DE102009009904A1 - Object i.e. road sign, identifying method, involves adding images and classifiers, which contain stored samples and sequences, as another set of samples, if sequences correspond to another set of sequences - Google Patents

Abstract

The method involves detecting an object (O) by a detection unit (2) integrated in a vehicle. The detected object is compared with stored samples (M1-Mn). Significances (BB1-BBn) of images of the object are determined from a driving condition (F) of the vehicle, and are compared with significances (B1-Bn) of pre-defined samples, which are stored in a memory unit (3). The images and classifiers (K1-Kn), which contain the stored samples and the sequences (B1-Bn), is added as another set of samples (mM1-mMn), if the sequences (BB1-BBn) correspond to the sequences (B1-Bn).

Description

The Invention relates to a method for identifying objects, in particular of traffic signs, wherein the objects by means of a recorded in a vehicle integrated detection unit and the captured objects for identification with stored patterns be compared.

It is well known that for automatic detection of Objects, for example for automatic recognition of traffic signs in the Road traffic, the objects by means of an optical detection unit recorded and compared with patterns. For this purpose will be creating the patterns using different methods and devices, the patterns are assigned a meaning and compared with the collected objects are stored the patterns.

A method for identifying patterns that are stored as stored information is known from US 5,156,237 DE 101 10 275 A1 known, wherein labeling proposals are generated, which are checked by an editor. The tagging suggestions are automatically generated in a computer system and the processor is presented with the pattern and the tagging suggestion for the pattern by means of a computer program, whereby the operator can accept the tagging suggestion to assign its meaning to the pattern. Alternatively, the engineer can also reject or change the labeling proposal.

Of the Invention is based on the object, a relation to the Prior art improved method for the identification of Specify objects by means of which an improved identification at the same time reduced effort is feasible.

The Invention is achieved by a method which the Having specified in claim 1 features.

advantageous Embodiments of the invention are the subject of the dependent claims.

at the method of identification according to the invention objects, in particular traffic signs, become the objects detected by means of a vehicle integrated in a detection unit and the detected objects for identification with stored patterns compared.

According to the invention a picture meaning of an image of a detected object from a Driving condition of the vehicle determined and with in a storage unit the vehicle deposited meanings of predefined patterns compared. The image is then a classifier, which at least one contains a predefined pattern and a meaning of an object, added as another pattern when the picture meaning corresponds to the meaning contained in the classifier.

by virtue of this inventive design of the method results in a particularly advantageous manner that the objects also then clearly identifiable, if their real appearance from a given ideal, in particular from virtual characters formed appearance differs because the predefined classifier by adding the other patterns with the same Meaning "adaptive" and thus updatable is. The determination of the image meaning of the detected object, for example the determination of a regulatory content of a traffic sign based on the driving condition of the vehicle provides reliable results and allows automatic and fast updating of the classifier, being a manual labor-intensive and time-consuming Expenses for updating the classifier can be omitted.

embodiments The invention will be described below with reference to a drawing explained.

there demonstrate:

1 schematically a block diagram of an apparatus for the detection and classification of objects.

The only 1 shows a device 1 for recognizing and classifying objects O, wherein the objects O are, in particular, traffic signs.

The device 1 is arranged in particular in a vehicle, not shown, wherein on the basis of the detected objects O preferably one or more driver assistance systems are controlled. In this case, for example, a driver can automatically be made visually and / or acoustically aware of a control content contained on a traffic sign, such as a speed limit, and / or automatic interventions in the drive of the vehicle can be listed.

For detecting the objects O, the device comprises 1 at least one optical detection unit 2 , In particular a camera, by means of which the objects O are detected by means of images A during operation of the vehicle in a street scene.

Furthermore, the device comprises 1 a storage unit 3 , in which several classifiers K ₁ to K _{n are} deposited. In the production of Vor direction 1 the classifiers K ₁ to K _{n are} predefined, each of the classifiers K ₁ to K _n having at least one pattern M ₁ to M _{n of} an object O and a meaning B ₁ to B _n assigned to the respective pattern M ₁ to M _{n of} the object O contains. The meaning B ₁ to B _{n of} the pattern M ₁ to M _{n of} the object O can, for. B. in a training of the object O as a road sign to be a regulatory content of a road sign.

The classifiers K ₁ to K _n serve to identify the detected objects O by means of one in the device 1 arranged data processing unit 4 , To identify the objects O, the data processing unit 4 by means of the detection unit 2 captured images A and supplied by the data processing unit 4 with those in the storage unit 3 deposited classifiers K ₁ to K _n , ie compared with the patterns M ₁ to M _n contained in the classifiers K ₁ to K _n . In one of a captured image A with one of the pattern M ₁ to M _n match, each corresponding to the pattern M ₁ to M _n corresponding meaning is B ₁ to B _n the object O displayed on the screen A is assigned, that the driver assistance system based on the Meaning B ₁ to B _{n of} the object O is controlled.

For example, as object O, a traffic sign is detected whose control content is a speed limit to a speed of 70 km / h. The image A of the traffic sign becomes the data processing unit 4 supplied with this in the storage unit 3 deposited patterns M ₁ to M _{n of} the classifiers K ₁ to K _n . If a match one of the patterns M ₁ to M _n, which are stored, for example as a virtual character, with the image A, the traffic sign, the _n associated importance to the matching pattern M ₁ to M B ₁ to B _n, namely the speed limit of 70 km / h, assigned.

Due to different environmental properties of the vehicle, in particular weather conditions, due to driving conditions F of the vehicle, for example its speed, and / or due to different representations of the objects O with the same meaning B ₁ to B _n is the deposit of a pattern M ₁ to M _n per classifier K ₁ to K _{n are} not sufficient to safely assign detected objects O the correct meaning B ₁ to B _n .

Since a consideration of all possible patterns M ₁ to M _n in the generation of the predefined classifiers K ₁ to K _{n is} very complicated, the method according to the invention provides that the captured images B of the objects O, which depend on the stored patterns M ₁ to M _n a classifier K ₁ to K _n but differ whose image significance BB ₁ BB _n a meaning B ₁ to B of the patterns M ₁ to M _n corresponding to this classifier K ₁ to K _n as a further pattern _n mM ₁ mM _n assigned and in the storage unit 3 so that a classifier K ₁ to K _{n can have} a plurality of patterns M ₁ to M _n and mM ₁ to mM _n with the same meaning B ₁ to B _n .

In order to determine whether the object O detected by means of the image B has the same image meaning BB ₁ to BB _n as one of the stored classifiers K ₁ to K _n , the image significance BB ₁ to BB _{n of} the object O becomes a driving state F of the vehicle determined during the detection of the object O. For this purpose, a driver profile of a behavior of the driver and other not-shown vehicle sensor system is created.

Thus, for example, a traffic sign is detected as object O, but due to the reasons already described above does not correspond to one of the stored patterns M ₁ to M _{n of} one of the classifiers K ₁ to K _n or is only identified incompletely. If the driver of the vehicle perceives the traffic sign, which in turn is, for example, the speed limit of 70 km / h, he generally carries out actions which change the driving state F of the vehicle and which can be detected by means of the vehicle sensor system.

Thus, the driver brakes the vehicle when the traffic sign z. B. from 70 km / h. Due to the detected deceleration process to 70 km / h during the detection of the traffic sign is determined as image meaning BB ₁ to BB _n the speed limit to 70 km / h, so that the image A of the object O the classifier K ₁ to K _n as a new pattern mM ₁ to mM _n , which as meaning B ₁ to B _n also includes the speed limit to 70 km / h.

Consequently it is possible in a particularly advantageous manner that z. B. in an internationally deviant representation of traffic signs whose regulatory content is reliably detected.

The image significance BB ₁ to BB _{n of} the detected object O can additionally be derived from a behavior, in particular a movement E, of another vehicle located in front of the vehicle, the movement E of the further vehicle being detected by the detection unit 2 and / or further, not shown sensors is detected. So z. B. detected during the insufficient detection of the traffic sign with the speed limit to 70 km / h, a deceleration of the other vehicle in passing the traffic sign to this speed. The driving behavior (acceleration, braking, Lenkbewe tion, declarations of intent, ...) is preferably carried out by means already existing in the vehicle assistance systems such. As a distance warning or a vehicle to vehicle communication determined. An additional integration of further assistance systems is furthermore advantageous, so that, for example, information from the traffic infrastructure can also be detected, for example by means of a lane assistant. In addition, it is possible to include other external sources of information (traffic control systems, radio services, ...) in determining the image meaning BB ₁ to BB _n .

Before the image A of the detected object O is preferably automatically added to one of the classifiers K ₁ to K _n , the determined image significance BB ₁ to BB _n can be automatically generated by the data processing unit 4 be checked, with z. For example, the driving state F of the own vehicle is compared with the behavior of the at least one preceding vehicle, so that in case of coincidence the image A of the object O is added as another pattern mM ₁ to mM _{n to} one of the classifiers K ₁ to K _n .

Alternatively or additionally, the image significance BB ₁ to BB _{n can} also be checked manually. Thus, for example, the driver of the vehicle, the captured image A of the object O and a picture meaning BB ₁ to BB _{n are} proposed, this of adding the captured image B as another pattern mM ₁ to mM _n to one of the classifiers K ₁ to K _n confirm or reject.

Furthermore, upon detection of an object O whose image meaning BB ₁ to BB _{n can} not be assigned to any of the classifiers K ₁ to K _n , new classifiers K ₁ to K _n can also be generated, wherein the generation takes place automatically and their final use and deposit are preferably based on an automatic and / or manual check confirmed or rejected.

It is also possible to read out the updated classifiers K ₁ to K _n, in particular during workshop stays of the vehicle, and these optionally with or without additional checking for correctness for updating or as a so-called update for similar devices as the illustrated device 1 use. Thus, it is also possible to the vehicle with the device 1 add updated classifiers K ₁ to K _{n of} other vehicles.

1

contraption

2

acquisition unit

3

storage unit

4

Data processing unit

A

image

B ₁ to B _n

importance

BB ₁ to BB _n

image meaning

e

Move

F

driving condition

K ₁ to K _n

classifier

M ₁ to M _n

template

mM ₁ to mM _n

template

O

object

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited patent literature

• DE 10110275 A1 [0003]

Claims (4)

Method for identifying objects (O), in particular traffic signs, wherein the objects (O) are detected by means of a detection unit (integrated in a vehicle) ( 2 ) and the detected objects (O) for identification with stored patterns (M ₁ to M _n ) are compared, characterized in that an image meaning (BB ₁ to BB _n ) of an image (A) of a detected object (O) from a Fahrzustand (F) of the vehicle is determined and with in a storage unit ( 3 ) of the vehicle (B ₁ to B _n ) predefined patterns (M ₁ to M _n ) is compared, the image (A) then a classifier (K ₁ to K _n ), which at least one predefined pattern (M ₁ to M _n ) and a meaning (B ₁ to B _n ) of an object (O) is added as a further pattern (mM ₁ to mM _n ) when the image meaning (BB ₁ to BB _n ) in the classifier (K ₁ meaning (B ₁ to K _n) contained to B _n). Method according to Claim 1, characterized in that the image significance (BB ₁ to BB _n ) is additionally derived from a detected movement (E) of at least one further vehicle. A method according to claim 1 or 2, characterized in that the image meaning (BB ₁ to BB _n ) before adding the image (A) of the object (O) to the classifier (K ₁ to K _n ) automatically and / or manually for correctness is checked. Method according to one of the preceding claims, characterized in that the image (A) of the detected object (O) the classifier (K ₁ to K _n ) as another pattern (mM ₁ to mM _n ) immediately after detection and determination and / or image significance checking (B ₁ to B _n ) is automatically assigned.