EP3769041A1 - System zur erzeugung von konfidenzwerten im backend - Google Patents
System zur erzeugung von konfidenzwerten im backendInfo
- Publication number
- EP3769041A1 EP3769041A1 EP19714341.5A EP19714341A EP3769041A1 EP 3769041 A1 EP3769041 A1 EP 3769041A1 EP 19714341 A EP19714341 A EP 19714341A EP 3769041 A1 EP3769041 A1 EP 3769041A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- backend
- objects
- data
- probability
- unit
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/38—Electronic maps specially adapted for navigation; Updating thereof
- G01C21/3804—Creation or updating of map data
- G01C21/3807—Creation or updating of map data characterised by the type of data
- G01C21/3815—Road data
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/26—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 specially adapted for navigation in a road network
- G01C21/34—Route searching; Route guidance
- G01C21/36—Input/output arrangements for on-board computers
- G01C21/3602—Input other than that of destination using image analysis, e.g. detection of road signs, lanes, buildings, real preceding vehicles using a camera
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/38—Electronic maps specially adapted for navigation; Updating thereof
- G01C21/3804—Creation or updating of map data
- G01C21/3833—Creation or updating of map data characterised by the source of data
- G01C21/3841—Data obtained from two or more sources, e.g. probe vehicles
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/38—Electronic maps specially adapted for navigation; Updating thereof
- G01C21/3804—Creation or updating of map data
- G01C21/3859—Differential updating map data
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/70—Determining position or orientation of objects or cameras
-
- 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
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- 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
Definitions
- the invention relates to a Obj ektkonfidenzwerter Wegungssystem for generating confidence values for objects in a digital road map, a vehicle with a rejennungsvor direction, a backend with a digital road map, a method for generating confidence values for objects in a digital road map, program elements and a computer-readable medium ,
- Vehicles are increasingly seen ver with driver assistance systems, which support the driver in the implementation of driving maneuvers. Furthermore, vehicles are increasingly equipped with highly or fully automatic driving functions. These highly or fully automated driving functions require highly accurate digital road maps to ensure safe and reliable vehicle navigation and to detect objects such as traffic signs or lane markings. Furthermore, these digital road maps must always have the current state of the roads and traffic signs to enable the highly or fully automatic driving functions. Furthermore, today's vehicles have a large number of sensors for detecting the vehicle environment.
- a first aspect of the invention relates to an object confidence value generation system for generating confidence values for objects in a digital road map.
- the object confidence value generation system has a backend and an object recognition device for a vehicle.
- the object recognition device in turn has a detection unit, an evaluation unit, a positioning unit and a transmitting / receiving unit.
- the detection unit is set up to capture environmental data from a vehicle and to divide the environmental data into a plurality of two- or three-dimensional segments.
- the positioning unit is configured to determine the positions of the segments and the objects contained therein.
- the evaluation unit is set up to recognize the objects and occlusions of the objects in the segments and to provide them with position information.
- the evaluation unit is further configured to determine a probability for a correct recognition for each of the objects and for each occlusion, wherein the probability of the relative position of the segment in which the object or the occultation is located to the object recognition device or depends on the vehicle.
- the transmitting / receiving unit is set up to transmit the data generated by the evaluation unit to the backend.
- the backend is configured to receive the data of the transceiver unit and to generate, modify or update the digital road map. Furthermore, each of the digital road map objects of the backend has a confidence value.
- the backend is adapted to increase the confidence value of the respective object depending on the determined probability, if the respective object is included in the received data, and to reduce the confidence value of the respective object depending on the determined probability, if the respective object is not in the received data is included, wherein the backend is adapted to perform in a masking of the object in the received data no reduction of the confidence value of the respective object by.
- each of the objects of the digital road map may have an individual confidence value indicating how reliable the information, eg existence, position or content, is about that object. This confidence value may be changed by the backend based on data of the object recognition device or a plurality of object recognition devices.
- the confidence value may be, for example, between 0 and 1, where 1 indicates a maximum reliability of the information about the respective object and 0 indicates a minimum reliability of this information. This may be based on the idea that a plurality of object recognition devices recognizes or does not recognize an object (because it no longer exists). Thus, individual errors in the detection by individual reerkennungsvor directions can be excluded on average, resolved or reduced. As a result, the confidence value of the object can be gradually adjusted and changes of the objects in the road traffic can be detected automatically and continuously by the backend, so that always a current Reconcesen tion of reality in the digital road map on the backend can be mapped.
- the confidence value of a new object exceeds a predefined threshold value, it can be assumed that this object is also present in reality and it can be added to the digital road map on the backend. If, however, the confidence value of an object falls below a predefined threshold value, it can be assumed that this object no longer exists in reality and thus it can be removed from the digital road map of the backend.
- the detection unit eg a camera, a stereo camera, a radar sensor, a lidar sensor, an ultrasound sound sensor, a laser scanner or a combination thereof, the collected environment data in a plurality of two- or three-dimensional segments.
- the segmentation and the respective position of the segment can be used to determine a probability for the correct recognition of the object or the occlusion.
- the relative position of the segment of the respective object or occlusion to the object recognition device or to the vehicle may influence the probability of the correct recognition. The greater the distance between the object and the object recognition device or the vehicle or the farther an object is in an edge region of the environment data of the detection unit, the lower the probability that the object or the occlusion has been recognized correctly.
- the probability of correct recognition may be higher if the object or occultation is centrally located in front of the detection unit at a small distance.
- a probability for a correct recognition of the object recognition device can be determined.
- the probability of correct recognition can be determined for the respective segments, whereby the probability for the correct recognition can depend on the position of the segment in the field of view (detection area) of the detection unit.
- the probability of correct recognition for the detected objects or occlusions may be determined by the segment in which the object or occlusion is located.
- the segments are constant, cuboidal volumes which have a width, a height and a depth which are constant in time.
- the division into segments can be based on Cartesian coordinates. It should be noted that the segments may be cuboids or cubes.
- the segments are circular sectors, which are characterized by a plurality of radii are broken, and which have an angle, a height and a start and an end radius.
- the segments can be divided based on cylindrical coordinates.
- a camera image can be divided into two-dimensional segments and the depth information is transmitted via another sensor, e.g. a lidar sensor.
- the backend may increase the confidence value of an object depending on the determined probability of correct recognition of that object if it has been detected by the object recognition device and included in the data transmitted to the backend. Further, the backend may reduce the confidence value of an object depending on the determined probability of correct recognition of that object if it has not been recognized by the object recognition device or is not included in the data transmitted to the backend. Further, the backend can not change the confidence value of the digital road map object if the area of the putative object is obscured, e.g. by another road user (truck) or a building (tree). Due to the occlusion, the object could not be recognized by the object recognition device because it was not visible.
- trucks road user
- tree building
- the backend can also incorporate other parameters in the change of the confidence value of the objects, such as the time course.
- the backend may be a new object that should be added to the digital roadmap.
- it may behave with an object that should be removed from the digital road map; for example, if an object has not been detected in the last 10 transmitted data without any obfuscation, the object may be removed from the digital roadmap as it is likely is no longer present in reality, although the confidence level can still be high.
- the most recent detection or non-recognition of the object can affect the confidence value more than older detections.
- the confidence levels may decrease over time unless confirmed by new detections.
- digital road maps are also road maps for advanced driver assistance systems (ADAS, Advanced Driver Assistance System) to understand without a navigation takes place.
- ADAS Advanced Driver Assistance System
- these digital road maps can be stored and created in a backend.
- An object in this case may be, for example, a traffic sign, a guardrail, a road marking, a traffic light system, a gyro, a crosswalk or a speed hill.
- the transceiver can transmit the data to the backend wirelessly, over the air.
- the wireless transmission or the wireless reception of the data can be via Bluetooth, WLAN (eg WLAN 802.1 la / b / g / n / ac or WLAN 802.1 lp), ZigBee or WiMax or by means of cellular radio systems such as GPRS, UMTS, LTE or 5G. It is also possible to use other transmission protocols.
- the mentioned protocols offer the advantage of already existing standardization.
- a backend can be understood to mean a computing unit which is located outside the driver's own vehicle and which can be used for a large number of vehicles or object recognition devices is available.
- the backend may be, for example, a server or a cloud which can be reached via the Internet or another network.
- the object identification device for the vehicle may be performed by the object identification device for the vehicle and some steps in the backend.
- the distribution between the evaluation by the evaluation unit of the object recognition apparatus and the backend can be adapted to the respective application case.
- the complete evaluation in the backend can be performed.
- the object recognition device of the vehicle in this case serves as a data acquisition unit for the backend.
- the evaluation can also be carried out on the object recognition devices and the result is then transferred or transmitted to the backend, in which the data of the individual vehicles are combined, so that the digital road map with corresponding confidence values for each object in the digital road map , eg Road signs, arises.
- the data transmitted by the transmitting / receiving unit is vector data.
- the data can be transmitted as vector data, so the information about the existence of an object and its position and not the entire image data can be transmitted.
- the Auswer teech is adapted to determine based on the distance and the angle of the detected objects to the object recognition device in the respective segment, the probability of correct detection of the objects or the obscurations of the objects, with a greater distance between the respective segment and the object recognition device leads to a lower probability for the correct recognition and a lesser distance between the respective segment and the object recognition device results in a higher probability of correct recognition.
- the probability of correct recognition may depend on the relative position of the segment to the object recognition device. The greater the distance and the farther the object is located at an edge area of the detection unit, the lower the probability of a correct recognition of the object or the concealment. Analogous to this, the probability of a correct detection can be higher, the more center and / or closer an object or a concealment with respect to the object recognition device be.
- the backend draws the object or the obscuration of the object only for the determination of the confidence value, if the likelihood for the correct detection is above achy certain threshold.
- the threshold value for the probability as of when the detected objects or occlusions are taken into account in the confidence value can be 30, 40 or 50%; all detected objects with a probability below this threshold value are not used to determine the confidence value in the confidence value digital road map of the backend used.
- the system may be located in a remote segment, such as B. in 80m distance to the object recognition device, recognize an object, which due to the geometry of a traffic sign could be.
- the object recognition device can remember the distance and thus actually the position of the object.
- accurate recognition of the object itself could not be done, so the likelihood of correctly recognizing this object may be low (it was a speed limit of 60 or 80, or none at all).
- the object is obscured due to viewing angle or a road user.
- the evaluation unit of the object recognition device would detect occlusion of the object, since the probability for the correct recognition lies below the predefined threshold value, and the adjustment of the confidence value of the object in the backend would not occur.
- an adjustment of the confidence value can be omitted if the object, its position or its content was not detected with a probability for a correct detection above a threshold value.
- the Auswer teech is adapted to evaluate the environment data of a covered by the vehicle portion of a route.
- the send / receive unit is set up to send the data of the entire section to the backend.
- the object recognition device in the vehicle can first collect and evaluate the data for a particular section, and then transmit the entire section or result for the entire section to the backend.
- the section can also be referred to as a snippet
- the recognition (visibility) or the occlusion (non-visibility) can be expanded by a second probability factor of the correct recognition. This increases if, in as many individual images (frames) of the respective section, the object was detected or it is obscured. In other words, the second probability value depends on the relative frequency of detection of the object or the concealment. Thus, not only discrete values such as 0 (occlusion) and 1 (recognition) may occur for the recognition or occlusion, but also arbitrary values in between.
- the respective sections can also be defined with a certain overlap with each other, so that at the beginning of the respective section also several pictures (frames) are present and the beginning does not consist of just one picture.
- the detections in the sections can be normalized so that the number of detections is divided by the number of total existing pictures which the respective object was able to recognize.
- the evaluation of recognizable and hidden areas can also be carried out completely in the backend.
- the vehicle or the object recognition device in the vehicle serves as a data acquisition unit and the data processing or the evaluation takes place in the backend.
- a highly accurate position is not needed for each object, but only for the beginning of the section.
- the remaining position determination can be realized on the basis of the image data in the backend, wherein the backend typically has a higher computing power than the evaluation unit.
- highly accurate position information must be transmitted to the backend only for the beginning and end of each section, and the backend can then determine or determine the respective positions, objects, and occlusions from the obtained data.
- the section is 100m long.
- any other length may be chosen for the section, e.g. 200m, 500m or 1km.
- the section can be adjusted depending on the road or the environment, so that the section is shorter in an urban environment than in a rural environment or on a highway. This can be advantageous, since in the urban environment typically more objects are present in the same section length.
- the section may also be defined on the basis of a fixed file size, e.g. 5MB, 25MB or 1GB, but the section or length of the section can also be defined by a combination of the above criteria.
- the evaluation unit is set up to classify all the detected segments in front of the vehicle as hidden areas in the respective section and if an object in a segment has been detected or no occlusion is detected in the segment, the corresponding segment as classify the visible area and determine the probability of correct recognition.
- the backend is set up not to include the hidden segments in the determination of the confidence values of the individual objects of the digital road map.
- the evaluation unit can evaluate the collected data in sections, ie for each section or snippet, and transmit them to the backend. In the evaluation of the section, it is also possible to proceed in such a way that first all areas are classified as hidden areas, ie areas in which a statement about any objects is not possible. These hidden areas are then classified by means of the evaluation by the evaluation unit gradually as recognized areas (recognized objects or ver Vercoverept including their probability for a correct recognition). In other words, successively, the recognized area becomes larger as more objects are detected therein. The unrecognized areas remain classified as hidden areas and a confidence value of an object that may be in this area is not adjusted because a reliable statement due to the occlusion or not recognition in this area is not possible.
- the object recognition apparatus further comprises a memory unit.
- a digital road map is stored with a plurality of objects on the storage unit.
- the evaluation unit is set up to compare the detected objects with the objects stored on the memory unit.
- the evaluation unit is further configured to notify detected objects that are not present in the digital road map of the storage unit or unrecognized objects that should be recognized according to the digital road map to the backend.
- the object recognition device may further comprise a memory unit on which a digital road map is stored. Furthermore, the digital roadmap on the memory card can be updated by the backend at regular intervals, so that the current version of the digital roadmap for the evaluation unit of the object identification device is always present. Further, the evaluation unit may compare the detected objects with the objects on the digital road map of the storage unit. The evaluation unit can then report detected or unrecognized objects directly to the backend. For example, an object is present in the digital road map, but has not been recognized by the evaluation unit, thus the confidence value of the Object in the backend, depending on the particular likelihood of correct recognition reduced.
- the confidence value of the object in the backend can be increased depending on the specific probability of correct recognition. If an object was detected by the evaluation unit, which is not contained in the digital road map, the detected object can be added to the digital road map in the backend.
- the data in the memory unit can be temporarily stored in the memory unit during the segmental evaluation of the acquired data until they are transmitted to the backend.
- the positioning is n istssussi a GPS module.
- GPS stands for all Global Navigation Satellite Systems (GNSS), such as e.g. GPS, Galileo, GLONASS (Russia), Compass (China) or IRNSS (India).
- GNSS Global Navigation Satellite Systems
- the position determination of the vehicle can also be done via a cell positioning. This is particularly useful when using GSM, UMTS, LTE or 5G networks.
- the environmental data are camera images, camera images of a stereo camera, laser images, lidar images or ultrasound images.
- the digital road map is intended for a vehicle or for the navigation of a vehicle.
- the digital roadmap created in the backend may be for a vehicle to provide navigation for it. Especially with highly or fully automated Vehicles require a precise and up-to-date digital road map.
- the backend is set up to transmit the digital road map to the memory unit of the object recognition device at periodic intervals.
- These periodic intervals may be, for example, once a month, every day or once a year.
- the backend can also transfer the digital road map to the vehicle as needed, for example when navigation on the vehicle has been started.
- the backend is set up to transmit only objects with a confidence value above a predefined threshold value to the memory unit of the object recognition apparatus.
- objects may be included in the digital road map and transmitted to the object recognition device if the confidence value of the digital road map objects on the backend exceeds a predefined threshold, e.g. have a confidence level of over 90%. This can serve to only transfer objects to the vehicle which are very likely to be present.
- the backend is configured to evaluate the received data and to remove or integrate unrecognized objects or new objects into the digital road map of the backend based on the received data.
- the backend can change the digital road map based on the received data from the transmitting / receiving unit of the object recognition device, update and possibly add new objects or remove old objects, so that there is always a current digital street map on the backend, which represents the current road conditions.
- the backend can update the exact position of the individual objects in the digital road map based on the received data, for example, because an increasing number of detections of the respective object can more accurately determine the position on average by the individual object recognition devices.
- the object recognition device has a detection unit, an evaluation unit, a positioning unit and a transmitting / receiving unit.
- the detection unit is set up to capture environmental data from a vehicle and to divide the environmental data into a plurality of two- or three-dimensional segments.
- the Positionie tion unit is set up to determine the positions of the segments and the objects contained therein.
- the evaluation unit is set up to recognize the objects and occlusions of the objects in the segments and to provide them with position information.
- the evaluation unit is further configured to determine for each of the objects and for each occlusion a probability of a correct recognition, whereby the probability of the relative position of the segment in which the object or the occultation is located depends on the vehicle.
- the transmitting / receiving unit is set up to transmit the data generated by the evaluation unit to a backend.
- the vehicle is, for example, a
- a motor vehicle such as a car, bus or truck, or even a rail vehicle, a ship, an aircraft, such as a helicopter or airplane, or, for example, a bicycle.
- Another aspect of the invention relates to a backend with a digital road map and confidence values for their objects.
- the backend is configured to receive the data of a transceiver unit and to generate, modify or update the digital roadmap. Furthermore, each of the digital road map objects of the backend has a confidence value.
- the backend is configured to increase the confidence value of the respective object depending on the determined probability, if the respective object is included in the received data, to reduce the confidence value of the respective object depending on the determined probability, if the respective object is not in the received data is included, wherein the backend is adapted to perform in a masking of the object in the received data no reduction of the confidence value of the respective object by.
- the backend may receive data from a vehicle object detection device.
- the objects or occlusions contained in the data may use the backend to change the confidence values of the objects in the digital road map, depending on the determined probabilities of correct recognition.
- the object recognition device can perform an evaluation before the data is sent to the backend or else the object recognition device sends all data acquired by the acquisition unit to the backend and the evaluation of the data takes place (mostly) in the backend.
- the detection unit can capture environmental data of a vehicle, these can then be evaluated by the evaluation of the object recognition device or by the backend. In this case, the data from the object recognition device can be transmitted continuously or in sections to the backend.
- the data of the object recognition device may include a plurality of different objects or occlusions of objects each having a probability of correct recognition, depending on the relative position of its respective segment to the object recognition device.
- the backend can change the confidence values of the objects contained in the digital roadmap on the backend.
- a confidence value of an object in the digital road may be increased depending on the determined probability of correct recognition when the object is included in the data of the object recognition device.
- a new object may be added to the digital road map when an object is included in the data of the object recognition device, but in the digital road map until now, there was no object at that position.
- the backend may reduce or reduce the confidence value of an object depending on the determined probability of correct recognition when an object contained in the digital road map has not been recognized by the object recognition device. This reduction of the confidence value can not be performed if the object recognition device detects occlusion in the area of the object. In other words, the detection unit of the object recognition device did not have the possibility to recognize an object, because it was hidden, there is no adjustment of the confidence value for this object, since no definitive statement can be made.
- the backend may further remove an object from the digital road map if the confidence value of that object falls below a predefined threshold, eg, 60% or 70%, or if the object has not been detected by a predefined number of vehicles, eg, 10, in recent times has been.
- a predefined threshold eg, 60% or 70%
- the backend directly an unrecognized object with the corresponding probability of correct recognition, ie an object which should have been recognized but was not recognized report. However, this will be a digital
- Road map needed as a comparison in the object recognition device this may for example be stored on a memory unit. Even newly recognized objects, that is, which until now were not present in the digital road map of the object recognition device, can be reported to the backend.
- the backend can receive and process data from a plurality of object recognition devices of different vehicles simultaneously.
- the respective object recognition devices of the vehicles can be used as sensors to keep the digital road map on the backend up to date, to update or to change and to add new objects or to delete old objects.
- the current digital road map can be returned to the object recognition devices by the backend
- Vehicles are transmitted so that they always have the current digital road map available.
- Another aspect of the invention relates to a method for generating confidence values for objects in a digital road map. The method comprises the following steps:
- a further aspect of the invention relates to program elements which, when executed on an evaluation unit and a back end of an object confidence value generation system, instructs the evaluation unit and the backend to perform the method described above and below.
- a further aspect of the invention relates to a computer-readable medium having stored thereon a program element which, when executed on an evaluation unit and a backend of an object confidence value generation system, instructs the evaluation unit and the backend to perform the method described above and below.
- FIG. 1 shows a block diagram of an object confidence value generation system according to an embodiment of the invention.
- Fig. 2 shows a schematic representation of environmental data with a recognized object according to an embodiment of the invention.
- FIG. 3 shows a schematic representation of environmental data with a concealment according to an embodiment of the invention.
- Fig. 4 shows a division of the field of view of the detection unit in cuboid according to an embodiment of the invention.
- Fig. 5 shows a division of the field of view of the detection unit by means of cylinder coordinates according to an embodiment of the invention.
- 6 shows a schematic illustration of a vehicle trajectory with a sectional evaluation of the environmental data according to an embodiment of the invention.
- Fig. 7 shows a vehicle having an object recognition device according to an embodiment of the invention.
- FIG. 8 shows a flowchart for a method for generating confidence values for objects in a digital map in the backend according to an embodiment of the invention.
- FIG. 1 shows a block diagram of an object confidence value generation system.
- the object confidence value generation system comprises an object recognition device 1 and a backend 2.
- the object recognition device 1 in turn has a Auswer teech 10, a positioning unit 11, a transmitting / receiving unit 12, a detection unit 13 and a memory unit 14 on.
- the object recognition device 1 and the backend 2 can exchange data with each other, this can for example be done wirelessly via mobile networks.
- On the backend 2 there may be a digital road map 3 which is updated, changed or generated with data of the object recognition device. Further, the backend 2 data, such as the current digital road map 3, to the object recognition device 1 transmitted.
- the transmitting / receiving unit 12 of the object recognition device 1 can be used for the data exchange.
- the detection unit 13 can be set up to capture environmental data, in particular, the detection unit can capture environment data of a vehicle by means of various sensors.
- the detection unit 13 can be, for example, a camera, a stereo camera, a lidar, a radar, an ultrasound sensor or a combination thereof.
- the detection unit 13 can capture time-sequential environment data, for example a video or a plurality of individual consecutive images (frames).
- the detection unit 13 can divide the environment data into several segments. The division can be based, for example, on Cartesian coordinates (cuboid or cube), cylindrical dinates (cylindrical or circular sectors) or spherical coordinates (spherical sectors).
- the positioning unit 11 may be configured to determine the position of the object recognition device 1 and objects and occlusions detected by the detection unit 13.
- the positioning unit 11 may be a GPS sensor.
- the memory unit 14 may include a digital road map and further, the memory unit may be used as a buffer when a transmission to the backend is done in sections.
- the evaluation unit 10 of the object recognition device 1 can be configured to recognize objects or occlusions in the acquired environment data and to provide them with a position.
- An object in this case may be, for example, a traffic sign, a guardrail, a road marking, a traffic light system, a gyro, a crosswalk or a speed hill.
- the expanding unit 10 can detect a concealment, ie a device, for example a truck or a tree, which covers the object to be detected, ie is located between the object to be detected and the detection unit 13. Thus, no object can be detected because the object was not even detected by the detection unit 13. Furthermore, the detected object or occlusion control unit 10 may determine a probability of correct recognition. This probability may in particular depend on the relative position of the respective segment to the object recognition device 1 or the vehicle.
- the evaluation unit 10 can be set up to compare the detected and recognized objects with the objects stored in the memory unit 14, so that the evaluation unit 10 can determine whether an object has been detected which is not present in the digital road map of the memory unit 14 or that an object should have been present according to the digital street card stored on the memory unit 14 but should not be recognized by the evaluating unit 10.
- These detected or unrecognized objects can report the object recognition device 1 directly to the backend 2.
- the evaluation can be carried out by the evaluation unit 10 and the result is sent to the backend 2 to there the digital road map. 3 adjust or update accordingly.
- the evaluation can also be done in the backend 2, to do this, the object recognition device 1 sends all captured field data, including the probability of correct detection, the detection unit 13 to the backend 2 and the backend 2 evaluates the data.
- computing power can be saved on the object recognition device 1.
- the evaluation of the environmental data can be carried out in sections, ie in fixed sections, of e.g. 100m, where the individual sections can be referred to as snippets.
- the evaluation unit 10 may be configured to first classify everything as hidden areas and to classify it successively when objects or occlusions have been detected, so that all areas that are not reliably recognized are considered to be occlusions and there no statement about the objects located there In this case, the confidence value of these objects will not be updated or adjusted. This procedure may be useful, in particular, in the case of the segmental evaluation, since an object can be detected over a certain period of time or over a certain distance by several different viewing angles and distances.
- multiple identifications of an object can also determine a second probability factor.
- a factor for each object in a respective section which reflects the number of correct detections of the object in a particular section.
- this factor may for example be between 0 and 1, where 1 means that the object in all temporally successive environment data by the evaluation unit 10 was detected and detected.
- the factor may be the number of detections of an object in the environment data of a section by the total number of available environment data. For example, a section consists of 1000 individual frames or frames, and a particular object has been detected in 850 of these pictures, so the factor can be determined to be 0.85.
- the factor may also be the number of detections of the respective object with respect to the number of possible detections.
- the position of the object may be in the middle of a section, so that the object could only be detected by half of the available environment data, e.g. the section consists of 1000 individual images, but due to its position, the object could only exist in 500 images or frames and the object was recognized on 460 of these 500 images, so the factor can be determined to be 0.92.
- the respective object is present on 92% of the environment data.
- This evaluated data with the respective factors and the probability for the correct detection can be transmitted to the backend 2, the backend 2 in turn can ba sierend on this data, the confidence value of the respective object in the digital road map 3 adapt, change or aktua.
- the transmission of the data by the transmitting / receiving unit 12 to the backup 2 can be done both continuously and from cut, in snippets.
- Fig. 2 shows an exemplary image of the environment data.
- a road shown.
- a traffic sign which is recognized by the evaluation unit or by the backend as object 31.
- This image or the position of the detected object 31 can be transmitted to the backend, which in turn the confidence value of this object in the digital road map in the backend depending on the Probability of correct recognition increased.
- the distribution of the environment data shown in different segments are represented by the dashed lines Darge.
- the segments were generated here using Cartesian coordinates.
- the probability of correct recognition may depend on the particular position of the segment in which the object or occlusion is present.
- the edge regions or distant segments may have a lower probability of correct recognition than central and proximal segments. This may in particular be due to optical effects in the detection unit and in the representation on the sensor (eg the camera), distant objects, for example, are represented by only a few pixels.
- FIG. 3 likewise shows an exemplary image of the environment data, but in this case the object is concealed by a concealment 21.
- the occlusion prevents detection of the object.
- the occlusion can arise, for example, by another road user, such as a truck, or by a tree. If an object to be detected is behind an occlusion, the confidence value of this object is not adjusted since no statement about the object can be made. Also in Fig. 3, the division of the environment data into several segments by the ge dashed lines can be seen.
- FIG. 4 shows the division or division of the environmental data acquired by the detection unit 13 into three-dimensional segments.
- the division or the definition of the segments takes place on the basis of Cartesian coordinates.
- the segments are cuboid.
- a side view of a vehicle 4 and a road is shown.
- various dashed lines are provided at different heights represents, which represent the division of the segments in height.
- the individual lines have a distance of about 50cm to each other in this example.
- a plan view of the vehicle 4 is shown.
- a detection unit 13 for example a camera
- the field of view of the detection unit 13 is symbolized by the two oblique dashed lines.
- the opening angle or the field of view may vary depending on the detection unit 13 used. Furthermore, a plurality of rectangles in the field of view of the detection unit 13 is shown. These define the majority of three-dimensional segments. In this case, the segments are cuboids having a width and a length of 2.5m and a height of 0.5m. It should be noted that other breakdowns are possible, for example, Im or 3m. Furthermore, the height and the length may differ from each other.
- FIG. 5 shows, like FIG. 4, the division or division of the environmental data into different segments, but in FIG. 5 the division is made on the basis of cylindrical coordinates.
- the opening angle or the field of view of the detection unit 13 is under-segmented by means of angles, for example 0.5 °, and the definition of the length is based on con centric circles which intersect the lines of sight at different locations.
- the definition of the height is equal to the method shown in FIG.
- Fig. 6 shows a course of a vehicle 4 over a certain time a) to e) along a certain path.
- This path is represented in FIG. 6 by the thick dashed black line.
- each of the parts a) to e) symbolizes an advanced period of time.
- the path can here be divided into several sections 5 (snippets).
- the different sections 5 are shown in FIG. 6 by the thick vertical bars transversely to the direction of the path.
- the sections 5 may have a fixed distance, for example 100m.
- the detection unit of the object recognition device of the vehicle 4 has a certain angle of view. This is represented by the two solid thin lines starting from the vehicle 4. In FIG.
- the detection unit can only detect the objects in front of the vehicle and the lateral only partially, the farther the vehicle moves along the way, the more often has the object recognition device the chance to detect an object 31 in the vehicle environment or the more often a Object 31 or a cover 21 detected.
- the area may also have been partially recognized by the object recognition device or may have been recognized only in a few image data or frames.
- Fig. 6 there are a vehicle 4 which moves along a path, an object 31 to be recognized and a cover 21.
- the vehicle 4 can recognize the object 31 since the view is clear, however because of the distance, the probability of a correct detection is still low.
- the visibility of the object 31 may be 100%, and the probability of correct detection may be 10%.
- the vehicle has already moved further on the way in the direction of the object 31. Furthermore, the object 31 is detected and the
- the probability of a correct recognition increases as the distance between the vehicle 4 and the object 31 becomes smaller.
- the visibility of the object 31 may still be at 100% and the probability for a correct one is 30%.
- the vehicle 4 has advanced even further on the way, so that the object 31 is covered by the cover 21.
- the occlusion 21 is between the object 31 and the vehicle 4.
- the visibility of the object 31 is now 0%, and the probability of correct recognition of the object 31 would be 75%.
- the vehicle 4 has advanced further and the object 31 is still obscured, so that, for example, the visibility is still at 0%, but the probability for the correct recognition based on the distance or the segment relative to the vehicle to 95% increases.
- the cover 21 continues to cover the object 31 so that no object 31 can be recognized.
- the visibility here can be 0% and the probability of a correct detection 100%.
- all areas can first be classified as unrecognized areas, and as soon as recognition takes place, this area can be classified as a recognized area.
- the confidence value of the recognized region may be increased with increasing recognition in a plurality of images or frames.
- FIG. 7 shows a vehicle 4 having an object recognition device 1.
- This object recognition device 1 can detect surrounding data about the vehicle 4 and recognize objects and occlusions therein. Furthermore, the object recognition device 1 can transmit the evaluated data to the backend. Furthermore, a multiplicity of vehicles 4 with an object recognition device 1 can record and evaluate environment data and transmit them to the backend.
- step S1 environment data can be acquired by a detection unit, and these are grouped into a plurality divided into two- or three-dimensional segments. These field data can be individual images or frames or an entire section of successive images or frames.
- step S2 the position determination of the detected environment data and the objects contained therein he can follow.
- An evaluation of the detected environment data can be performed segment by segment in step S3 and in step S4 objects and occlusions in the respective segments of the environment data can be detected.
- step S5 the probabilities for a correct recognition may be determined for each of the detected objects or occlusions, the probabilities depending on the relative position of the segment in which the object or occlusion is located, to the object recognition device or the vehicle.
- step S6 the objects and occlusions detected in step S5 may be provided with positional information.
- step S7 the transmission of the evaluated data to a backend can be carried out by means of a transmitting / receiving unit. Subsequently, in step S8, the digital road map on the backend may be generated, changed or updated based on the received data.
- step S9 a confidence value of a respective object may be increased depending on the determined probability of correct recognition when the respective object is included in the received data.
- step S10 a confidence value of an object may be reduced depending on the probability of a correct recognition if the object is not included in the received data unless the object has been obscured so that it could not be recognized in the acquired environment data.
Landscapes
- Engineering & Computer Science (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Traffic Control Systems (AREA)
- Navigation (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| DE102018204501.1A DE102018204501B3 (de) | 2018-03-23 | 2018-03-23 | System zur Erzeugung von Konfidenzwerten im Backend |
| PCT/EP2019/057090 WO2019180143A1 (de) | 2018-03-23 | 2019-03-21 | System zur erzeugung von konfidenzwerten im backend |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP3769041A1 true EP3769041A1 (de) | 2021-01-27 |
Family
ID=65991768
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP19714341.5A Withdrawn EP3769041A1 (de) | 2018-03-23 | 2019-03-21 | System zur erzeugung von konfidenzwerten im backend |
Country Status (4)
| Country | Link |
|---|---|
| US (1) | US20210003420A1 (de) |
| EP (1) | EP3769041A1 (de) |
| DE (1) | DE102018204501B3 (de) |
| WO (1) | WO2019180143A1 (de) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102018118215B4 (de) * | 2018-07-27 | 2020-04-16 | Man Truck & Bus Se | Verfahren zur Aktualisierung einer Umgebungskarte, Vorrichtung für die fahrzeugseitige Durchführung von Verfahrensschritten des Verfahrens, Fahrzeug, Vorrichtung für die zentralrechnerseitige Durchführung von Verfahrensschritten des Verfahrens sowie computerlesbares Speichermedium |
| EP4107485A1 (de) * | 2020-02-20 | 2022-12-28 | TomTom Global Content B.V. | Verwendung von kartenänderungsdaten |
| DE102020125448A1 (de) * | 2020-09-29 | 2022-03-31 | Daimler Ag | Kartenplausibilitätsprüfungsverfahren |
| JP7511504B2 (ja) * | 2021-02-16 | 2024-07-05 | 三菱重工業株式会社 | 移動体、移動制御システム、移動体の制御方法及びプログラム |
| JP7301897B2 (ja) * | 2021-03-09 | 2023-07-03 | 本田技研工業株式会社 | 地図生成装置 |
| US12299984B2 (en) * | 2022-06-17 | 2025-05-13 | Comcast Cable Communications, Llc | Systems, methods, and apparatuses for selecting/generating visual representations of media content |
| DE102022003088A1 (de) * | 2022-08-23 | 2024-02-29 | Mercedes-Benz Group AG | Verfahren zur Speicherung und zur Qualitätsbewertung von Kartendaten eines Fahrzeugs |
| WO2025085510A1 (en) * | 2023-10-17 | 2025-04-24 | BrightAI Corporation | Relevance based weighting |
| US20260079026A1 (en) * | 2024-09-17 | 2026-03-19 | Tomtom Global Content B.V. | Automated Digital Map Update Based on Media Messages |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19916967C1 (de) | 1999-04-15 | 2000-11-30 | Daimler Chrysler Ag | Verfahren zur Aktualisierung einer Verkehrswegenetzkarte und kartengestütztes Verfahren zur Fahrzeugführungsinformationserzeugung |
| US8818024B2 (en) * | 2009-03-12 | 2014-08-26 | Nokia Corporation | Method, apparatus, and computer program product for object tracking |
| US20100256981A1 (en) * | 2009-04-03 | 2010-10-07 | Certusview Technologies, Llc | Methods, apparatus, and systems for documenting and reporting events via time-elapsed geo-referenced electronic drawings |
| US9565403B1 (en) * | 2011-05-05 | 2017-02-07 | The Boeing Company | Video processing system |
| US8849567B2 (en) * | 2012-05-31 | 2014-09-30 | Google Inc. | Geographic data update based on user input |
| DE102012220158A1 (de) | 2012-11-06 | 2014-05-22 | Robert Bosch Gmbh | Verfahren zur Aktualisierung von Kartendaten |
| DE102014220687A1 (de) | 2014-10-13 | 2016-04-14 | Continental Automotive Gmbh | Kommunikationsvorrichtung für ein Fahrzeug und Verfahren zum Kommunizieren |
| KR101678095B1 (ko) * | 2015-07-10 | 2016-12-06 | 현대자동차주식회사 | 차량, 및 그 제어방법 |
| EP3131020B1 (de) | 2015-08-11 | 2017-12-13 | Continental Automotive GmbH | System und verfahren zur zweistufigen objektdatenverarbeitung durch ein fahrzeug und server-datenbank zur erzeugung, aktualisierung und abgabe einer präzisen strasseneigenschaftsdatenbank |
| DE102016216154A1 (de) * | 2016-08-29 | 2018-03-01 | Bayerische Motoren Werke Aktiengesellschaft | Verfahren und Auswerteeinheit zur Ermittlung der Position von Verkehrszeichen |
| US11067995B2 (en) * | 2017-03-20 | 2021-07-20 | Mobileye Vision Technologies Ltd. | Navigation by augmented path prediction |
-
2018
- 2018-03-23 DE DE102018204501.1A patent/DE102018204501B3/de not_active Expired - Fee Related
-
2019
- 2019-03-21 EP EP19714341.5A patent/EP3769041A1/de not_active Withdrawn
- 2019-03-21 US US16/982,884 patent/US20210003420A1/en not_active Abandoned
- 2019-03-21 WO PCT/EP2019/057090 patent/WO2019180143A1/de not_active Ceased
Also Published As
| Publication number | Publication date |
|---|---|
| US20210003420A1 (en) | 2021-01-07 |
| DE102018204501B3 (de) | 2019-07-04 |
| WO2019180143A1 (de) | 2019-09-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| DE102018204501B3 (de) | System zur Erzeugung von Konfidenzwerten im Backend | |
| EP3769042B1 (de) | System zur erzeugung von konfidenzwerten im backend | |
| DE69635569T2 (de) | Vorrichtung zum Bestimmen der lokalen Position eines Autos auf einer Strasse | |
| DE102020102725B4 (de) | Verfahren und vorrichtung für eine kontextabhängige, aus der crowd-source stammende, spärliche, hochauflösende karte | |
| DE102010006828B4 (de) | Verfahren zur automatischen Erstellung eines Modells der Umgebung eines Fahrzeugs sowie Fahrerassistenzsystem und Fahrzeug | |
| EP2149132B1 (de) | Verfahren und vorrichtung zur erkennung von verkehrsrelevanten informationen | |
| DE112018006665T5 (de) | Verfahren zum zugreifen auf ergänzende wahrnehmungsdaten von anderen fahrzeugen | |
| DE102018005869A1 (de) | System zur Erstellung eines Umgebungsmodells eines Fahrzeugs | |
| DE102012208974A1 (de) | System und Verfahren zum sensorbasierten Aufbauen eines Umgebungsmodells | |
| EP3207538A1 (de) | Kommunikationsvorrichtung für ein fahrzeug und verfahren zum kommunizieren | |
| EP3465652A1 (de) | Verfahren vorrichtung und system zur falschfahrererkennung | |
| DE102015225900B3 (de) | Verfahren und Vorrichtung zur kamerabasierten Verkehrszeichenerkennung in einem Kraftfahrzeug | |
| EP3380810B1 (de) | Verfahren, vorrichtung, kartenverwaltungseinrichtung und system zum punktgenauen lokalisieren eines kraftfahrzeugs in einem umfeld | |
| EP3999806B1 (de) | Verfahren und kommunikationssystem zur unterstützung einer wenigstens teilweise automatischen fahrzeugsteuerung | |
| EP3292423A1 (de) | Diagnoseverfahren für einen sichtsensor eines fahrzeugs und fahrzeug mit einem sichtsensor | |
| DE102018112888A1 (de) | Systeme und Verfahren zur Überprüfung von Strassenkrümmungskartendaten | |
| WO2020043246A1 (de) | Lokalisierungsvorrichtung zur visuellen lokalisierung eines fahrzeugs | |
| DE102016220581A1 (de) | Verfahren und vorrichtung zur bestimmung eines umfeldmodells | |
| DE102018117830A1 (de) | Digitale Umfeldkarte mit Sensorreichweiten | |
| EP3465655A1 (de) | Verfahren vorrichtung und system zur falschfahrererkennung | |
| EP3465653A1 (de) | Verfahren, vorrichtung und system zur falschfahrererkennung | |
| WO2017148851A1 (de) | Verfahren zur präzisen ortsbestimmung eines kraftfahrzeuges | |
| EP3690399B1 (de) | Entfernen von objekten aus einer digitalen strassenkarte | |
| DE102022106248A1 (de) | System zum Steuern einer Kamera zur Unterstützung der menschlichen Überprüfung von Sensorinformationen | |
| DE102024004195B3 (de) | Verfahren zur Verortung eines Fahrzeugs auf einer Fahrspur und Fahrzeug |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20201023 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| AX | Request for extension of the european patent |
Extension state: BA ME |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| RAP1 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: CONTINENTAL AUTOMOTIVE TECHNOLOGIES GMBH |
|
| GRAP | Despatch of communication of intention to grant a patent |
Free format text: ORIGINAL CODE: EPIDOSNIGR1 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: GRANT OF PATENT IS INTENDED |
|
| INTG | Intention to grant announced |
Effective date: 20220914 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE APPLICATION IS DEEMED TO BE WITHDRAWN |
|
| 18D | Application deemed to be withdrawn |
Effective date: 20230125 |