of step S30 mono-embodiment in Fig. 1；
Fig. 5 is the schematic diagram of the first embodiment that marking area changes in video image in vehicle collision avoidance method for early warning of the present invention；
Fig. 6 is the schematic diagram of the second embodiment that marking area changes in video image in vehicle collision avoidance method for early warning of the present invention；
Fig. 7 is the schematic diagram of the 3rd embodiment that marking area changes in video image in vehicle collision avoidance method for early warning of the present invention；
Fig. 8 is the high-level schematic functional block diagram of vehicle collision avoidance prior-warning device one embodiment of the present invention；
Fig. 9 is the refinement high-level schematic functional block diagram of closely front truck judge module in Fig. 8；
Figure 10 is the refinement high-level schematic functional block diagram that in Fig. 8, marking area determines module.
The realization of the object of the invention, functional characteristics and advantage will in conjunction with the embodiments, are described further referring to the drawings.
Detailed description of the invention
Should be appreciated that specific embodiment described herein, only in order to explain the present invention, is not intended to limit the present invention.
With reference to the schematic flow sheet that Fig. 1, Fig. 1 are vehicle collision avoidance method for early warning one embodiment of the present invention.In the present embodiment, described vehicle collision avoidance method for early warning includes:
Step S10, gathers the picture frame of this car road ahead video image；
In the present embodiment, shooting this car road ahead environment especially by picture pick-up device (such as photographic head) and generate corresponding video image, this video image is specially RGB color image.It should be noted that, picture pick-up device shooting both can be to shoot in this car driving process, can also be to shoot when vehicle stops (under state of parking) simultaneously, the shooting angle scope of this car road ahead environment captured by picture pick-up device does not limits simultaneously, can be specifically the traveling lane for this car place or for a plurality of track on same bar road.
In the present embodiment, picture pick-up device is not limited, the mobile phone of such as integrated camera or panel computer, or independent photographic head is connected with vehicle-mounted computer.Riding position simultaneously for picture pick-up device does not limits, and such as, mobile phone or panel computer etc. are fixed on this car the optional position that can photograph this track road forward image by support.Additionally, for being applicable to all kinds of special environment, such as rain, dense fog, night etc., special camera can be selected to be lifted at the image-capable under above-mentioned special environment.
In the present embodiment, by gathering the picture frame of this car road ahead video image, and then it is easy to analyze this front side vehicle condition or analyze this car vehicle condition according to front vehicle condition.Such as, analyze this front side and whether there is in-plant front truck (within such as 3-5 rice), or analyze this car and whether there occurs and slip car, and then according to analysis result to carry out anticollision early warning or prompting.
By described picture frame, step S20, determines whether this car road ahead exists predefined closely front truck；
In the present embodiment, closely front truck is defined to the potential trend correlation that whether can collide with specific reference to front truck with this car, wherein, in-plant definition is the most relevant to local environment during vehicle collision avoidance, the inventive method is preferably applied to low speed, in-plant environment, such as during road congestion at a slow speed with car, wait traffic lights or by the environment such as charge station, therefore, closely may be defined as the spacing within 3-5 rice, and the definition of front truck both can be with the front truck in track with this car, it is also possible to be and the front truck on the adjacent lane in track, this car place.Such as, as in figure 2 it is shown, wherein A represents this car, B1-B5 can represent front truck, then B1-B3 can be defined as closely front truck.
In the present embodiment, for determining whether this car road ahead exists the processing mode of closely front truck and do not limit by picture frame, such as, texture, gradient and marginal informations etc. according to vehicle train grader to identify front vehicles, and by detect this car kinestate and and front vehicles between relative distance and the information such as relative velocity to determine the closely front truck that there is this car.
Step S30, when determine there is closely front truck time, from described picture frame, determine described front truck tail feature marking area；
In the present embodiment, front truck tail feature marking area specifically includes car plate, car light, bumper etc..Identifying that from picture frame the mode of above-mentioned marking area is a lot, such as edge detection algorithm, contour detecting algorithm etc., therefore do not do in the present embodiment and too much repeat.
Step S40, follows the tracks of described marking area, and records described marking area position in the picture frame of described video image and/or size；
In the present embodiment, do not limit for following the tracks of the processing mode of marking area, such as TLD algorithm, STC algorithm, KCF algorithm etc., the most do not do and too much repeat.It should be noted that, when there is many closely front trucks in this front side, potential risk of collision is caused for preventing the unexpected lane change of closely front truck in other tracks, therefore, in the present embodiment, both can be only to follow the tracks of the closely front truck (monotrack) on the same track of this car, it is also possible to be to follow the tracks of the closely front truck (multiple target tracking) on a plurality of track, be configured with specific reference to being actually needed.
Step S50, it is judged that the described marking area recorded position in the picture frame of described video image and/or the spacing state corresponding to change in size, and send vehicle collision avoidance early warning according to corresponding spacing state.
In the present embodiment, based on parallel perspective principle, when such as Ben Che and front truck have relative movement, follow the tracks of target have in video image positions and dimensions change (position near farsighted low, size is near big and far smaller), thus can determine whether that Ben Che is closer or far from state with the spacing state of front truck, if and near state, then needing to issue the user with vehicle collision avoidance early warning.
The present embodiment only need to use picture pick-up device (such as photographic head) can carry out vehicle collision avoidance early warning process, thus save hardware cost.By gathering this car road ahead video image, and then by analysis video image to determine whether this car road ahead exists closely front truck and front truck tail feature marking area, and by following the tracks of front truck tail feature marking area and judging front truck tail feature marking area position in the picture frame of video image and/or the spacing state corresponding to change in size, send vehicle collision avoidance early warning under corresponding spacing state.In the present embodiment, by obtaining this car road ahead video image, thus the detection angles that this front side is bigger can be obtained, promote the effect of detection.Additionally, in the present embodiment, by tracking front truck tail feature marking area to determine the spacing state between this car and front truck, thus expand the multiple applicable situation of vehicle collision avoidance early warning, such as parking anti-slip, front truck lane change etc..
With reference to the refinement schematic flow sheet that Fig. 3, Fig. 3 are step S20 in Fig. 1.Based on above-described embodiment, in the present embodiment, for ease of identifying the closely front truck in picture frame, therefore, the common classification algorithm preferably employed in the present embodiment in machine learning algorithm carries out off-line training grader, specifically comprises the following steps that
Step one: gather the closely front truck sample of this car road ahead, remote front truck sample and without multiclass samples such as car samples as the off-line training sample of grader；Wherein, the training sample of grader is configured with specific reference to practical situation, is such as arranged corresponding training sample etc. by user according to actual driving experience.
Step 2: the feature such as HOG, haar or LBP of extracting each sample.
Step 3: by sorting algorithm conventional in machine learning, features described above is trained, obtains may determine that the grader of closely vehicle, and then this grader obtained by training can carry out classification process to picture frame.Wherein, grader specifically refers to the general designation of the method in machine learning classified sample, and such as conventional sorting algorithm has neural network algorithm, support vector machine etc..
In the present embodiment, above-mentioned steps S20 specifically includes:
Step S201, the described grader obtained according to training in advance, analyze the characteristics of image of described picture frame, obtain the classification results corresponding to described picture frame；
Step S202, carries out comprehensive descision according to the described classification results obtained with the vehicle characteristics preset, and to determine whether this car road ahead exists closely front truck, described vehicle characteristics at least includes the correlated characteristic of car plate and/or car light and/or bumper.
In the present embodiment, although the vehicle condition of this car road ahead can be determined more accurately by grader, but, for avoiding the unexpected factor in actual environment or the impact on judged result of other outside environmental elements, therefore, comprehensive descision is carried out further according to the vehicle characteristics preset, such as car plate color, car light color, car plate shape and length-width ratio, bumper shape etc., thus realize the accurate differentiation of the vehicle condition to this car road ahead, and then precisely determine whether this car road ahead exists closely front truck, and improve the safety of the inventive method accordingly.
With reference to the refinement schematic flow sheet that Fig. 4, Fig. 4 are step S30 in Fig. 1.Based on above-described embodiment, in the present embodiment, above-mentioned steps S30 includes:
Step S301, when determine there is closely front truck time, from described picture frame, extract some image outlines of described front truck tail feature, and generate the external bounding box of some described image outlines；
Step S302, according to car plate and/or car light and/or the correlated characteristic of bumper, the some specific external bounding box of the correlated characteristic meeting car plate and/or car light and/or bumper is filtered out from some described external bounding boxes, and determine that described specific external bounding box region corresponds to described front truck tail feature marking area, wherein, described correlated characteristic at least includes that car plate, car light, bumper are distinguished one or more in the symmetry characteristic feature of corresponding shape facility, size characteristic, color characteristic and car light.
In the present embodiment, when arranging too high with front truck at a distance of the position crossing near and picture pick-up device in view of Ben Che, this car headstock may shelter from the car plate part of front truck and cause extracting the license plate outline of front truck, therefore, it is further contemplated that extract front truck vehicle lamp area (vehicle lamp area to arrange the general relatively car plate in position higher, car light has the features such as redness, yellow, white and symmetry simultaneously) as front truck tail feature marking area, concrete processing procedure and the processing procedure basic simlarity of car plate, the most do not do and too much repeat.
Further alternative, in vehicle collision avoidance method for early warning one embodiment of the present invention, above-mentioned steps S50 specifically includes following a few class:
(1) when described marking area is when the current image frame of described video image is with the position in previous image frame and/or size constancy, it is determined that Ben Che and the spacing of described front truck correspond to keep constant；
In the present embodiment, marking area change in location in current image frame with previous image frame specifically refers to marking area change in location on the longitudinally perpendicular direction of video image, as shown in Figure 5.
(2) when described marking area is when the current image frame of described video image is with the position in previous image frame and/or change in size, if the position that described marking area is in the current image frame of described video image moves up relative to the position in previous image frame and/or described marking area size in the current image frame of described video image diminishes relative to the size in previous image frame, then judge that this car corresponds to increase with the spacing of described front truck；
As shown in Figure 6, Ben Che corresponds to increase with the spacing of front truck, namely now spacing state corresponds to be relatively distant from state.
(3) if described marking area position in the current image frame of described video image moves down relative to the position in previous image frame and/or described marking area size in the current image frame of described video image is relative to becoming large-sized in previous image frame, then judge that this car corresponds to reduce and send vehicle collision avoidance early warning with the spacing of described front truck.
As it is shown in fig. 7, the spacing of Ben Che and front truck corresponds to reduce, namely now spacing state corresponds to relatively close state, therefore, under this spacing state, needs to issue the user with vehicle collision avoidance early warning.
In the present embodiment, the mode for early warning does not limits, and can be alarm song, it is also possible to be corresponding prompting sound etc..Furthermore, it is necessary to explanation, the present embodiment is applicable not only to the anticollision early warning of vehicle running state, can equally be well applied to the anticollision early warning of vehicle parking state (slip forward or backward car).It addition, in view of in reality Driving Scene, due to this car overtake other vehicles front truck, front truck lane change or the situation such as turn around and when causing following the tracks of failed, by next picture frame of Resurvey and repeat the process step in above-described embodiment and process to proceed early warning.
With reference to the high-level schematic functional block diagram that Fig. 8, Fig. 8 are vehicle collision avoidance prior-warning device one embodiment of the present invention.In the present embodiment, described vehicle collision avoidance prior-warning device includes:
Image capture module 10, for gathering the picture frame of this car road ahead video image；
In the present embodiment, image capture module 10 is by gathering the picture frame of this car road ahead video image, and then is easy to analyze this front side vehicle condition or analyze this car vehicle condition according to front vehicle condition.Such as, analyze this front side and whether there is in-plant front truck (within such as 3-5 rice), or analyze this car and whether there occurs and slip car, and then according to analysis result to carry out anticollision early warning or prompting.
By described picture frame, closely front truck judge module 20, for determining whether this car road ahead exists predefined closely front truck；
In the present embodiment, closely front truck is defined to the potential trend correlation that whether can collide with specific reference to front truck with this car, wherein, in-plant definition is the most relevant to local environment during vehicle collision avoidance, the inventive method is preferably applied to low speed, in-plant environment, such as during road congestion at a slow speed with car, wait traffic lights or by the environment such as charge station, therefore, closely may be defined as the spacing within 3-5 rice, and the definition of front truck both can be with the front truck in track with this car, it is also possible to be and the front truck on the adjacent lane in track, this car place.Such as, as in figure 2 it is shown, wherein A represents this car, B1-B5 can represent front truck, then B1-B3 can be defined as closely front truck.
In the present embodiment, for determining whether this car road ahead exists the processing mode of closely front truck and do not limit by picture frame, such as, texture, gradient and marginal informations etc. according to vehicle train grader to identify front vehicles, and by detect this car kinestate and and front vehicles between relative distance and the information such as relative velocity to determine the closely front truck that there is this car.
Marking area determines module 30, for when determine there is closely front truck time, from described picture frame, determine described front truck tail feature marking area；
In the present embodiment, front truck tail feature marking area specifically includes car plate, car light, bumper etc..Identifying that from picture frame the mode of above-mentioned marking area is a lot, such as edge detection algorithm, contour detecting algorithm etc., therefore do not do in the present embodiment and too much repeat.
Tracking module 40, is used for following the tracks of described marking area, and records described marking area position in the picture frame of described video image and/or size；
In the present embodiment, do not limit for following the tracks of the processing mode of marking area, such as TLD algorithm, STC algorithm, KCF algorithm etc., the most do not do and too much repeat.It should be noted that, when there is many closely front trucks in this front side, potential risk of collision is caused for preventing the unexpected lane change of closely front truck in other tracks, therefore, in the present embodiment, both can be only to follow the tracks of the closely front truck (monotrack) on the same track of this car, it is also possible to be to follow the tracks of the closely front truck (multiple target tracking) on a plurality of track, be configured with specific reference to being actually needed.
Early-warning judgment module 50, for judging the described marking area position in the picture frame of described video image and/or the spacing state corresponding to change in size that are recorded, and sends vehicle collision avoidance early warning according to corresponding spacing state.
In the present embodiment, based on parallel perspective principle, when such as Ben Che and front truck have relative movement, follow the tracks of target have in video image positions and dimensions change (position near farsighted low, size is near big and far smaller), thus can determine whether that Ben Che is closer or far from state with the spacing state of front truck, if and near state, then needing to issue the user with vehicle collision avoidance early warning.
The present embodiment only need to use picture pick-up device (such as photographic head) can carry out vehicle collision avoidance early warning process, thus save hardware cost.By gathering this car road ahead video image, and then by analysis video image to determine whether this car road ahead exists closely front truck and front truck tail feature marking area, and by following the tracks of front truck tail feature marking area and judging front truck tail feature marking area position in the picture frame of video image and/or the spacing state corresponding to change in size, send vehicle collision avoidance early warning under corresponding spacing state.In the present embodiment, by obtaining this car road ahead video image, thus the detection angles that this front side is bigger can be obtained, promote the effect of detection.Additionally, in the present embodiment, by tracking front truck tail feature marking area to determine the spacing state between this car and front truck, thus expand the multiple applicable situation of vehicle collision avoidance early warning, such as parking anti-slip, front truck lane change etc..
With reference to the refinement high-level schematic functional block diagram that Fig. 9, Fig. 9 are closely front truck judge module in Fig. 8.Based on above-described embodiment, in the present embodiment, described closely front truck judge module 20 uses machine learning algorithm to carry out off-line training grader, and wherein, the off-line training sample of described grader at least includes the closely front truck sample of this car road ahead, remote front truck sample and without car sample；
Described closely front truck judge module 20 includes:
Taxon 201, for the grader obtained according to training in advance, analyzes the characteristics of image of described picture frame, obtains the classification results corresponding to described picture frame；
Comprehensive descision unit 202, for carrying out comprehensive descision according to the described classification results obtained with the vehicle characteristics preset, to determine whether this car road ahead exists closely front truck, described vehicle characteristics at least includes the correlated characteristic of car plate and/or car light and/or bumper.
For promoting image processing speed, promote the precision of result simultaneously, the present embodiment specifically uses machine learning algorithm off-line training grader, taxon 201 are classified by picture frame by grader, the setting of the result of classification specifically corresponding training sample.Such as, A picture frame classification results is that this car road ahead exists closely front truck, and B picture frame classification results is that this car road ahead exists remote front truck, and C picture frame classification results is that this car road ahead is without car.
In the present embodiment, although the vehicle condition of this car road ahead can be determined more accurately by grader, but, for avoiding the unexpected factor in actual environment or the impact on judged result of other outside environmental elements, therefore, comprehensive descision is carried out according to default vehicle characteristics further by comprehensive descision unit 202, such as car plate color, car light color, car plate shape and length-width ratio, bumper shape etc., thus realize the accurate differentiation of the vehicle condition to this car road ahead, and then precisely determine whether this car road ahead exists closely front truck, and improve the safety of the inventive method accordingly.
It is the refinement high-level schematic functional block diagram that in Fig. 8, marking area determines module with reference to Figure 10, Figure 10.Based on above-described embodiment, in the present embodiment, described marking area determines that module 30 includes:
Contours extract unit 301, for when determine there is closely front truck time, from described picture frame, extract some image outlines of described front truck tail feature, and generate the external bounding box of some described image outlines；
Marking area determines unit 302, for the correlated characteristic according to car plate and/or car light and/or bumper, the some specific external bounding box of the correlated characteristic meeting car plate and/or car light and/or bumper is filtered out from some described external bounding boxes, and determine that described specific external bounding box region corresponds to described front truck tail feature marking area, wherein, described correlated characteristic at least includes that car plate, car light, bumper are distinguished one or more in the symmetry characteristic feature of corresponding shape facility, size characteristic, color characteristic and car light.
In the present embodiment, when arranging too high with front truck at a distance of the position crossing near and picture pick-up device in view of Ben Che, this car headstock may shelter from the car plate part of front truck and cause extracting the license plate outline of front truck, therefore, it is further contemplated that extract front truck vehicle lamp area (vehicle lamp area to arrange the general relatively car plate in position higher, car light has the features such as redness, yellow, white and symmetry simultaneously) as front truck tail feature marking area, concrete processing procedure and the processing procedure basic simlarity of car plate, the most do not do and too much repeat.
Further alternative, in vehicle collision avoidance prior-warning device one embodiment of the present invention, described early-warning judgment module 50 specifically for:
(1) when described marking area is when the current image frame of described video image is with the position in previous image frame and/or size constancy, it is determined that Ben Che and the spacing of described front truck correspond to keep constant；
In the present embodiment, marking area change in location in current image frame with previous image frame specifically refers to marking area change in location on the longitudinally perpendicular direction of video image, as shown in Figure 5.
(2) when described marking area is when the current image frame of described video image is with the position in previous image frame and/or change in size, if the position that described marking area is in the current image frame of described video image moves up relative to the position in previous image frame and/or described marking area size in the current image frame of described video image diminishes relative to the size in previous image frame, then judge that this car corresponds to increase with the spacing of described front truck；
As shown in Figure 6, Ben Che corresponds to increase with the spacing of front truck, namely now spacing state corresponds to be relatively distant from state.
(3) if described marking area position in the current image frame of described video image moves down relative to the position in previous image frame and/or described marking area size in the current image frame of described video image is relative to becoming large-sized in previous image frame, then judge that this car corresponds to reduce and send vehicle collision avoidance early warning with the spacing of described front truck.
As it is shown in fig. 7, the spacing of Ben Che and front truck corresponds to reduce, namely now spacing state corresponds to relatively close state, therefore, under this spacing state, needs to issue the user with vehicle collision avoidance early warning.
In the present embodiment, the mode for early warning does not limits, and can be alarm song, it is also possible to be corresponding prompting sound etc..Furthermore, it is necessary to explanation, the present embodiment is applicable not only to the anticollision early warning of vehicle running state, can equally be well applied to the anticollision early warning of vehicle parking state (slip forward or backward car).It addition, in view of in reality Driving Scene, due to this car overtake other vehicles front truck, front truck lane change or the situation such as turn around and when causing following the tracks of failed, by next picture frame of Resurvey and repeat the process step in above-described embodiment and process to proceed early warning.
These are only the preferred embodiments of the present invention; not thereby the scope of the claims of the present invention is limited; every equivalent structure utilizing description of the invention and accompanying drawing content to be made or equivalence flow process conversion; or directly or indirectly it is used in other relevant technical fields, the most in like manner it is included in the scope of patent protection of the present invention.
Claims (8)
1. a vehicle collision avoidance method for early warning, it is characterised in that described vehicle collision avoidance method for early warning includes:
Gather the picture frame of this car road ahead video image；
Determine whether this car road ahead exists predefined closely front truck by described picture frame；
When determine there is closely front truck time, from described picture frame, determine described front truck tail feature marking area；
Follow the tracks of described marking area, and record described marking area position in the picture frame of described video image and/or size；
Judge described marking area position in the picture frame of described video image and/or the spacing state corresponding to change in size recorded, and send vehicle collision avoidance early warning according to corresponding spacing state.
2. vehicle collision avoidance method for early warning as claimed in claim 1, it is characterized in that, using machine learning algorithm to carry out off-line training grader, wherein, the off-line training sample of described grader at least includes the closely front truck sample of this car road ahead, remote front truck sample and without car sample；
Described determine whether this car road ahead exists predefined closely front truck and include by described picture frame:
The described grader obtained according to training in advance, analyzes the characteristics of image of described picture frame, obtains the classification results corresponding to described picture frame；
Carrying out comprehensive descision according to the described classification results obtained with the vehicle characteristics preset, to determine whether this car road ahead exists closely front truck, described vehicle characteristics at least includes the correlated characteristic of car plate and/or car light and/or bumper.
3. vehicle collision avoidance method for early warning as claimed in claim 1 or 2, it is characterised in that described when determine there is closely front truck time, from described picture frame, determine that described front truck tail feature marking area includes:
When determine there is closely front truck time, from described picture frame, extract some image outlines of described front truck tail feature, and generate the external bounding box of some described image outlines；
According to car plate and/or car light and/or the correlated characteristic of bumper, the some specific external bounding box of the correlated characteristic meeting car plate and/or car light and/or bumper is filtered out from some described external bounding boxes, and determine that described specific external bounding box region corresponds to described front truck tail feature marking area, wherein, described correlated characteristic at least includes that car plate, car light, bumper are distinguished one or more in the symmetry characteristic feature of corresponding shape facility, size characteristic, color characteristic and car light.
4. vehicle collision avoidance method for early warning as claimed in claim 3, it is characterized in that, the described marking area that described judgement is recorded position in the picture frame of described video image and/or the spacing state corresponding to change in size, and send vehicle collision avoidance early warning according to corresponding spacing state and include:
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or size constancy, it is determined that Ben Che and the spacing of described front truck correspond to keep constant；
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or change in size, if the position that described marking area is in the current image frame of described video image moves up relative to the position in previous image frame and/or described marking area size in the current image frame of described video image diminishes relative to the size in previous image frame, then judge that this car corresponds to increase with the spacing of described front truck；
If the position that described marking area is in the current image frame of described video image moves down relative to the position in previous image frame and/or described marking area size in the current image frame of described video image is relative to becoming large-sized in previous image frame, then judge that this car corresponds to reduce and send vehicle collision avoidance early warning with the spacing of described front truck.
5. a vehicle collision avoidance prior-warning device, it is characterised in that described vehicle collision avoidance prior-warning device includes:
Image capture module, for gathering the picture frame of this car road ahead video image；
By described picture frame, closely front truck judge module, for determining whether this car road ahead exists predefined closely front truck；
Marking area determines module, for when determine there is closely front truck time, from described picture frame, determine described front truck tail feature marking area；
Tracking module, is used for following the tracks of described marking area, and records described marking area position in the picture frame of described video image and/or size；
Early-warning judgment module, for judging the described marking area position in the picture frame of described video image and/or the spacing state corresponding to change in size that are recorded, and sends vehicle collision avoidance early warning according to corresponding spacing state.
6. vehicle collision avoidance prior-warning device as claimed in claim 5, it is characterized in that, described closely front truck judge module uses machine learning algorithm to carry out off-line training grader, wherein, the off-line training sample of described grader at least includes the closely front truck sample of this car road ahead, remote front truck sample and without car sample；
Described closely front truck judge module includes:
Taxon, for the grader obtained according to training in advance, analyzes the characteristics of image of described picture frame, obtains the classification results corresponding to described picture frame；
Comprehensive descision unit, for carrying out comprehensive descision according to the described classification results obtained with the vehicle characteristics preset, to determine whether this car road ahead exists closely front truck, described vehicle characteristics at least includes the correlated characteristic of car plate and/or car light and/or bumper.
7. the vehicle collision avoidance prior-warning device as described in claim 5 or 6, it is characterised in that described marking area determines that module includes:
Contours extract unit, for when determine there is closely front truck time, from described picture frame, extract some image outlines of described front truck tail feature, and generate the external bounding box of some described image outlines；
Marking area determines unit, for the correlated characteristic according to car plate and/or car light and/or bumper, the some specific external bounding box of the correlated characteristic meeting car plate and/or car light and/or bumper is filtered out from some described external bounding boxes, and determine that described specific external bounding box region corresponds to described front truck tail feature marking area, wherein, described correlated characteristic at least includes that car plate, car light, bumper are distinguished one or more in the symmetry characteristic feature of corresponding shape facility, size characteristic, color characteristic and car light.
8. vehicle collision avoidance prior-warning device as claimed in claim 7, it is characterised in that described early-warning judgment module specifically for:
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or size constancy, it is determined that Ben Che and the spacing of described front truck correspond to keep constant；
When described marking area is when the current image frame of described video image is with the position in previous image frame and/or change in size, if the position that described marking area is in the current image frame of described video image moves up relative to the position in previous image frame and/or described marking area size in the current image frame of described video image diminishes relative to the size in previous image frame, then judge that this car corresponds to increase with the spacing of described front truck；
If the position that described marking area is in the current image frame of described video image moves down relative to the position in previous image frame and/or described marking area size in the current image frame of described video image is relative to becoming large-sized in previous image frame, then judge that this car corresponds to reduce and send vehicle collision avoidance early warning with the spacing of described front truck.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610148003.9A CN105825185B (en) | 2016-03-15 | 2016-03-15 | Vehicle collision avoidance method for early warning and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610148003.9A CN105825185B (en) | 2016-03-15 | 2016-03-15 | Vehicle collision avoidance method for early warning and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105825185A true CN105825185A (en) | 2016-08-03 |
| CN105825185B CN105825185B (en) | 2018-02-06 |
Family
ID=56987745
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610148003.9A Active CN105825185B (en) | 2016-03-15 | 2016-03-15 | Vehicle collision avoidance method for early warning and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105825185B (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106157697A (en) * | 2016-08-31 | 2016-11-23 | 东软集团股份有限公司 | A kind of braking method for early warning and device |
| CN106778480A (en) * | 2016-11-22 | 2017-05-31 | 武汉大学 | A kind of high accuracy based on car plate closely vehicle distance measurement method |
| CN106904121A (en) * | 2017-02-27 | 2017-06-30 | 北京汽车研究总院有限公司 | A kind of method and device that early warning is carried out to vehicle lane-changing behavior |
| CN106971627A (en) * | 2017-05-22 | 2017-07-21 | 佛山市神风航空科技有限公司 | A kind of avoidance prior-warning device and determination methods |
| CN108307066A (en) * | 2018-01-25 | 2018-07-20 | 广东欧珀移动通信有限公司 | Fall data reporting method and relevant apparatus |
| CN108417086A (en) * | 2017-02-09 | 2018-08-17 | 通用汽车环球科技运作有限责任公司 | The anticollision of personal mobile device |
| CN108515967A (en) * | 2018-04-25 | 2018-09-11 | 深圳卡安兴科技有限公司 | A kind of anti-collision early warning method and device based on Car license recognition |
| CN110569791A (en) * | 2019-09-02 | 2019-12-13 | 东北林业大学 | Expressway anti-collision early warning method |
| CN110816401A (en) * | 2019-11-07 | 2020-02-21 | 中国第一汽车股份有限公司 | Method and device for warning vehicle sliding of front vehicle, vehicle and storage medium |
| CN110930436A (en) * | 2019-11-27 | 2020-03-27 | 深圳市捷顺科技实业股份有限公司 | Target tracking method and device |
| CN111009166A (en) * | 2019-12-04 | 2020-04-14 | 上海市城市建设设计研究总院（集团）有限公司 | Road three-dimensional sight distance checking calculation method based on BIM and driving simulator |
| CN111178215A (en) * | 2019-12-23 | 2020-05-19 | 深圳成谷科技有限公司 | Sensor data fusion processing method and device |
| CN111243290A (en) * | 2020-04-28 | 2020-06-05 | 江苏货斯基网络科技有限公司 | Driving behavior data acquisition and analysis method and system |
| CN112349142A (en) * | 2020-10-29 | 2021-02-09 | 深圳市元征科技股份有限公司 | Vehicle rear-end collision early warning method and related equipment thereof |
| CN112365741A (en) * | 2020-10-23 | 2021-02-12 | 淮阴工学院 | Safety early warning method and system based on multilane vehicle distance detection |
| CN112885112A (en) * | 2021-04-29 | 2021-06-01 | 腾讯科技（深圳）有限公司 | Vehicle driving detection method, vehicle driving early warning method and device |
| CN111422190B (en) * | 2020-04-03 | 2021-08-31 | 北京四维智联科技有限公司 | Forward collision early warning method and system for rear car loader |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090143986A1 (en) * | 2004-04-08 | 2009-06-04 | Mobileye Technologies Ltd | Collision Warning System |
| CN102642510A (en) * | 2011-02-17 | 2012-08-22 | 汽车零部件研究及发展中心有限公司 | Image-based vehicle anti-collision early warning method |
| WO2014047465A2 (en) * | 2012-09-21 | 2014-03-27 | The Schepens Eye Research Institute, Inc. | Collision prediction |
| CN104268589A (en) * | 2014-09-16 | 2015-01-07 | 奇瑞汽车股份有限公司 | Front vehicle detection method |
-
2016
- 2016-03-15 CN CN201610148003.9A patent/CN105825185B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090143986A1 (en) * | 2004-04-08 | 2009-06-04 | Mobileye Technologies Ltd | Collision Warning System |
| CN102642510A (en) * | 2011-02-17 | 2012-08-22 | 汽车零部件研究及发展中心有限公司 | Image-based vehicle anti-collision early warning method |
| WO2014047465A2 (en) * | 2012-09-21 | 2014-03-27 | The Schepens Eye Research Institute, Inc. | Collision prediction |
| CN104268589A (en) * | 2014-09-16 | 2015-01-07 | 奇瑞汽车股份有限公司 | Front vehicle detection method |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106157697A (en) * | 2016-08-31 | 2016-11-23 | 东软集团股份有限公司 | A kind of braking method for early warning and device |
| CN106778480A (en) * | 2016-11-22 | 2017-05-31 | 武汉大学 | A kind of high accuracy based on car plate closely vehicle distance measurement method |
| CN108417086A (en) * | 2017-02-09 | 2018-08-17 | 通用汽车环球科技运作有限责任公司 | The anticollision of personal mobile device |
| CN106904121A (en) * | 2017-02-27 | 2017-06-30 | 北京汽车研究总院有限公司 | A kind of method and device that early warning is carried out to vehicle lane-changing behavior |
| CN106971627A (en) * | 2017-05-22 | 2017-07-21 | 佛山市神风航空科技有限公司 | A kind of avoidance prior-warning device and determination methods |
| CN108307066A (en) * | 2018-01-25 | 2018-07-20 | 广东欧珀移动通信有限公司 | Fall data reporting method and relevant apparatus |
| CN108515967A (en) * | 2018-04-25 | 2018-09-11 | 深圳卡安兴科技有限公司 | A kind of anti-collision early warning method and device based on Car license recognition |
| CN110569791A (en) * | 2019-09-02 | 2019-12-13 | 东北林业大学 | Expressway anti-collision early warning method |
| CN110816401A (en) * | 2019-11-07 | 2020-02-21 | 中国第一汽车股份有限公司 | Method and device for warning vehicle sliding of front vehicle, vehicle and storage medium |
| CN110930436B (en) * | 2019-11-27 | 2023-04-14 | 深圳市捷顺科技实业股份有限公司 | Target tracking method and device |
| CN110930436A (en) * | 2019-11-27 | 2020-03-27 | 深圳市捷顺科技实业股份有限公司 | Target tracking method and device |
| CN111009166A (en) * | 2019-12-04 | 2020-04-14 | 上海市城市建设设计研究总院（集团）有限公司 | Road three-dimensional sight distance checking calculation method based on BIM and driving simulator |
| CN111178215A (en) * | 2019-12-23 | 2020-05-19 | 深圳成谷科技有限公司 | Sensor data fusion processing method and device |
| CN111178215B (en) * | 2019-12-23 | 2024-03-08 | 深圳成谷科技有限公司 | Sensor data fusion processing method and device |
| CN111422190B (en) * | 2020-04-03 | 2021-08-31 | 北京四维智联科技有限公司 | Forward collision early warning method and system for rear car loader |
| CN111243290A (en) * | 2020-04-28 | 2020-06-05 | 江苏货斯基网络科技有限公司 | Driving behavior data acquisition and analysis method and system |
| CN112365741A (en) * | 2020-10-23 | 2021-02-12 | 淮阴工学院 | Safety early warning method and system based on multilane vehicle distance detection |
| CN112365741B (en) * | 2020-10-23 | 2021-09-28 | 淮阴工学院 | Safety early warning method and system based on multilane vehicle distance detection |
| CN112349142A (en) * | 2020-10-29 | 2021-02-09 | 深圳市元征科技股份有限公司 | Vehicle rear-end collision early warning method and related equipment thereof |
| CN112885112A (en) * | 2021-04-29 | 2021-06-01 | 腾讯科技（深圳）有限公司 | Vehicle driving detection method, vehicle driving early warning method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105825185B (en) | 2018-02-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105825185A (en) | Early warning method and device against collision of vehicles | |
| US10152638B2 (en) | Early detection of exit only and shared lanes using perception technology | |
| JP6773540B2 (en) | In-vehicle image processing device | |
| US7046822B1 (en) | Method of detecting objects within a wide range of a road vehicle | |
| CN102765365B (en) | Pedestrian detection method based on machine vision and pedestrian anti-collision warning system based on machine vision | |
| US9064418B2 (en) | Vehicle-mounted environment recognition apparatus and vehicle-mounted environment recognition system | |
| US8670592B2 (en) | Clear path detection using segmentation-based method | |
| CN106647776B (en) | Method and device for judging lane changing trend of vehicle and computer storage medium | |
| US10699567B2 (en) | Method of controlling a traffic surveillance system | |
| US20130286205A1 (en) | Approaching object detection device and method for detecting approaching objects | |
| JP4246766B2 (en) | Method and apparatus for locating and tracking an object from a vehicle | |
| CN109871728B (en) | Vehicle type recognition method and device | |
| EP2717219B1 (en) | Object detection device, object detection method, and object detection program | |
| CN104361350A (en) | Traffic sign identification system | |
| JP2003084064A (en) | Device and method for recognizing vehicle in front side | |
| CN103761877A (en) | Radar-triggering-based system and method for automatically recording red light running | |
| CN113793533A (en) | Collision early warning method and device based on vehicle front obstacle recognition | |
| JP2018055597A (en) | Vehicle type discrimination device and vehicle type discrimination method | |
| CN115147587A (en) | Obstacle detection method and device and electronic equipment | |
| US8971579B2 (en) | Windshield localization for occupancy detection | |
| Álvarez et al. | Perception advances in outdoor vehicle detection for automatic cruise control | |
| Beresnev et al. | Automated Driving System based on Roadway and Traffic Conditions Monitoring. | |
| WO2021250934A1 (en) | Image processing device and image processing method | |
| WO2022209373A1 (en) | Information processing device, information processing method, and information processing program | |
| Wang et al. | Research on Vehicle Speed Estimation Algorithm with Traffic Camera |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CP03 | Change of name, title or address | ||
| CP03 | Change of name, title or address |
Address after: 14 / F, Beidou building, 6 Huida Road, Jiangbei new district, Nanjing, Jiangsu Province 210000 Patentee after: Jiangsu Zhongtian Anchi Technology Co., Ltd Address before: 518000 Taoyuan Street Lishan Road, Nanshan District, Shenzhen City, Guangdong Province Patentee before: SHENZHEN ZHONGTIAN ANCHI Co.,Ltd. |