Method for Identifying Soiling On a Transparent Screen

Abstract

Camera for identifying soiling on a protective screen. The camera is focused on a scene behind the protective screen and can be used to identify soiling and for other applications, e.g. to identify driving lanes and/or to identify objects. Soiling is identified merely by evaluating frames recorded in succession, artificial reference frames or reference objects are not required. A prerequisite for a successful measurement is that the relative speed vrel of the camera and of at least one recorded object in the surroundings is not equal to zero and its trajectory is predicted in the depiction. By comparing the relevant frame sections, possible soiling on subregions of the protective screen is identified.

Description

The present invention relates to a method for identifying soiling on a protective screen. This invention is used e.g. in motor vehicles. Camera sensors are used here to an increased extent to record the surroundings, in particular to identify driving lanes and/or to identify objects, with a view to the driving direction or to the backward region. The camera sensors are usually attached in the interior of the vehicle behind a windscreen. A faultless function of the sensor is ensured only if the view is not masked by foreign objects on the screen.

Numerous methods for measuring soiling of screens e.g. in motor vehicles are known. In DE 10230200 a detector is described to identify objects on a surface of a transparent element. In this case, the detector is at least approximately focused on the surface and for identifying soiling the contrast distribution of the recorded frames is evaluated. A further method to identify objects located on a windscreen is shown in DE 19749331. For this purpose, a sensor array is focused on the exterior of the windscreen. An analysis of the spatial frequencies in the recorded frame provides information on a soiling of the windscreen. Due to the focusing of the camera on the surface both arrangements cannot be used for other camera-based applications, e.g. for coverage of the surroundings. In DE 19700665 a camera-based sensor is described, which detects foreign materials on a transparent screen. For this purpose the frame of a firmly installed light emitting reference object is recorded through the screen and is compared with a reference frame. It is disadvantageous with this arrangement that a reference object is needed to measure soiling.

It is the object of the invention to economically identify soiling of a protective screen in the field of view of a camera, which is used also for other applications in particular for coverage of the surroundings.

This object is achieved in accordance with a method according to patent claim 1. The dependent patent claims show advantageous forms of embodiment and further developments of the invention.

In accordance with the invention this object is achieved by a passive camera-based approach. For this purpose, a camera is provided, which is focused on the surroundings behind a transparent screen. It is advantageous with this arrangement that the camera can be used to identify soiling and to cover the surroundings. In a given cycle frames are read out from the camera, e.g. 25 frames/s. The recorded frames are evaluated in terms of soiling on the protective screen. Thus, no further artificial reference frames or reference objects are needed, so that for the user no further costs and/or adjustment expenditure arise here. Soiling can only be identified according to the presented method, if the relative velocity v_rel of the camera and of at least one object recorded in the surroundings is not equal to zero. This is ensured in an application e.g. in the motor vehicle by intrinsic speed of the vehicle V_vehide>0 m/s. Thus, the relative velocity towards static objects in the vehicle surroundings is v_rel>0 m/s. Objects with v_rel>0 m/s continuously change their position in relation to the camera and thus also their position in the frame. The calculated process of this movement in the frame is called in the following trajectory. This means that the same object is depicted chronologically offset in different frame sections. Now if a frame section is totally or partly masked by soiling, then the object is changed or not depicted at all. By comparing the frame sections, possible soiling on subregions of the protective screen is identified. In order to select the frame sections suitable for the comparison, the trajectory of at least one object is predicted with v_rel>0.

In an advantageous embodiment of the invention measuring windows are set along the predicted trajectory of objects, so that the same object is depicted successively in the different measuring windows with a suitable read-out frequency and/or a suitable number of averagings of a measurement. The same applies for several similar objects, in particular for road markings, along a predicted trajectory. The latter, too, are visible in the measuring windows along their trajectory with a suitable read-out frequency and/or a suitable number of averagings of a measurement. If the illustration range of a measuring window is masked by soiling, the object is changed or not recorded at all in this measuring window. From the comparison of the object illustrations in corresponding measuring windows soiling in subregions of the protective screen is identified. Measuring windows along a predicted trajectory are called here corresponding measuring windows, in which an object illustration is expected according to the prediction.

In an advantageous arrangement of the invention in a measuring window the contrast is evaluated and/or the average value of the brightness is determined. The average value is not always sufficient to identify soiling. Shadows e.g. of vehicles, trees, buildings etc. can result in a local change of the brightness. However, if the average value and the standard deviation of the brightness are taken into consideration, it is possible to make a conclusion, since a shadow in front of the camera causes a local change of the average value, however, it does not lead to a substantial change of the standard deviation within the measuring window. The gray tone differences can be evaluated in addition to the other values or exclusively. If the difference of the determined values in corresponding measuring windows exceeds a threshold value, soiling is identified.

In case of a linear proper movement of the camera, the trajectories of static objects within the illustration range of the camera run along straight lines, which proceed radially from the vanishing point of the frame. In a preferred form of embodiment measuring windows are set along the cited straight lines. In case of a non-linear proper movement of the camera or of a non-linear movement to be expected, in an advantageous embodiment of the invention the trajectory of objects is calculated based on the characteristics of the surroundings and/or the instantaneous camera movement and/or the mounting position of the camera. If, for example, the camera is used to measure soiling in a motor vehicle, an object trajectory, in particular that of a static object, can be calculated from the estimated driving lane parameters such as curvature and offset, the mounting position of the camera, its height, angle of pitching, yawing and roll, as well as from the intrinsic camera parameters, the focal width, main point, etc. If there are no driving lane parameters available, as the system due to soiling or lacking road markings is not in the position to estimate a driving lane, the curvature can be determined on the basis of the steering angle. The measuring windows are arranged along the predicted trajectory.

An advantageous embodiment of the invention is a motor vehicle with a camera for observing the surroundings, e.g. to identify driving lanes and/or objects with a view to the driving direction or to the backward region, and to measure soiling.

Further advantages and characteristics of the invention are described in detail by way of examples on the basis of one example of embodiment and of two illustrations, in which

FIG. 1: shows a determination of the column c of a measuring window placed in a line y with lateral offset x.

FIG. 2: shows a determination of the distance of a point d to the camera.

The example of embodiment describes a motor vehicle with a camera to identify; soiling on a protective screen. The camera is focused on the surroundings in front of the vehicle and not on the protective screen and is used for coverage of the surroundings. The vehicle moves in a temporal average on a straight line. Static objects in the surroundings thus move within the illustration range of the camera on straight lines, which proceed radially from the vanishing point of the frame. Naturally, this applies also to the median strip and to the side limits of a road, the course of the roadway and oncoming vehicles. If measuring windows are positioned along the radially running straight lines, under ideal conditions, i.e. no damping or soiling, in the corresponding measuring windows, e.g. in the illustration range of the median strip or the roadway, it is to be reckoned with identical contrasts, average values, standard deviations and gray tone differences. If disturbances (soiling) are in the close range of the camera, they can be identified on the basis of local discontinuities. In the simplest case a disturbance can be identified on the basis of the differences between the measured values averaged along the radials and the measured values chronologically averaged in the measuring windows. If the difference exceeds a threshold it can be assumed that the associated measuring window is masked by soiling. The strength of the discontinuity depends on the roadway textures, the degree of soiling, the lighting, as well as on the size of the measuring windows.

However, also non-static objects with a relative velocity not equal to zero, e.g. oncoming vehicles can be used to measure soiling. Also these objects move within the illustration range of the camera on radials, which proceed from the vanishing point of the frame. The measuring windows are set along the radials, so that the object is depicted in all measuring windows along the given radials within a time period Δt with a suitable read-out frequency and/or a suitable chronological averaging of the measured values. Also in this case soiling will be identified on the basis of discontinuities.

As already represented above, the camera can also be used for coverage of the surroundings. If the vehicle is equipped with an surroundings detection system, the course of the driving lane and the associated markings can be calculated in advance. The measuring windows are set along the lane. For this purpose, at first the camera picture is divided into N×M regions. For each pre-calculated object trajectory the regions lying closest are determined and are set as measuring windows. If no marking is found in a predicted measuring window, a counter z(n, m) is incremented. If a marking is found in a predicted measuring window, a counter z(n, m) is reset. If a counter z(n, m) exceeds a fixed or a speed-dependent threshold, the region (n, m) is classified as soiled.

In order to determine size and position of the measuring windows in the frame, the roadway can be divided into regions of equidistant width. Size and position of the measuring windows in the frame result from the illustration model of the camera. In FIG. 1 the simplest case of a camera directed downwards is represented. The column C of a measuring window placed in a line y with lateral offset x calculates to

$c=\frac{x·f}{d·\eta }.$

Here d is the distance of a point from the camera, c the frame column of the point, η the pixel size and f the camera focal width. FIG. 2 shows the distance determination of a point d to the camera. The distance is determined from h the camera mounting height, a the camera pitch angle, y the frame line of the point, η the pixel size and f the camera focal width to

$d=h·\frac{1-t·\tan \alpha }{t+\tan \alpha }\mathrm{with}t=y·\frac{\eta }{f}$

Claims

1. A method to identify soiling on a protective screen, wherein a camera, in particular suitable to cover the surroundings in the motor vehicle, is focused on a scene behind the protective screen, and is characterized in that the relative velocity vrel of the camera and of at least one recorded object in the surroundings is not equal to zero,the trajectory of an object with vrel is predicted to be not equal to zero andthe conformity of the object recorded in different frame regions and/or several similar objects is analyzed and a conclusion is derived therefrom about soiling of the screen.

2. A method according to claim 1, characterized in that measuring windows are set along the predicted trajectory,the same object is detected in more than one measuring window in frames recorded in chronologically offset manner and/orseveral similar objects, in particular roadway markings, along a predicted trajectory are detected in more than one measuring window.

3-6. (canceled)