Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; 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
• • B—PERFORMING OPERATIONS; TRANSPORTING
  • B60—VEHICLES IN GENERAL
  • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
  • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
  • B60W40/02—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to ambient conditions
  • B60W40/06—Road conditions
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/50—Depth or shape recovery
  • G06T7/55—Depth or shape recovery from multiple images
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; 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]
  • G06V10/464—Salient features, e.g. scale invariant feature transforms [SIFT] using a plurality of salient features, e.g. bag-of-words [BoW] representations
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; 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
  • G06V10/60—Extraction of image or video features relating to illumination properties, e.g. using a reflectance or lighting model
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; 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
  • G06V20/588—Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road
• • B—PERFORMING OPERATIONS; TRANSPORTING
  • B60—VEHICLES IN GENERAL
  • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
  • B60W2552/00—Input parameters relating to infrastructure
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30248—Vehicle exterior or interior
  • G06T2207/30252—Vehicle exterior; Vicinity of vehicle
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
  • G06V10/00—Arrangements for image or video recognition or understanding
  • G06V10/20—Image preprocessing
  • G06V10/24—Aligning, centring, orientation detection or correction of the image
  • G06V10/247—Aligning, centring, orientation detection or correction of the image by affine transforms, e.g. correction due to perspective effects; Quadrilaterals, e.g. trapezoids

Definitions

• the invention relates to a method for detecting and evaluating reflections on a road. Furthermore, the invention relates to a device for carrying out the aforementioned method and a vehicle with such a device.
• Image capture allows the use of camera-based
• Driver assistance systems that are placed behind the windshield capture the visual perception of the driver corresponding to the apron of the vehicle.
• the functional scope of these systems extends from the high-beam automatic lighting system to the detection and display of speed limits to the warning of lane departure errors or imminent collision.
• the task of digital image processing as a stand-alone function or in fusion with radar or lidar sensors is primarily to recognize objects, to classify them and to follow them in the image section.
• Classic objects are usually a variety of vehicles such as cars, trucks, two-wheelers or pedestrians.
• cameras take over the capture of signs,
• the inertial sensors provide a good impression of the current driving condition of the vehicle and the entire driving situation. From this the criticality of driving situations can be deduced and the
• the design of the warning and intervention times is basically based on a dry roadway with high traction potential between the tire and the road.
• the object of the present invention can therefore be to provide a method and a device of the type mentioned at the outset with which the road condition or even the available friction coefficient of the roadway can be determined so that driver alerts and system interventions can be made more targeted and the effectiveness of accident avoidance driver assistance systems can be increased.
• the object is solved by the subject matters of the independent claims. Preferred embodiments are subject of the dependent claims.
• the method according to the invention as claimed in claim 1 serves to detect and evaluate reflections of at least one point on a roadway.
• a camera is provided, by means of which at least two digital images of the at least one roadway point are generated, wherein the generation of the images from different
• a road condition information is determined, in particular a road condition information which describes the friction coefficient of the roadway or makes a statement as to whether the roadway is dry, wet or icy.
• the invention makes use of the fact that reflections can generally be divided into three categories and each with a change in perspective or perspective
• Image processing with the aim to robustly detect road reflections for the detection of moisture and ice in particular.
• Roadway point which represents the roadway, draw a conclusion on a current road condition by using in the camera generated from two different perspectives images of the roadway point by using digital
• Image processing algorithms are searched for specific features that allow it to close on the current road condition.
• the method is preferably performed on a sufficiently lit scene that enables the production or recording of usable images.
• Prerequisite for the method is a change in perspective in a sequence of at least two images. In the case of diffuse reflections (indicator for a dry lane), the change of the viewing angle to a fixed point on the roadway has no visual effect, since the light is equally reflected in all directions. Changing the perspective changes that
• the method according to the invention is preferably used in a vehicle. Deploying the camera can do this especially inside the vehicle, preferably behind the windshield, so that a visual
• Perception of a driver of the vehicle according to the apron of the vehicle can be detected.
• the generation of the images from two different perspectives can be effected in particular by a driving movement of the vehicle.
• a digital camera is provided with which the at least two appearances directly digital
• a monocamera or a stereo camera can be used to generate the appearances, since depth information from the image can also be used for the algorithm, depending on the severity.
• depth information from the image can also be used for the algorithm, depending on the severity.
• Generating at least two digital images of the at least one roadway point by means of the camera takes place, wherein the generation of the images from different recording perspectives takes place with a stereo camera.
• a particular advantage of the method according to the invention is that specular reflections can be reliably distinguished from shadows (diffuse reflections), since they are a
• Method includes the additional method steps
• the Roadway condition information thus serves as an input for an accident-avoiding driver assistance system of a vehicle in order to be able to adapt warning and intervention times of the driver assistance system particularly effectively.
• ADAS Advanced Driver Assistance Systems
• Road condition information is used as important information of the driving environment in automating and is preferably supplied to a corresponding system control for autonomous driving. In this sense, it is provided according to a further advantageous embodiment that the
• Automated vehicle is included and driving strategy, as well as determination of delivery times between a
• a region comprising a plurality of lane points is used in the images of the camera representing the lane.
• the region can also be a segmented section.
• a region in the form of a trapezium is particularly preferred, wherein the trapezoidal region is transformed into a rectangular bird's-eye view by means of an estimated homography.
• the provision of the camera takes place in a vehicle.
• the first image is in a first position of the vehicle from a first
• Recording perspective generated The vehicle is moved to a second position, e.g. method, wherein the second position is different from the first position, i. the first and second positions do not match.
• This is followed by generating the at least second image in the at least second position of the vehicle from an at least second recording perspective.
• the at least two images of the at least two different recording perspectives are then transformed into a respective top view.
• This is followed by registering the at least two generated plan views with means of digital image processing including vehicle dynamics parameters of the vehicle and a comparison of the appearances of the at least one roadway point in the at least two registered plan views.
• the registration is realized in this embodiment by a simple translation and rotation, since the scene has been transformed into a plan view.
• the compensation may preferably be done by incorporating individual driving dynamics parameters, e.g. Vehicle speed, steering angle, etc. or entire models, e.g. Groundplane model and vehicle dynamics models are made or supported.
• individual driving dynamics parameters e.g. Vehicle speed, steering angle, etc.
• entire models e.g. Groundplane model and vehicle dynamics models are made or supported.
• the individual features form a feature vector, which is then assigned by at least one class from a classification system (classifier).
• classifier is a mapping of a feature descriptor to a discrete number representing the classes to be recognized.
• classifier is preferably a random
• Decision Forest Random Decision Forest
• Decision trees are hierarchically arranged
• Classifiers that iteratively split the classification problem. Beginning in the root, the path to a leaf node is made on the basis of the decisions made, in which the final classification decision takes place. Due to the learning complexity, very simple classifiers, the so-called decision stumps, which separate the input space orthogonally to a coordinate axis, are preferably used for the inner nodes.
• Decision forests are collections of decision trees that contain randomized elements in training trees, preferably at two sites. First, each tree is trained with a random selection of training data, and second, for each binary decision, only a random selection of allowed dimensions is used. In the leaf nodes, class histograms are stored that contain a
• Class histograms store the frequency at which a feature vector of a particular lane condition occurs
• the decision trees preferably be assigned a probability that is calculated from the class histograms.
• a probability that is calculated from the class histograms.
• This decision per input image is preferably followed by an optimization.
• This optimization may take into account temporal context or other information provided by the vehicle.
• Temporal context is preferably taken into account by using the most common class from a previous period or by using a so-called hysteresis threshold method.
• hysteresis threshold method the change from one road condition to the other is regulated by means of threshold values. A change takes place only when the
• the at least two images produced, particularly preferably the generated plan views are averaged to a
• Average image and the associated column average is formed.
• a basic assumption of this embodiment is that a region moves through the entire image area. It does not look at a particular region in itself, but their path. Thus, more preferably, more than two images are generated. It is further assumed that there is a straight-line and continuous change in the perspective, preferably a uniformly rectilinear movement of the vehicle. This assumption may preferably be confirmed by vehicle motion parameters as contextual knowledge. Under this condition, the frames, preferably the transformed frames, are averaged from the sequence to obtain an average image. In order to minimize storage space or to weight younger events higher, the calculation of a moving average is also possible.
• the absolute difference or the quadratic difference is formed between each pixel in the average image and the associated mean of the column.
• the extraction of features of the average image can be done considering the column averages, preferably using a "bag-of-visual-words" approach, in which the occurrence of certain prototypical values or value tuples is detected on the basis of a histogram.
• the resulting image can be used to evaluate the presence of specular reflections, such as statistical moments or, in a particularly advantageous form, local features (preferably "local binary patterns") in a "bag-of-visual words". Approach.
• specular reflections such as statistical moments or, in a particularly advantageous form, local features (preferably "local binary patterns") in a "bag-of-visual words”.
• local features preferably "local binary patterns”
• Mean column values are very similar when moving in a straight line, whereas, in specular reflections, the change in the appearance of the passaged regions is very different from the mean in the column. This method is based - as mentioned above - in
• Another advantage is the required computing time, which, in contrast to the former method, is greatly reduced.
• the calculations are limited to averaging and some subtractions.
• the ever-changing exposure time is taken into account, which causes changes in the appearance of the regions in the sequence (for example, brightness changes) and the detection of reflections negative
• the device according to the invention for detecting and evaluating reflections of at least one point on a roadway according to claim 13 comprises a camera, which is adapted to at least two digital images of the at least one
• the device is set up using algorithms of digital image processing
• FIG. 1 a and b show a schematic representation of a
• the device 1 according to the invention shown by FIGS. 1 a and 1 b comprises a digital camera 2, which is set up to record at least two digital images of a roadway point 3 from different recording perspectives, wherein the different recording perspectives are represented by two different positions A and B, respectively the respective camera 2 are shown.
• the camera 2 is arranged in a vehicle, not shown, behind its windshield, so that a visual perception of a driver of the vehicle can be detected corresponding to the apron of the vehicle.
• a driving movement of the vehicle is moved from a first position to a second position. In the first position, in which the camera 2 covers the recording perspective A shown on the right in FIGS. 1a and 1b, a first image of the roadway point 3 is taken in each case.
• the vehicle is moved to the second position, in which the
• Recording perspective B is covered, from each of which a second image of the roadway point (3) is recorded.
• a shot perspective from A to B because an incoming light beam 4 is equally reflected in all directions by a dry road surface 5. This corresponds to a diffuse reflection, which is an indicator of a dry road surface.
• the device 1 compares the first and second images with each other. Using digital image processing algorithms, the device 1 recognizes that the first and the second image do not differ from each other or only to an extent in that a diffuse reflection must be present. Due to the detected or detected diffuse
• Reflection determines the device 1 a
• Road condition information which includes that the road surface 5 is dry. This value is transmitted to an unrepresented driver assistance system.
• the image of the roadway point 3 changes from A to B when the recording perspective changes, because an incident light beam 6 changes in only one direction from an iced or wet one
• Road surface 7 is reflected. This corresponds to a specular reflection, which is an indicator of a wet or icy road surface.
• the device 1 compares the first and second images with each other. Using digital image processing algorithms, the device recognizes that the first and second images are so much different from each other differ that a reflective reflection must be present. Due to the detected or reflected specular reflection, the device determines road condition information, which includes that the road surface is wet or iced. This value is sent to an unillustrated

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Multimedia (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Software Systems (AREA)
• Transportation (AREA)
• Mechanical Engineering (AREA)
• Mathematical Physics (AREA)
• Automation & Control Theory (AREA)
• Image Analysis (AREA)
• Traffic Control Systems (AREA)
• Image Processing (AREA)
• Control Of Driving Devices And Active Controlling Of Vehicle (AREA)

Priority Applications (6)

Applications Claiming Priority (2)

Publications (1)

Family

ID=56097956

Family Applications (1)

Country Status (7)

Cited By (3)

Families Citing this family (10)

Citations (4)

Family Cites Families (12)

• 2015
  • 2015-05-06 DE DE102015208429.9A patent/DE102015208429A1/de not_active Withdrawn
• 2016
  • 2016-05-04 CN CN201680026128.2A patent/CN107667378B/zh active Active
  • 2016-05-04 JP JP2017557347A patent/JP6453490B2/ja active Active
  • 2016-05-04 DE DE112016002050.3T patent/DE112016002050A5/de active Pending
  • 2016-05-04 KR KR1020177031347A patent/KR101891460B1/ko active Active
  • 2016-05-04 US US15/572,010 patent/US10442438B2/en active Active
  • 2016-05-04 WO PCT/DE2016/200207 patent/WO2016177371A1/de not_active Ceased
  • 2016-05-04 EP EP16726768.1A patent/EP3292510B1/de active Active

Patent Citations (4)

Non-Patent Citations (3)

Also Published As

Similar Documents

Legal Events