DE102015105784A1 - Distributed system for detecting and protecting vulnerable road users - Google Patents

Abstract

The invention relates to a distributed recognition and communication system (1) for detecting and protecting vulnerable road users (RU, K, L, M, N). Recognition results (31) on identified objects are exchanged between two or more communication participants (CU, A, B, C) of the system (1). In a communication subscriber, vulnerable road users (RU, K, L, M, N) are detected by means of a recognition device and identified as objects, in particular as relevant destination objects for the protection of vulnerable road users. The recognition results (31) on identified objects are stored in an ordered data structure (30) and communicated to the other communication subscribers (CU, B, C) in the system (1). The dispatch of messages with recognition results (31) is controlled on the basis of an analysis of the own (new) recognition results against the stored information (31) about known objects and the known communication participants (CU, 30 '). Various measures are proposed for the control of the shipment and the efficiency of information usage in the system (1) is improved.

Description

The invention relates to a distributed recognition and communication system for recognizing and protecting vulnerable road users, wherein identification results about identified objects are exchanged between the participants of the system. In particular, the invention relates to a recognition and communication method as part of this system, which is provided for execution in a self-communication subscriber. A self-communication subscriber may in particular be a vehicle or a safety device which is arranged on a vehicle and has access to a detection means as well as a communication means.

In practice, it is known that the protection of vulnerable road users can be improved if information is exchanged by several subscribers in a communication system. It has been proposed, in particular, to exchange sensor data, such as images or radar signals, between two vehicles, in order to allow the respective other vehicle additional detection in the sensor data communicated. The exchange of sensor data requires standardized recognition systems in the participants of the system and involves a high load on the communication channels.

On the other hand, the present invention relates to a distributed recognition and communication system or an associated recognition and communication method, in which an object-communication participant performs object recognition on the basis of sensor data, which is determined by a separate recognition device, and exclusively results of recognition to others Communication participants are communicated in the system. There is thus a fundamental difference to the aforementioned systems in which sensor data such as camera images, radar signals, etc. are exchanged.

In other words, in the distributed detection and communication system discussed herein, descriptive information about detected objects is shared among the communication participants in a common traffic environment so as to help each communication participant take action to help protect vulnerable road users.

In part, a generic distributed discovery and communication system according to the present disclosure is abbreviated to "system." A communication participant in the system is sometimes abbreviated as "participant".

The previously known systems do not offer optimal use of the available information. It is an object of the present invention to provide a distributed detection and communication system for the protection of vulnerable road users, in which a more efficient use of the recognition results allows and thus the protection of vulnerable road users are improved.

It is assumed below that in the distributed detection and communication system at least two communication participants are in a common traffic environment and each perform a recognition and communication method according to the independent main claim.

The function of the distributed system arises from the interaction of the communication participants and is explained from the perspective of a participant, which is referred to below as the "self-communication participant". The other participants in the system are referred to as "external communication participants".

The recognition and communication method described for execution in the self-communication subscriber may be executed in the same or a similar manner at one or all of the foreign communication parties.

In the self-communication participant, an object recognition is performed by means of any recognition device. Such recognition devices are known in various variants and are in constant evolution. For example, it is assumed that it is a camera-based recognition device which determines matches between reference objects and acquisition data of an imaging sensor and identifies objects via them. Further, it is believed that the recognition device provides, beyond the existence of an object, further descriptive information that more accurately identifies the object. For this purpose, a relative or absolute Position of the object, a relative or absolute size (height / width) or a type or a type of the object (hereinafter also referred to as "classification").

The recognition device can preferably track an identified object over a time course (object-tracking) and thus determine whether an object moves within the traffic environment and, if necessary, at what speed and in which direction. Also from the tracking of an object descriptive information can be determined as recognition results.

The recognition device is designed to recognize objects of various types in the traffic environment, including at least vulnerable road users. Vulnerable road users are, in particular, people and / or animals who are in the common traffic environment with at least one communication user. These include persons and animals in the area of the traffic and adjacent traffic routes, such as pedestrians, cyclists, skateboard riders (or users of comparable sports or recreational vehicles), children playing, livestock and so on. Preferably, the recognition device has access to a plurality of reference objects that enable recognition and identification. The reference objects can be stored in an extensible database and possibly be partially or completely unified within the system.

The recognition device may also recognize and identify other (real) objects that are not vulnerable road users and access corresponding other reference objects. These include, for example, other vehicles (especially trucks and passenger cars), structures, plants, boundaries of the traffic routes (especially curbs, lane markings, guardrails, delineators, pylons, site boundaries, signs, light markings, etc.).

Detection results on the identified objects (vulnerable road users and possibly other objects) are stored in an ordered data structure of the self-communication subscriber. Furthermore, messages about identified objects are exchanged between the communication participants. The information about identified objects communicated by foreign communication participants is also stored in an ordered data structure. Preferably, there may be a common data structure in which own recognition results and reported recognition results are summarized.

In addition, descriptive information about its own state can be exchanged between the communication participants. Such exchange of self-information is known as vehicle-to-vehicle communication using status messages (so-called "beacon messages").

A distributed recognition system of the type described above can show very different effectiveness depending on the traffic situation, the number of communication participants as well as the number of identified objects and the results to be communicated.

If there are too many communication subscribers communicating each identified objects, a variety of messages may occur, with recognition information about a single real object being communicated from multiple sources multiple times. This leads to unnecessary redundancies. Furthermore, when sending a large number of messages, the transmission channel may become overloaded, so that some of the messages can not be received or read out correctly. This has a negative impact on the quality of the information base shared in the system.

As a result of different recognition quality, it may also happen that two or more communication participants identify a real object, but determine different and possibly contradictory recognition results and communicate them in the system. As a result, erroneous views can be generated in the system.

Especially with the other communication participants, who themselves can not perform their own recognition of such a real object, this leads to erroneous views on the traffic, which are not identifiable from their own competence. Incorrect views in the shared state of knowledge of the system can accordingly cause erroneous reactions, which then do not contribute to the protection of the vulnerable road users or even lead to a hazard. This also affects the efficiency of the system.

The present invention provides that, in the case of a self-communication subscriber, the transmission of messages with recognition results (own and / or received recognition results) is controlled based on an analysis of the own recognition results against the stored information, including information about known objects and the known communication participants , The transmission of messages is preferably controlled by a prioritization, in particular by favoring and / or limiting the sending of messages as well as the message contents to be exchanged.

Message dispatch control is directed to (a) avoiding unnecessary redundancies and / or (b) avoiding misunderstandings in the shared knowledge of the system and / or (c) existing misconceptions in the common knowledge of the system based on the own detection results are detected and / or eliminated and / or (d) an efficiency reduction due to excessive transmission frequency or excessive amounts of information is avoided and / or (e) potential and / or acute threats to vulnerable road users can be managed prioritized.

Each of the above-mentioned effects, by themselves and by combining them, improves the distributed recognition and communication system and its efficiency.

In the dependent claims, the following description and the accompanying drawings advantageous embodiments and applications of the invention are given.

The invention is illustrated by way of example and schematically in the drawings. Show it:

1 a plan view of a traffic situation with multiple participants of a distributed detection and communication system and multiple vulnerable road users;

2 u. 3 : Examples of recognition results stored in an ordered data structure as well as information about communication participants of the system;

4 : Examples of possible detection ranges of a communication subscriber or a recognition device;

5A u. 5B : An example of the origin, detection and elimination of misconception in the common knowledge of the system.

1 shows a plan view of a traffic environment ( 10 ) with several road users. On a first street there is a vehicle (A), which is also understood below as a self-communication participant (A). Disposed on the vehicle (A) is a security device for recognizing and protecting vulnerable road users, which has a computing unit for carrying out the recognition and communication method according to the present disclosure. The security device has access to a communication means as well as a recognition means.

The safety device may be an integral part of the vehicle and may be connected to one or more vehicle control devices in terms of data technology. Alternatively, the safety device can be integrated in a vehicle control unit. Again alternatively, the safety device may be formed as a separate and optionally retrofittable device. Accordingly, the communication means and / or the recognition means may be fixed components of the vehicle which the safety device accesses or the safety device may have its own communication means and / or its own recognition means. A recognition device evaluates the sensor data obtained from at least one detection means and supplies recognition results about identified objects.

Communication within the system ( 1 ) can be wirelessly via one or more arbitrary communication channels, in particular via mobile radio or known vehicle-to-vehicle communication systems. The communication means is designed accordingly.

A detection means may comprise one or more sensors which sensory monitor one or more surrounding areas of the communication party for detection purposes. Such sensors may be imaging sensors such as cameras, 3D cameras or LiDAR / LaDAR sensors. Alternatively or additionally, other sensors may be provided, such as radar sensors, thermal imaging cameras or ultrasonic sensors.

In the context of the present disclosure, it is assumed that a recognition device accesses the sensor data from at least one recognition means and identifies objects within the sensor data that represent real objects in the environment of the communication subscriber. Such real objects belong in the traffic environment ( 10 ) according to the 1 example shown other vehicles (B, C), persons (K, L, M, N), objects ( 14 ), Plants ( 11 . 13 ) or a building ( 15 ) and, if necessary, the other objects mentioned above.

In the example according to 1 For example, in the vehicle (A), the vehicle (B), and the vehicle (C), a recognition and communication method according to the present disclosure is executed. The vehicles (B, C) are each considered in the following description as a foreign communication participants. It is further assumed that the vehicles (A, B, C) possibly exchange status messages about their current state, in particular by sending status messages about their own position, direction, driving speed, size (height / width), etc.

The vehicles (A, B, C) can each have safety devices and recognition devices of different design. Further, the vehicles (A, B, C) may comprise any known active or passive collision avoidance systems as well as personal protection devices. Brake assistants, evasive assistants and the like, for example, come into consideration as a collision avoidance system. Personal security devices can reduce the risk of injury to a vulnerable road user in the event of an unavoidable collision. Such personal protection devices include, for example, external airbags.

Within the scope of the present disclosure, a potential or acute collision situation with a vulnerable road user can be determined by a safety device on the basis of the own and / or received recognition results, and any measures can be taken to avert or mitigate the collision. Such measures may be, inter alia, the display of warnings to the driver, the making of the own vehicle to a road user at risk of conflict and / or the actuation of one of the aforementioned active or passive systems.

In the traffic situation ( 10 ) according to 1 the vehicle (A) drives in a first direction of travel on a priority road in a residential area, passing a footpath on which a woman (K) with a stroller and a child (L) with a ball are. Next to the footpath is a lawn ( 12 ) with a park bench ( 14 ) on which a man (N) sits. The lawn area ( 12 ) is in the direction of travel of the vehicle (A) by a hedge ( 11 ) limited. Between the hedge ( 11 ) and the park bench ( 14 ) is a tree ( 13 ).

In 1 The current movements of the road users are illustrated by block arrows.

On the back of the hedge ( 11 ) is a narrow side street ( 17 ) with the priority road ( 16 ) crosses. On the other side, the narrow side street ( 17 ) through a house ( 15 ), which is built near the roadside.

In the side street ( 17 ) is the second vehicle (B). It is assumed, for example, that the vehicle (B) backwards into the priority road ( 16 ) would like to enter.

In the opposite direction to the vehicle (A) moves on the priority road ( 16 ) a third vehicle (C) and behind a cyclist (M). The cyclist (M) wants in his view to the left in the side street ( 17 ).

It is now further assumed that the child (L) starts suddenly, in the direction of the side street ( 17 ), for example because the ball rolls off in that direction.

It is obvious that this traffic situation ( 10 ) can develop dangerously in the further course. The child (L) could collide with the reversing vehicle (B). The woman (K) could run after the child (L) and let go of the stroller, which may be in the direction of the priority road ( 16 ) could roll where a collision with the vehicle (A) threatens. The driver of the vehicle (A) could by Avoiding the complex situation by avoiding a collision with the stroller (K) by avoiding it, but then bringing it on with the cyclist (M) starting to turn.

As a result of the respectively different recognition devices, the vehicles (A, B, C) have different perceptual surrounding areas, ie sections of the outside environment definable with reference to the respective vehicle (A, B, C), in which vulnerable road users are fundamentally vulnerable by evaluating the sensor data are identifiable. 4 exemplifies the coverage areas ( 20 - 25 ) of the vehicle (A).

A front-entry area ( 20 ) is monitored by one or more high-resolution cameras and at the same time by radar systems that have a long range. A hedge field ( 21 ) is detected by a reversing camera with a wide and substantially downward detection range. A left and a right side detection area ( 22 . 23 ) are monitored by a turn-off assistance system with side-mounted cameras in the fenders. In a left and right rearview mirror detection area ( 24 . 25 ) there is a detection with cameras and / or ultrasonic sensors of a lane change assistant instead. In addition, a parking assistance system with proximity range sensors is arranged on the vehicle (A).

The vehicle (B), however, has a lower sensory equipment. It only has one camera system monitoring a front detection area. In all other surrounding areas the vehicle (B) is sensorially blind.

The vehicle (C) has a front camera, a rear-facing camera and cameras in the fenders, which essentially correspond to the systems of vehicle (A).

It is easy to see that in the traffic situation shown ( 10 ) some of the collision possibilities described above are not recognizable because of a lack of sensor technology by one or more of the vehicles themselves. For example, the vehicle (B) can not determine the approach of the child (L) itself in the absence of a rear sensors. The view of the cyclist (M) could be limited for the front-facing camera of the vehicle (A) by the oncoming vehicle (C), so that the cyclist (M) or only very late by the vehicle (A) are detected can. It could also be that during the onward journey of the vehicle (A), the stroller of the woman (K) is no longer detectable by the front camera and still not by the side cameras when rolling, or that a separate detection takes place too late to a collision to anticipate.

It can now further be assumed that there are numerous other communication subscribers in the wider environment (not shown) and that there is already a considerable utilization of the communication channels.

In order that the distributed detection and communication system can effectively protect the vulnerable road users, various actions executable from the perspective of the vehicle (A) will be explained below, which individually or collectively contribute to the desired effects described above. These measures can preferably also be carried out by the other communication participants (B, C).

The measures are based on the idea that each participant in the system ( 1 ) controls the dispatch of messages with recognition results in such a way that the intervention possibilities of the other participants for the protection of vulnerable road users are thereby promoted or specifically made possible.

A first measure envisages that the self-communication participant tracks objects within the sensor data available to him, ie not only selectively evaluates the descriptive data for each object but comparatively analyzes them for consistency over a certain period of time, for example at least one to two seconds , The reliability of the recognition results is thereby improved. A self-communication subscriber may prefer recognition results via an object newly identified by him only after reaching an adjustable confidence threshold in the system ( 1 ), unless special circumstances suggest an immediate notification of the results (see below). Such a confidence threshold can be defined in different ways, for example, as a minimum time period over which an object has been tracked. Alternatively, or a confidence threshold may be defined based on a confidence score of the recognition device, wherein the confidence score indicates which probability a match between the sensor data and a reference object has been determined.

The aforementioned measure can be used alone or preferably in combination with an additional object classification. Such a classification may preferably distinguish between target objects relevant to the protection of vulnerable road users and other objects. The self-communication participant (A) then preferentially or exclusively sends recognition results about relevant target objects, while recognition results about other objects are subordinated or not communicated at all. The degree of prioritization, in particular the delay or preference of the shipment, can be adjustable depending on further criteria, in particular as a function of a determined utilization on the communication channel.

The two above-mentioned measures, individually or in combination, lead to the fact that the shared state of knowledge in the distributed system is first and / or fastest extended and updated with reliable and primarily relevant information. In contrast, less reliable or less relevant recognition results are not communicated or communicated until the load on the communication channel permits it.

Classification may be based on the descriptive information about the identified objects. For example, all objects to which a collision with a communication participant is anticipated can be classified as a relevant target object, which includes, for example, the woman's stroller. Furthermore, objects identified as vulnerable road users can be classified as relevant destination objects. The classification may possibly be made dependent on the current movement and / or the movement history. Thus, the man sitting on the park bench could not be considered relevant because he does not move for a long time and has a considerable distance from the traffic routes. Further below, further aspects of the classification of relevant target objects are explained.

A further measure, which is applicable on its own or in combination with the abovementioned measures, provides that newly determined recognition results on known objects are communicated as a function of the age of the last exchanged information. A known object can be any object to which recognition results have already been sent or received by the intrinsic communication user.

In the example of 1 For example, the woman (K) with the stroller could have been detected by both the vehicle (A) and the vehicle (C) for an extended period of time and possibly classified as a relevant target object. Furthermore, objects such as the seated man (N), the park bench ( 14 ), The Hedge ( 11 ) and the tree ( 13 ) has long been identified and recognition results in the system ( 1 ). In contrast, the recognition of the child (L) - since it is a small and difficult to identify object - only in the 1 state shown (see also explanations to 5A and 5B below). In particular, it could be that the vehicle (A) and the vehicle (C) substantially simultaneously recognize the child (L) and possibly classify as a relevant target object.

It is positive for the efficiency of the system in the said case when new recognition results about the child (L) are exchanged preferentially, in particular since no history of movement is yet known about the newly recognized child (L).

2 shows an example of an ordered data structure ( 30 ), for example in the form of an object list, in the recognition results ( 31 ) are stored over different objects. In this ordered data structure ( 30 ), preferably own recognition results as well as recognition results communicated by foreign communication participants can be summarized.

Preferably, an object reference number (ID) is assigned as an identifier for each object. This reference number may be included when sending messages. The object reference number is selected here by way of example in accordance with the reference numbers for the vulnerable road users (K, L, M, N). For each identified object, data is preferably stored that enables spatial localization. The spatial localization can be made possible in any way, in particular by absolute position information (x-coordinate, y-coordinate) and / or position information in relation to the communication participants (A, B, C) (on a defined half-line in space / within a Surface area, etc.). Such data can be stored and communicated in any suitable form. In the example of 2 absolute space coordinates (XY / lat-long) are provided for defining a position point or a position area.

Further, a detected moving direction (Dir.) And a current moving speed (V) can be detected and notified. Optionally, a movement history for one or more objects can be recorded over a longer period.

If available, additional descriptive data such as height / height (H) or width (W) of an identified object may be collected and reported.

For each object, an age of the stored information can be detected, i. the time since the last update of the recognition results (Ta). If necessary, a distinction can be made between the age of the last information communicated and the age of the last own recognition results.

3 shows another ordered data structure ( 30 ' ) in which descriptive data about the communication participants (A, B, C) in the system ( 1 ). In the ordered data structure ( 30 ' ) the same or similar descriptive information can be stored according to the example shown as in the ordered data structure ( 30 ). In other words, the known communication subscribers (A, B, C) can also be treated as objects of the traffic environment. The descriptive data can be detected by the communication participants (A, B, C) with high accuracy and exchanged as part of the status messages. Alternatively or additionally, a self-communication subscriber (A) can use his recognition device to ascertain and store recognition results for the foreign communication subscribers (B, C), in particular in order to supplement information from the received status messages.

The classification of the identified objects can be done in any way. For example, the reference objects by which the object recognition is performed may define certain classes. These can be classes such as trees, lane boundaries, buildings, pedestrians, motorcyclists, etc. Alternatively or additionally, the further descriptive information from the recognition results ( 31 ), ie, for example, the position, size, width and speed. It may be advantageous if, within the system ( 1 ) an at least partially unified classification database is used, which is for example regularly obtained in the current version of a provider and updated.

In the example of 2 and 3 For example, the object (K) as an adult, the object (L) as a child, the object M as a bicycle, and the object N as an adult are also classified. If necessary, the external communication participants (B, C) can enter their own classification as part of the status reports, in each case as a car (car).

A further preferred measure, which can be used alone or in any combination with the preceding measures, provides that blind spot surrounding areas (BE) of the communication participants (A, B, C) are determined. It is provided, in particular, that the intrinsic communication subscriber (A) determines whether an object, in particular a relevant target object (K, L, M) is assigned to a blind spot surrounding area of a foreign communication subscriber (B). If such an association is established, recognition results about such relevant target objects (K, L, M) are preferably communicated.

A blind spot environment area (BE) is one such surrounding area of a foreign communication subscriber (B) in which the latter itself can not perform recognition and / or (currently) can not perform recognition. In the following, various variants for determining blind spot environmental areas (BE) are explained.

A first variant provides that a communication participant (A, B, C) in the system ( 1 ) sends a message about in which perceptual environment (AE) he himself performs or can perform a recognition. This can be done, for example, by providing information about the recording systems available to a communication subscriber (A, B, C) and, in particular, the available coverage areas ( 20 - 25 ) be replaced. Such information may preferably be standardized. It is also preferable here to have a uniform database via standardized coverage areas and / or detection systems to be provided, which is obtained from a provider and regularly updated.

Alternatively, detection areas can be identified by a generally defined grid, with a communication participant notifying which of the defined grid areas are covered by his recognition device. In 1 With reference to the vehicle (B) such a grid is exemplified. It defines with a first grid area ( 40 ) a front environment, with a second grid area ( 41 ) a rear environment and with corresponding further grid areas ( 42 - 45 ) front left and right environments as well as back left and right environments, which essentially correspond to the definition of the detection areas ( 20 - 25 ) out 4 correspond. Alternatively, any other grid and / or any other subdivision of the environment may be chosen.

In the example off 1 Thus, the vehicle (B) could have communicated from the perspective of the vehicle (A) as a foreign communication participants that it only in its front environment ( 40 ) can perform its own acquisition. Further, based on the status messages, the vehicle (A) may know that the vehicle (B) is moving backwards, ie, in a direction in which it can not perform detection itself. It is further assumed that the vehicle (A) detects the child (L) as a relevant target object and recognizes its position and, if appropriate, movement. Based on the position information on the vehicle (B) and the child (L), the vehicle (A) can determine that the child (L) with its current position of the rear environment ( 41 ) and / or the rear right environment ( 45 ) of the vehicle (B) is assigned. It can also conclude that the anticipated progression of the child (L) of the backward environment ( 41 ) of the vehicle (B) is assigned. Accordingly, the vehicle (A) can determine that the child (L) is in a blind spot surrounding area (BE) of the vehicle (B) in which the child himself does not perform recognition. The vehicle (A) preferably sends as a self-communication participant a message with recognition results about the child (L), in order to specifically assess the state of knowledge of the system ( 1 ) and in particular the foreign communication subscriber (B) to improve.

From the substantially consistent consideration, the vehicle (A) can also communicate recognition results about the woman (K) with the stroller, as well as their positions the blind spot environment area (BE, 45 ) of the vehicle (B) is assigned.

According to an alternative or additional variant, blind spot surrounding areas (BE) of a foreign communication subscriber (B) can be based on recognized or notified obstacle objects (B). 11 . 13 . 15 ) be determined. The vehicle (A) can, for example, the hedge ( 11 ) as an object, whereby the position, height and width of the hedge ( 11 ) are known. Based on the known position of the vehicle (B), the descriptive data on the hedge ( 11 ) and the recognition results on the child (L) and the woman (K) with the stroller, the vehicle can determine that with respect to the vehicle (B) the hedge ( 11 ) represents an obstacle object that denies the vehicle (B) (at least momentarily) its own detection of the child (L) and the woman (K), ie that the child (L) and the woman (K) in a blind Spot environment area (BE) of the foreign communication party (B) are located. This determination is independent of whether the self-communication participant (A) the perception areas (AE) of the vehicle (B) are known.

Preferably, the two aforementioned variants for determining a blind spot surrounding area (BE) of a foreign communication subscriber (B) can be combined. It can also be provided that a self-communication subscriber (A) in the system ( 1 ) informs itself that one of its coverage areas ( 20 - 25 ) is restricted by an obstacle object. In other words, a self-communication participant (A) can detect and communicate own blind spot environment areas (BE).

Another measure, which can be used alone or in any combination with the aforementioned measures, provides that a self-communication participant (A) has potential collision situations between known relevant target objects (L, M) and known external communication participants (B, C) analyzed. For this purpose, in particular a momentary movement or a movement history for the anticipation of the further movement of the are evaluated. A self-communication subscriber (A) preferably sends messages with recognition results about such relevant target objects (L, M) to which a collision situation with a foreign communication subscriber (B, C) is anticipated. For example, the shipment may be made at higher frequencies or under the effect of other communications. In the example, the communicating party (A) may anticipate that the child (L) is moving on a path that overlaps temporally and spatially with the path of movement of the vehicle (B). Thus, an acute collision hazard is detected.

In the same way, the vehicle (C) could analyze the movement of the woman's pushchair (K) or the turning of the cyclist (M) against the movement of the vehicle (A) and anticipate the aforementioned collision situations. The vehicle (C) would thus preferably communicate recognition results about the stroller and the cyclist (M) and make it available to the vehicle (A). The preferred shipping is useful here, although the vehicle (A) itself has previously communicated recognition results on these relevant target objects, because a collision situation was anticipated.

In 5A By way of example, an image section is shown in which an adult and a child stand close together. Only the adult is fully visible. The child is partially covered by the adult. It may be difficult for a recognizer to discriminate the adult from the child, so that only the adult is detected as an object (X) having a size (H (X)) and a width (W (X)). The two real objects (child, adult) are mistakenly understood as a single object (X). Also, the above-mentioned tracking of objects within the sensor data may not avoid such a misconception. Accordingly, it may be that only the recognition results about the object (X) in the system ( 1 ). It is also possible that even through multiple communication participants the same inadequate interpretation of the traffic environment occurs, which may lead to an increased manifestation of the misconception in the system ( 1 ) can lead.

The representation of 5B shows an image capture at a later time. Here, the adult within the captured image has moved away from the child, for example because of an actual distance from the child, or because the image comes from a different perspective. Accordingly, a separate identification of an object (X '= adult) and an object (Y = child) will now take place, with different descriptive data for these objects (X', Y) resulting from the recognition results. In particular, it can be stated that the object (Y) has a different size (H (Y)) and / or a different position and / or a different classification than the object (X ') or the preceding object (X).

According to a further measure, which can be used alone or in any combination with the above-mentioned measures, a self-communication participant (A) can determine an association between a detected object (X, X ', Y) and its descriptive data, and in particular analyze its classification against the previously recognized recognition results (own and received). In particular, the new recognition results may indicate that a new association occurs between an identified object and its classification as one or more relevant target objects. The changed classification may result from the descriptive information or from the fact that suddenly one of two or more previously known objects of known classification is dropped. Furthermore, descriptive data about a known object may change abruptly while the above-mentioned confidence threshold is reached or exceeded. The self-communication device can deduce from such deviations that at least one erroneous view exists within the recognition results communicated in the system.

From such a knowledge different reactions can be derived. On the one hand, the self-communication subscriber (A) can preferably send a message with the new recognition results in order to connect the other communication subscribers (B, C) in the system ( 1 ) with better or more up-to-date descriptive information. Alternatively or additionally, in the system ( 1 ) a warning is sent that there is a misconception regarding the one or more known objects. Such a warning may cause the foreign communication participants (B, C), who perform their own detection for these objects, also to carry out a consistency check and possibly to send additional new recognition results preferably.

Modifications of the invention are possible in various ways. In particular, the features shown or described with reference to the exemplary embodiments can be combined with one another in any desired manner, replaced with one another, supplemented or omitted. All of the above measures can be used in any combination and supplemented if necessary by further measures. In addition to the above-mentioned systems protection of vulnerable road users can be supported for example by targeted illumination with alignable vehicle headlights or by showing position-related markings in a head-up display. A detection device may alternatively or in addition to the above sensors have a thermal imaging camera. Furthermore, a recognition device can merge different sensor data or intermediate results of the object recognition with one another in order to recognize vulnerable road users. Thus, for example, sensor data of a camera, a radar, and a thermal imaging camera can be evaluated simultaneously, overlapping or successively become. From the image data of the camera, in combination with the thermal image data, an object can be well classified as a pedestrian or an animal, while using the radar data, a particularly accurate position determination is possible. Alternatively or additionally, the position determination can be derived by the distance information acquired by a 3D camera from stereo images. Similarly, any other sensor data merging or merging of intermediate results of the object recognition is possible. LIST OF REFERENCE NUMBERS A First vehicle / owner First vehicle / Own communication stations communication participant B Second vehicle / third party Second vehicle / communication stations Foreign communication participant C Third vehicle / third party Third vehicle / Foreign communication stations communication participant K First road user First road user / First / first relevant target relevant target object Object / woman with / Woman with buggy stroller L second Second road user / Road users / Second relevant target second relevant target object / child with Object / child with ball ball M third Third road user / third Road users / relevant target object third relevant target / bicyclist Object / cyclist N fourth Forth road user / Road users / Sitting man Sitting man RU vulnerable Vulnerable road user road users CU communication stations Communication in the system participant within the system AE perceptive Perception surroundings surrounding area area BE Blind spot blind spot surroundings surrounding area area ID Object Reference Object reference number Pos. position position H Height / size Height / size W width width To you. direction Direction v speed speed Class object class Object class Ta Age of information / Age of information / Time since update Time since updating X Known object Known Object X ' Replacement Property Substitute object Y Additional object Additional Object 1 Recognition and Detection and communication system communication system 10 traffic environment Traffic environment 11 Hedge / First obstacle Hedge / First obstacle object object 12 lawn Grass field 13 Tree / second obstacle Tree / Second obstacle object object 14 Park bench Park bench 15 House / third obstacle House / Third obstacle object object 16 priority road Priority road 17 side street byroad 20 Front detection range Front detection area 21 Rear detection area Back detection area 22 Left side Left lateral detection detection range area 23 Right side Right lateral detection detection range area 24 Left rearview mirror Left rearview mirror detection range detection area 25 Right rearview mirror Right rearview mirror detection range detection area 30 Ordered data structure / Organized data object list structure / object list 30 ' Ordered data structure / Organized data List over structure / list of communication stations communication participants 31 recognition results Detection results 32 subscriber information Participants information 40 Front environment Front surrounding 41 Rear environment Rear surrounding 42 Front left environment Front left surrounding 43 Front right environment Front right surrounding 44 Rear left environment Rear left surrounding 45 Rear right environment Rear right surrounding

Claims (13)

Recognition and communication method for execution in a self-communication subscriber (A) and as part of a distributed recognition and communication system ( 1 ) for detecting and protecting vulnerable road users (RU, K, L, M, N) by communicating recognition results via identified objects between two or more communication users (CU, A, B, C) of the system ( 1 ), wherein in the self-communication subscriber (A) by means of a recognition device vulnerable road users (RU, K, L, M, N) in its recognition environment ( 20 - 25 ) and identified as objects and recognition results ( 31 ) about identified objects in an ordered data structure ( 30 ) and to external communication users (CU, B, C) in the system ( 1 ), characterized in that in the self-communication subscriber (A) the sending of messages with recognition results (A) 31 ) based on an analysis of the own recognition results against the stored information ( 31 ) about known objects and known communication participants (CU, 30 ' ), in particular by favoring and / or limiting the sending of communications. A method according to claim 1, characterized in that in the self-communication participant (A) within the sensor data of the recognition device objects (K, L, M, N) are tracked, and wherein these objects (K, L, M, N) classified to that effect whether it is target objects relevant to the protection of vulnerable road users. A method according to claim 1 or 2, characterized in that the self-communication participant (A) sends only messages with recognition results on relevant target objects. Method according to claim 1, 2 or 3, characterized in that the notification of newly determined recognition results ( 31 ) is controlled via known objects as a function of the age (Ta) of the last exchanged information. Method according to one of the preceding claims, characterized in that the self-communication participant (A) preferably recognition results ( 31 ) reports on such relevant target objects (K, L, M), which correspond to a blind spot surrounding area (BE, 41 . 44 . 45 ) of a foreign communication subscriber (B) are assigned, in which by this foreign communication participant (B) itself no detection is possible or possible. Method according to one of the preceding claims, characterized in that the self-communication participant (A) sends a message about in which perception environment areas (AE) he himself can perform a detection or perform, and / or that the self-communication participant ( A) receives a message from a foreign communication subscriber (B) about which perceptual surrounding areas (AE, 40 ) this can perform or perform a detection. Method according to one of the preceding claims, characterized in that a blind spot surrounding area (BE, 41 . 44 . 45 ) is determined from the communicated information about the perceptual environment area (AE) of the foreign communication subscriber (B). Method according to one of the preceding claims, characterized in that the self-communication participant (A) has a blind spot surrounding area (BE, 41 . 44 . 45 ) of a foreign communication subscriber (B) on the basis of detected or notified obstacle objects ( 11 . 13 . 15 ) in the environment of this external communication subscriber (B) determined. Method according to one of the preceding claims, characterized in that the self-communication participant (A) analyzes potential collision situations between relevant target objects (L, M) and foreign communication participants and recognizes recognition results ( 31 ) preferably sent via such relevant target objects (L, M), to which a collision situation is anticipated. Method according to one of the preceding claims, characterized in that a message with new recognition results ( 31 ) is preferably sent, if from the new recognition results ( 31 ) yields a new association between a detected object (K, L, M, N) and its classification (Class) as one or more relevant target objects. Security device for detecting and protecting vulnerable road users (RU) in conjunction with a distributed detection and communication system ( 1 ), wherein the security device has an arithmetic unit for carrying out a recognition and communication method and has access to a communication means and to a recognition means, characterized in that the security device is designed to detect vulnerable road users (RU) on the basis of the sensor data of the recognition means to carry out and to carry out a method according to one of the preceding claims. Safety device according to claim 11, characterized in that the safety device is part of a motor vehicle. Software product for providing a recognition and communication method as part of a distributed recognition and communication system ( 1 ), whereby the system ( 1 ) for recognizing and protecting vulnerable road users (RU, K, L, M, N) with notification of recognition results ( 31 ) over identified objects (K, L, M, N) between two or more communication participants (CU, A, B, C) of the system ( 1 ), characterized in that the software product comprises programmed instructions by which a method according to one of claims 1 to 10 is executed when executed on a computer.

Images