Method and device for identifying an object in an image of a camera

The invention relates to a method for identifying a Object in a camera image.

The term "augmented reality application" are camera applications known in which, for example, text information or modeled artificial objects in a filmed real environment at respective associated real objects are also displayed. For display of additional Information in a current camera image, make the necessary objects associated with the real environment quickly and be reliably identified. known for this purpose are Methods of digital image processing used, the particularly in edge-based and texture-based image analysis methods are divided.

While the former designed for detection of edges of an object are, are the latter to detect a texture or a Pattern of the object optimized. For example, there is a Texture analysis using a Fourier analysis in the frequency domain. A reliable object detection and object identification is with the methods of conventional digital image processing only to ensure when used in each case Image analysis methods to the respective object-specific Features tuned.

For the simultaneous detection of a plurality of objects a recorded using a camera real environment has gained an image information from the environment and both are examined gradually as well as feature-dependent. Below then is a suitable image analysis methods selected for object recognition. Such investigation methods generally cause a high amount of calculation and for this reason commercial in a variety Camera applications not in use. This applies in particular For single or video cameras with wireless communication terminal, Augmented Reality applications include.

Conventional methods for augmented reality applications for this reason using predefined markers performed in a filmed or recorded around are installed. The image analysis methods used are tuned to geometric properties of the markers. This makes it possible that a picture information according to the specific Properties of the markers are analyzed. disadvantageous is the installation to be carried out only with great effort the marker.

In addition, the above image analysis methods must at a rapidly occurring change of the recorded Environment are initialized again, as in this case a product obtained result is falsified or lost.

It is therefore an object of the present invention, a fast to be implemented, reliable method for object identification specify.

This object is achieved by a method having the features of Patent claim 1. Advantageous developments are specified in the subclaims.

In the present process, for example by filmed camera motion an environment or recorded. In the recorded ambient significant objects determines which one hand respective position information and on the other hand to a suitable object identification Image analysis methods are specifically assigned.

If now at a later time, a section of the real Environment using the camera as the current camera image recorded, so the current image with the recorded is Environment compared.

With the help of a machine vision-based motion estimator, consisting of augmented reality gaming applications on wireless communication terminals is known, the position of the environment section is determined in the recorded environment. For the identification of objects of the environment section is the image analysis method used, which its associated Position in the recorded environment according to the determined Position the ambient cutout closest lies.

The inventive method thus increases the probability a faultless object identification, since the selected image identification method depending on position becomes.

By continuously comparing a once recorded Environment with the current camera images will be compared the prior art, saves computing time.

Fast camera movements or disruptions in the current camera image have in the selection of the object identification process not interfere, since the respective positioning is rapidly repeatable.

The inventive method is an identification security individual objects increases without markers having to place objects around.

The invention based on a drawing is closer explained.

Here 1 shows a section of a pedestrian street in the inner city Area as recorded environment. An unhidden in the section of the pedestrian area radio communication terminal M has an integrated camera and a processor unit. By moving the Camera in an angular range, the pedestrian environment as recorded and stored.

In the recorded ambient example be as significant objects a steeple KT and a flag F detected. The steeple KT is a first position x0, y0 and assigned an edge-based image analysis techniques, while the flag F to a second position x1, y1 and a texture-based Image analysis method is assigned.

Displays a current camera image after a few screen changes to the steeple KT, as the current image with the will recorded around compared. As the current position Camera image, with respect to the modeled environment x0 the first position y0 determined. Since this position x0, y0 is an edge-based image analysis methods associated is this selected and identification of the church tower KT used.

This is for example at a augmented reality application allows text or object information that the Steeple KT associated, fixed in the current camera image see the church tower KT.

Advantageously, in the selection of the image analysis method an environment likely considered. For the first Position x0, y0 and their surroundings is preferred an edge-based image analysis method for object recognition used, while for the second position x1, y1 and their surroundings preferably a texture-based image analysis methods used for object recognition or -Identification becomes. The feature analysis is thus on the respective image content optimally tuned.

Advantageously carried out a dynamic, adaptive adjustment of Image analysis method to the objects of the current camera image. This has the consequence that with a change of the camera image from the first position x0, y0 to the second position x1, y1, the image analysis of an edge-based analysis method gradually to a texture-based analysis method is changed and thus adaptively adjusted to the environment.

The relevant data for the positions x0, y0 above and x1, y1 and the respective associated image analysis method advantageously in a look-up table of the radio communication terminal stored and are thus available at any time quickly.

Above it was a current camera image with the recorded Environment compared. Alternatively, it is also possible one of the recorded environment appropriate model form and modeled the current camera Image compare environment. The environment might, for example, are modeled as mathematical model in three dimensions. For the Inventive process steps is thus in place of recorded around the replicated as a model environment used.