US-12627782-B2 - Method for deriving at least one item of information from images of a stereo camera

US12627782B2US 12627782 B2US12627782 B2US 12627782B2US-12627782-B2

Abstract

A method for deriving at least one item of information from images of a stereo camera. A resource-saving and at the same time accurate derivation of information from the images is achieved by alternately using and processing the images with methods of monocular image processing and deriving at least one item of information from the results. A motor vehicle comprising a stereo camera and comprising a control device which carries out the method, are also described.

Inventors

Moritz Michael Knorr

Assignees

ROBERT BOSCH GMBH

Dates

Publication Date: 20260512
Application Date: 20230111
Priority Date: 20220118

Claims (10)

1 . A method for deriving at least one item of information from images of a stereo camera including a first camera and a second camera, the method comprising the following steps: capturing, at successive points in time, first images from the first camera, and second images from the second camera; in a monocular step, processing only a single new image at each respective point in time, wherein the single new image is selected alternately from the captured first images and second images to generate a time-alternating sequence of processed images using methods of monocular image processing, and outputting results of the monocular step including the time-alternating sequence of processed images; and in a derivation step, deriving at least one item of information from the results of the monocular step including the time-alternating sequence of processed images.
2 . The method according to claim 1 , wherein in the monocular step, an optical flow of the single new image processed at each respective point in time, that is selected alternately from the captured first images and second images, is determined and output as a result.
3 . The method according to claim 1 , wherein, in the monocular step, at least one image feature of the single new image processed at each respective point in time, that is selected alternately from the captured first images and second images, is tracked over time and output as a result.
4 . The method according to claim 1 , wherein monocular deep learning methods are used in the monocular step.
5 . The method according to claim 1 , wherein at least one monocular method for deriving at least one of the at least one item of information is used in the derivation step.
6 . The method according to claim 1 , wherein, in the derivation step, a movement of the stereo camera itself is derived as information.
7 . The method according to claim 1 , wherein the derivation step includes warping.
8 . The method according to claim 1 , wherein the derivation step derives an estimation of depths and/or detection of self-moving objects, in an environment of the stereo camera as information.
9 . A non-transitory computer-readable storage medium on which is stored a computer program for deriving at least one item of information from images of a stereo camera including a first camera and a second camera, the computer program, when executed by a computer, causing the computer to perform the following steps: capturing, at successive points in time, first images from the first camera, and second images from the second camera; in a monocular step, processing only a single new image at each respective point in time, wherein the single new image is selected alternately from the captured first images and second images to generate a time-alternating sequence of processed images using methods of monocular image processing, and outputting results of the monocular step including the time-alternating sequence of processed images; and in a derivation step, deriving at least one item of information from the results of the monocular step including the time-alternating sequence of processed images.
10 . A motor vehicle, comprising: a stereo camera including a first camera which captures first images during operation, and a second camera which captures second images during operation, wherein the first images and the second images are captured at successive points in time; and a control device configured to derive at least one item of information from the captured first and second images of the stereo camera, the control device configured to: in a monocular step, process only a single new image at each respective point in time, wherein the single new image is selected alternately from the captured first images and second images to generate a time-alternating sequence of processed images using methods of monocular image processing, and output results of the monocular step including the time-alternating sequence of processed images; and in a derivation step, derive at least one item of information from the results of the monocular step including the time-alternating sequence of processed images.

Description

CROSS REFERENCE The present application claims the benefit under 35 U.S.C. § 119 of German Patent Application No. DE 10 2022 200 542.2 filed on Jan. 18, 2022, which is expressly incorporated herein by reference in its entirety. FIELD The present invention relates to a method for deriving at least one item of information from images of a stereo camera. The invention also relates to a motor vehicle comprising a stereo camera and a control device for carrying out the method. BACKGROUND INFORMATION In many applications, it is desirable to obtain information about the current environment of the application. One such application is a motor vehicle, for example. For this purpose, at least one camera which captures images of the environment of the application during operation can be used. Information is then derived from the images. Depth estimation in particular is of special interest here, because it is important for driving assistance, for example, and/or for at least partially autonomous driving of motor vehicles. The use of a single camera for this purpose is available in the related at. In monocular systems, information is derived from the images of the camera by processing the successive images of the camera using methods for monocular image processing. The use of a stereo camera comprising two cameras is available for this purpose as well, wherein the respective camera captures associated images. In the corresponding stereo methods, the new images of both cameras are used and processed at each point in time. In particular the stereo disparity between simultaneously captured images and/or an optical flow between successively captured images is determined. This allows improved depth estimation and improved detection of self-moving objects. Corresponding systems and methods are described, for example In German Patent Application No. DE 10 2012 014 994 A1 and European Patent Application No. EP 1 580 092 A2. The monocular systems and methods in particular have the disadvantage that a depth estimation for self-moving objects is not possible or is subject to high error rates. The stereo systems and methods in particular have the disadvantage that the processing of the images involves a high consumption of resources. SUMMARY The present invention has an object of providing improved or at least different embodiments for a method for deriving at least one item of information from images of a stereo camera and for a motor vehicle comprising such a stereo camera, which in particular may eliminate disadvantages from the related art. This object may be achieved according to features of the present invention. Advantageous embodiments of the present invention are disclosed herein. The present invention includes alternately using and processing the images captured by cameras of a stereo camera at successive points in time in order to derive information from the images. Therefore, at each successive point in time a new image has to be taken into account. This means that methods designed for monocular processing are applied to an alternating image sequence of a stereo camera. Compared to the stereo methods available in the related art, this results in a considerable reduction of required resources. This in particular saves bandwidths and/or memory and/or the consumption of electrical energy. This in particular avoids the calculations of disparities between simultaneously captured stereo images and/or calculations of the change over time between pairs of successive images which are usually necessary in stereo methods. Compared to stereo methods, the solution according to the invention further offers the advantage that two different methods, in particular algorithms, are not required. Compared to monocular systems available in the related art, there can be no linear movement of the cameras at successive points in time. As a result, improved information can be derived from the environment, in particular from objects moving in parallel. Compared to monocular systems, there is furthermore the advantage that a pseudo movement can be determined even when stationary, which can be advantageous for a depth estimation. Another advantage over monocular systems is that, due to the usually known distance between the two cameras of the stereo camera, there is a fixed reference value that, for example, makes it possible to eliminate or at least reduce the problem of scale ambiguity of monocular systems. In accordance with an example embodiment of the present invention, in a step which hereinafter is also referred to as a monocular step, the images of the first camera and the second camera are used and processed with methods of monocular image processing alternately in terms of time and output as results in order to derive at least one item of information from images of a stereo camera comprising a first camera and a second camera. In a step, which hereinafter is also referred to as a derivation step, at least one item of information is deriv