Methods for recording features in the environment

Abstract

Method for capturing features in an environment using at least one camera, a flash unit, and an evaluation device, in which at least four images are taken with substantially the same orientation, wherein at least two different camera positions and different lighting conditions are used for the images, and in which the positions of the features are determined by comparing the at least four images, and a mutual verification of the results of different acquisitions of the features is carried out by combining different acquisitions of the features, wherein at least two images of the features are taken with flash as a first acquisition, and by different illumination in the at least two images, in which the flash unit is once further away from the object and once closer to the object,The distances of the features to the flash unit and / or the camera are determined by comparing different flash photographs with known distances between a first flash unit location and a second flash unit location, such that the distances to the features are determined by flashing once from a shorter and once from a greater distance to the features, and • wherein, as a second acquisition, a size of the features is used to determine their distances by taking two photographs with two different camera locations, once with the camera closer to the features and once further away, and • wherein, as a third acquisition, a photograph is taken with flash, the flash unit being at a different distance than in the first photograph, and a base illumination is subtracted from the photographs taken with flash at different distances.which is determined from a photograph taken without flash, and the distances to the features are determined by calculating quotients.

Description

It is known to use optical systems, such as cameras or laser scanners, preferably for detecting features in an environment. The environment, for example in front of a vehicle in the direction of travel or in front of a stationary device in the area of danger, can also be detected with radar sensors or ultrasonic sensors. Optical sensors usually have a higher resolution compared to radar and ultrasound and are also used for two-dimensional or three-dimensional detection or measurement of rooms, buildings, objects etc. or also, for example, only of characteristic features on surfaces. The optical sensors also include those which operate in the ultraviolet or infrared range. Environment is hereinafter meant to mean the environment in front of the environment-detecting device, in particular the area which is of particular interest. It does not therefore have to be the complete environment. The features include, for example, recognizable edges, lines, surfaces, marks, stationary or movable objects, including real-worlds or parts thereof, persons, etc. When the features are displayed by a camera, two dimensions are generally recorded. Edges and lines, i.e. quasi-one-dimensional features, are perceived at least by their contrast. Features may occur in the singular as well as in the plural even if only one of the two forms is mentioned below. It is of interest to describe, recognize and / or ascertain features in the environment. The detection of features in the environment generally relates first to the recording sensor, e.g. a camera. As soon as the features in the camera have become clear, the alignment with respect to the camera is thus also relatively easy to establish. For the determination of the positions with respect to the camera, direction and distance are usually sought. This distance determination is initially difficult with a two-dimensionally operating camera.US 2006 / 0 171 566 A1 relates to a method for detecting distance values by means of a digital camera, with which a first digital image of a scene is recorded under a first illumination state, wherein the first digital image detects a plurality of objects at different distances from the digital camera. Subsequently, a second digital image of the scene is captured under a second illumination state. As a result, distance values of the distances from the various objects in the scene are determined by means of a single digital camera, in that the method makes use of the inverse quadratic dependence of the distance to the light source.DE 10 2009 005 861 A1 describes a vehicle environment recognition system having a stereo image recording device for simultaneously recording images of an environment around a relevant vehicle by means of a pair of cameras and for outputting the images as a reference image and as a comparison image. By means of a stereo matching device, parallaxes or distances are calculated by stereo matching on the basis of the reference image and the comparison image. Alternatively, the calculation may be performed on the basis of two images formed by subjecting the reference image and the comparison image to a predetermined first preprocessing operation. In this way, a distance image is determined by assigning the calculated first parallaxes or first distances to pixel blocks in the reference image.US 4 724 480 A relates to an arrangement for aligning a first object relative to a second object by projecting a light pattern from the second object onto the first object. An image of the pattern on the second object is transferred to the first object with a camera. From the camera image, it is possible to determine the relative position and orientation of the second object relative to the first object. In an alternative arrangement, the second light pattern and the first light pattern form a uniquely identifiable overlap pattern when the second object is oriented in a predetermined manner with respect to the first object. From the camera image of the overlapping pattern, the position and orientation of the second object relative to the first object can be determined.DE 10 2004 044 695 A1 relates to a method for distance measurement in which an object is illuminated in a structured manner in chronological succession, in particular from different directions, and an image of the object is recorded for each illumination and the distance of the object is calculated from the recorded images from the local offset of the illumination structures on the object. In this case, a plurality of images of the object are recorded from the same angle of view with different illumination. For this purpose, structured illumination with an uneven spatial distribution of the intensity of the radiation used is required.GB 2 453 163 A relates to a system for determining a three-dimensional representation of an object and comprises one or more light sources and one or more image recording devices arranged to record a plurality of images. In this way, the three-dimensional surface of the scene is determined according to the principle of photometric stereo systems. The positions of the light sources and of the image capture cameras are detected for this purpose by means of position sensors.US 2004 / 0 212 725 A1 discloses a method for generating a nonphotorealistic rendering in which shape features of real scenes and objects are captured without the reconstruction of a 3D model being required for this purpose. For this purpose, a camera with a plurality of light sources is used, for example a plurality of flash devices which are arranged distributed around the lens of the camera. The geometric relationship between the camera and flash devices is then utilized to detect depth. By forming the difference, the recognized properties can be emphasized and unnecessary details can be suppressed.US 2002 / 0 149 694 A1 relates to a three-dimensional image acquisition device which, in addition to a first and second light source, also has in particular a distance sensor. For this purpose, the light sources are embodied as pulse-modulated distance measuring light beams. A distance processor detects at least one of the light beams reflected on the measurement object of the two distance measurement light beams. The distance information of the measurement object is determined on the basis of electrical charges of the pixels.U.S. Pat. No. 4,843,565 A discloses a measurement method for distance measurement which is based on the ratio between the intensities of the reflections on an object detected by means of an intensity sensor, starting from two light sources which are arranged at different distances from the object.In the method known from US 2008 / 0231 835 A1 for determining three-dimensional data of detected objects, the distance from the object is also calculated on the basis of the intensity difference of the image data of the object detected with different illumination.If the environment is well illuminated, camera systems generally operate passively, i.e. without actively emitting energy itself, e.g. in the form of a flash. Radar, laser and ultrasound, on the other hand, must emit energy and rely on the reflection of this energy. In the case of unfavourable lighting conditions, however, it is also necessary for cameras to irradiate the interesting scene in the environment to be examined for the recording time, for example with a flash from a flash device.Stereo technology is occasionally used to detect the environment, but this also has its disadvantages. As a rule, there are too many contrasts in the images, which lead to ambiguities and also to other errors. The evaluation is very complicated and unreliable. It is not possible to recognize a wall with stereo technology if no contrasts are present on this wall.The object of the invention is to avoid the disadvantages mentioned and to improve the detection of features in the environment with little effort. In particular, features are to be reliably detected and their position determined.This object is achieved according to the invention by the features listed in claim 1. Advantageous embodiments are the subject matter of the dependent claims.According to the invention, at least one camera, a flash device and an evaluation device are used for detecting features in the environment according to claim 1. At least four recordings can be compared with the evaluation device. An electronic computer with memory and a suitable program, for example, is suitable for the evaluation unit. At least four images are taken of the entire environment or of a part of the environment, which images can then be compared.Features can advantageously be established by forming the difference between the two recordings. Since the camera generally records not the complete environment but only a part of the environment, possibly records a scene, the camera should be aligned substantially with the same scene or comparable scene for the recordings.To distinguish between at least two recordings, at least one of the two recordings can be stored in the recording memory. It is advantageous to first store the images and then compare the contents of the image memories. The comparison is expediently carried out in the evaluator device, for example with a suitable electronic computer. The pixels corresponding to one another in the images can be subtracted from one another in terms of their level in such a way that a difference image is produced. The regions which have approximately the same intensity, in other words brightness or level, in the two recordings will have a very small level or virtually disappear completely in the difference image.If the images are actually the same, nothing will be shown in the difference formation. The signal is quasi zero. However, since differences are preferably to be displayed, the conditions for the recordings must differ. This can be done by illuminating the scene differently for at least two recordings at different locations for the recordings or using an illumination body. The lighting element, e.g. a flash device, can be used only for one or more recordings or can also change its location. By different camera locations or illumination contrasts and features are disclosed in comparison with the recordings or in the difference formation and are determined in their position.The use of stereo technology is advantageous for determining the distance from features. A stereo camera typically includes two cameras. More complete, however, is the use of three or even more cameras. The cameras can be arranged next to one another or one above the other, but an arrangement one behind the other is also possible.In principle, the use of only one camera is also possible, wherein this camera then has to change its position for the individual recordings. For simplicity, the method will be described below predominantly with only two cameras which are arranged next to one another. The other constellations mentioned above can likewise be used and advantageous.The application of this technique is also interesting when a scene, ideally a relatively planar plane, e.g. a roadway or a wall, is observed at a known distance, e.g. in order to detect obstacles thereon or other objects in front thereof. Firstly, equalization is carried out of at least one image which is to be compared with another in such a way that the contrasts, structures and details on this plane on the one image are assigned to identical pixels of the other image. Thus, both receptacles become coincident, so to speak. If both images were then still produced in the same light, this has the advantage that, in the case of difference formation, the features in this plane cancel one another out (cancellation plane). It is also possible that this scene is uneven or that some objects or features are permanently present in the scene. For these unevennesses or features, which are located at a different distance or plane, an equalization of the recordings can also be carried out in such a way that an extinction occurs in the event of a difference formation. Planes at different distances are possible in principle in several areas of the scene, so that several different planes can be cancelled out in two images in the difference formation. Thus, several cancellation planes can also exist next to each other at different distances.Two images can now be easily compared with one another, which must be resolved in the difference formation despite different locations but because of the equalization. This preferably applies to a standard or reference scene.The advantage is now that new, additional features are disclosed very quickly in terms of their position and distance: if a feature now appears in the scene, for example, at a smaller distance than in the known distance, no extinction takes place at this point. The deviation is determined by the simple difference formation. The striking comparisons of a pixel in one recording with a plurality of pixels in the other recording, as is otherwise necessary in stereo evaluation, can thus be omitted.If an interesting feature has been found, the distance of the feature in both the equalized and non-equalized templates can be calculated according to the known method of stereo technology. However, it is now advantageous that it is not necessary to calculate the distance of all the details in the recording, but rather that it is now possible to focus on the interesting feature alone. Meaningful is such an equalization with difference formation, for example in order to detect obstacles above or on a roadway. The camera or cameras can be directed directly downward from a vehicle or preferably also obliquely forward in the direction of travel. The roadway is then a known plane, which is however inclined. If the vehicle does not roll too much, the inclined plane, i.e. the roadway, will remain the same in its position with respect to the vehicle even during travel. In the evaluator, an equalization for the recordings must be carried out for this plane arrangement, depending on the camera position. An obstacle on the roadway is recorded by contrast. The difference formation then does not lead to a cancellation and the obstacle is thus detected. The obstacle is the closer to the camera positions, the greater the feature displacement in two images is to one another.In principle, features can be determined in their distance if they are located in front of or behind a scene or a plane formed virtually or by calculation.The extinction of two images in the difference can also be realized for a specific plane by aligning the two adjacent cameras in such a way that their optical axes meet in this plane. The "cancellation plane" is not truly planar, however, but rather diffracted, which in practice can be tolerated to some extent or compensated by optical or computational corrections.At least two images are produced according to the invention with artificial illumination. The illumination intensity and / or illumination time are to be adapted according to the requirements. For this purpose, the use of a flash device or flash light or flash for short is suitable. For comparison, another shot with or without flash is used. Two exposures can be made with flash, but then the flash must fall from a different direction onto the scene so that other exposed and unexposed surfaces form.The advantage of this comparison of the images is that the differences in the difference image become very clear. The sign can also be taken into account in this case, so that negative or positive differences can arise. The larger amounts in the difference are important. If a value is particularly large, if applicable also has exceeded a certain threshold, the changes due to the flash light technique have become particularly clear. Thus, it is relatively easily possible to easily eliminate the differences simply by the subtraction. This process helps to simplify and speed up the recording evaluation. It is advantageous here that images with different flash positions or also in comparison with a non-flashed image, which show features more clearly by lightening (type two-dimensional) and / or by shadow formation (type three-dimensional feature), are identified than in a single image. The detection of these features is more difficult or not possible with a single recording.The term flash is to be understood relatively broadly. As a rule, the illumination for the recording is used only for the very short opening time of the camera. Therefore, the term flash is then appropriate. In many cases, however, artificial lighting is provided just in the twilight or at night in order to lighten a space to be monitored in such a way that a person can recognize the sought details or features. This also applies in particular to the headlight on vehicles. For recording without flash, the driving light can then be switched off for a very short time, for example on the vehicle. When it is turned off in the range of up to several hundredths of a second, it is practically hardly perceived by human. Suitable for this purpose are illuminations with LEDs. In this case, therefore, the phase with flash is then very long, or the typical state and the phase without flash are very short.Only because one of the two recordings is not flashed results in different representations. In principle, it is also possible to operate with reduced flash instead of without flash. The difference from the flash with full power is then smaller, however. It is then advantageous that the flash-free flash is very short. This is because differences in power are known, so that the flash effect can nevertheless be used for the detection.Whereas in the case of natural light, for example, a landscape or else an industrial environment are shown to be approximately equally bright both in the front and in the background, in the case of recording with flash, the surfaces in the removed background are not or are hardly illuminated, but the marks in the foreground, that is to say at a small distance from the flash device, are significantly illuminated in the color of the flash light. It is advantageous in that other contrasts are thereby created, which can be determined again in their distance using a stereo camera, for example, or that features are displayed differently solely on the basis of the brightness difference between the first and second recordings and are therefore emphasized and can be identified better. The fact that the recording without flash is subtracted from the recording with flash results in the pure effect of the flash as the difference.Not only can the general illumination over a broad range of flash light be used. If the flash light has only a certain narrow spectrum, it may be sufficient to detect the flash effect only in this certain spectrum.According to the invention, by making two images with flash of one feature and by having the distance of the flash device from the first image to the second image varied to a known extent in the direction of the feature, the distance to the feature can be determined by the brightness differences in the two images. In this case, it can be assumed in a simplified manner that the flash device for the emission region is a punctiform radiator. If the flash device is not punctiform or a plurality of flash devices are used at the same time, the following considerations apply in an adapted manner. However, a different flash light distribution in space must be taken into account. Flashing for distance determination may be sufficient, for example, at night when there is no disturbance light. However, in the case of a quasi-natural light present, at least one further recording without flash is advantageous to be made by the sun or by lamps, so that the disturbing light can then be calculated out in the distance determination. The flash exposures are thus freed from the disturbing light. The distance determination then uses the statement that the light of the flash with the reciprocal of the distance to the features whiten the surface of the features in the square. In other words, at half the distance, the flash light is illuminated four times brighter on the same surface.For the detection of features in front of a vehicle, it is appropriate to flash one after the other in time. If the second shot is taken after the vehicle has traveled a known distance toward the features, more flash energy will fall on the features and thereby lighten the features. In stationary devices, the use with two flashes one behind the other is more appropriate. The distance between the feature and the camera can thus also be determined if the position of the camera with respect to the flash device is known.If the camera is aligned with a plane, such as a camera on a vehicle, which obliquely looks at the roadway in front of the vehicle, then the near region of the plane, for example in the lower part of the camera photograph, is imaged and the far region, for example the horizon, is correspondingly further at the upper part of the photograph. The term "plane" here and below stands not only for the strict mathematical plane or a two-dimensional vector space, but also for inclined or wavy, curved planes or surfaces, in the sense of a geometrically defined or natural surface. If features are determined by comparing at least two recordings, preferably for features that touch or stand on the plane, for example for objects or persons, the distance from the camera to the feature can be determined by determining how far the feature appears in the lower or upper part of the recording. The distance to the feature can be calculated via the position, i.e. the height in the image. However, it should be noted that the lower part of the feature is used for the distance determination. For example, if the feature is a standing person, the feet on the floor must be viewed and not the head for distance determination. The person can be recognized by his shadow on the roadway. For this purpose, it is advantageous that the flash device is oriented with the person from the side toward the scene. A flash exposure and one without flash is then sufficient to expose the shadow. However, it is also possible to produce two recordings with flash. In this case, it is then possible to flash, for example, once from the left and once from the right. Here, a right shadow and a left shadow are then correspondingly produced already at the feet of the person.In a further advantageous embodiment, however, the person or another feature can be visibly separated from the plane already by a direct lightening. For example, in the case of a photograph without flash, but illumination from the ceiling or from behind, the person or another feature is not illuminated frontally itself. The person is thus more or less in the shadow of the natural light. When the person is photographed with a frontal flash, the person becomes much brighter, however. In the difference, the person can then be separated well from the plane. Instead of for persons, this example also applies to other features.A raised feature, for example an object which is located on a roadway in front of a vehicle, that is to say on a known plane, can be detected with the aid of the invention even if it does not differ significantly in the degree of remission from the degree of remission of the roadway per se. If the camera is oriented flat, i.e. predominantly more or less on the horizon, and the feature, e.g. an object or a person, represents a sufficient elevation above the roadway, in contrast to the non-flashed recording, a flashed recording will light up the person in the foreground much more than the remote background.However, even if persons and the roadway have approximately the same reflectance and thus the feet of the persons also have the same brightness as the roadway and thus the person cannot be initially distinguished from the roadway, the upper body will emit more lightning energy than the background removed. In particular, using a stereo camera, the contrast of the upper body in particular can be used for the direct determination of the distance. Stereo technology is also referred to as a camera (stereo camera), although usually two cameras are used internally. Both cameras are usually closely adjacent. However, the method is also suitable if there are greater distances between the cameras (base).The stereo method works well when distinct contrasts are present. However, features cannot be clearly emphasized, particularly in the case of similar reflectances and diffuse illumination. Therefore, the use of flash light, optionally also of a plurality of flashes at different locations in combination with the stereo method, is particularly useful. Features which are difficult to recognize quasi in daylight or conventional artificial illumination can also be better recognized by means of shadows and contrast formation by one flash or a plurality of flashes using stereo cameras.The recordings of e.g. left and right cameras, each with and without flash, i.e. 4 recordings each, result in addition to the two stereo recordings also a comparison (e.g. difference formation) of the recordings of the left camera with and without flash and a comparison of the recordings of the right camera with and without flash.In a further embodiment, further possible comparisons of images are obtained with further images, optionally also with further cameras and flash devices at different locations. Even the result of a comparison (e.g., difference) may be compared to other plots or other differences. Thus, the feature is ultimately more clearly determined in its position.Since, according to the invention, the camera is once closer or once farther away from two recordings, the position relative to the feature can also be determined. The relationship is inversely proportional: When the distance between camera and feature is doubled, the feature on the recording is halved in terms of height and width. Here too, differences from recording comparisons with and without lightning provide the clearer representation of the features. The use of a camera is appropriate if the camera can be moved or, for example, also for moving vehicles. The degree of approach (distance) is then determined by the distance travelled. With two cameras mounted at a known distance and aligned with the feature, the distance to the feature can be determined even if flashing is done only once. Both cameras then take photographs simultaneously.It is also advantageous if the camera and flash device approach one another with respect to the features, because it is true here that when the distance between the camera and flash device and the features is halved, the features must double in height and width and the effect of the flash light becomes noticeable by four times reflection.In order to keep the influence of disturbing external light low, it is advantageous in a further embodiment to use the flash energy as narrowband as possible. The camera should of course be particularly sensitive in the narrow band range.Working in the UV and / or infrared range is particularly effective. The sunlight and also the artificial, electrical illumination intended for us are hardly present or relatively weak in the UV or infrared light. Flash light is UV or infrared because it is hardly or not perceived by humans. It is also advantageous to work with LED flashes, since they are inherently present in narrow-band variants as well.In a further embodiment, correspondingly narrowband filters are suitable for the camera. However, extraneous light can also be suppressed in some cases by means of high-pass and low-pass filters. The filter can then cover the entire receptacle or only a part of the receptacle. Thus, a filter may have a different effect in the upper region than in the lower region. For example, if the camera in front of a vehicle is directed obliquely forward onto the roadway (scene), remote features are perceived differently than near features. If the filter is electrically adjustable or replaceable, two different images can be taken with a camera or the effect of the filter can be determined. A similar advantage is the use of color cameras, for example with sensitivity to blue, green and red. If flashing is then carried out with only one specific color, the flashing effect can be detected by lightening this color and not the other colors. Contrasts can also be emphasized in particular by working and comparing with different frequencies, thus also in the blue, red or green range. When using corresponding filters, it is then also possible for black / white cameras in the various colors to recognize contrasts.If the reflectance of the surface of the features in the environment is known, the lightning effect on these features can be used as a measure of the distance from the features. With known flash action, the brightness received in the camera can be used for distance estimation or measurement. The flash effect can be determined by subtracting the recording with flash and the recording without flash-i.e. only with quasi-natural illumination. It is advantageous for the measurement accuracy if the flash light is significantly brighter than the natural illumination on the surfaces.In a further embodiment, the invention can also be combined with laser, ultrasound or radar. All methods have their advantages and disadvantages. In combination, the weakness of one can be compensated for by the other. Here too, there is redundancy which leads to higher reliability by checking.If the size of a feature is known per se, the feature can be determined better in distance as soon as it is differentiated from the environment by the different recordings or their comparison. In particular, the evaluation of landmarks for the navigation of vehicles is appropriate. As a rule, the height of the landmark is more unambiguous than the width if it is not exactly known how the landmark is turned relative to the vehicle.Known features in the environment can advantageously be used as a reference in a further embodiment. This serves for checking the function and is thus relevant to safety. The known feature must be detected in the expected distance and / or direction. On the other hand, the reflectance of the feature can also be used for calibration in order to possibly carry out evaluations or measurements more accurately.Thus, by calibration at a feature with known reflectance at known distance, the evaluation with flash and camera can be calibrated such that another feature with known reflectance, but previously from unknown distance, can be determined at the distance.In addition to the safety check, the known features are also suitable in a further embodiment for determining the location or for navigation of vehicles, for example.A particular advantage is achieved in a further embodiment when using retroreflective surfaces. They emit the light almost exclusively in the direction of the flash. The contrast of the retroreflectors is even extreme with respect to matt, white surfaces. Reflection marks are particularly suitable as landmarks for the navigation of vehicles.A particular source of interference is mirrors. In a further embodiment, mirrors or mirror surfaces can be distinguished from the retroreflective markings by the specular reflections traveling along a moving vehicle from the view of the camera. Retroreflective markers, on the other hand, retain their known location. In stationary applications of the camera, the reflection on the reflecting surface in different directions can be detected by flash devices at different locations or by using two cameras at different locations. For example, in stereo cameras, the reflection is not presumed on the mirror itself, but rather apparently at the same distance behind it.By using flash devices at different locations, different, apparent reflective locations are suspected. This reveals the mirror surface. In contrast, retroreflective markers provide only visible reflection when the flash device is directly in proximity to the camera.Under certain conditions, it may be advantageous in a further embodiment to use already present external light for the recordings. It is known that some electric light sources with the grid frequency or twice the grid frequency are switched on and off. Thus, the interruption or the light phase of the external light can be used for one recording each. In addition, at least one further recording can be produced with flash. The images compared better the features.In some applications, especially in mobile applications, features which were not in the region of the flash device up to now are illuminated more strongly when the flash device approaches them and thus stand out more clearly from the background. Features are thus disclosed earlier or later by the changed light conditions and associated contrast and shadow values.It is therefore advantageous to continue the detection of the marks by further recordings. This serves not only generally to increase the accuracy, but also, if appropriate, to detect weaknesses or errors. Thus, disadvantageous constellations, for example, foreign and disturbing light sources, in particular the light and shadow effect of the sun, can cause errors which can only be detected by further recordings. It is of advantage here especially in obstacle detection that the effect of the flash is particularly intensive in the near region. At a greater distance, the effect of the flash is less, so that even in the difference formation of two recordings with or without flash, hardly any differences become evident. In a very long distance, the recognition of obstacles is also rather of minor importance.The continued detection of the environment, in particular with flash, assists the ongoing detection of the change in position, for example of movable objects. If an object moves with respect to the camera, then the previous course of the position changes, i.e. the previous movement, can possibly also be used to predict the future tendency to move.If the camera and flash are located on a vehicle, acceleration and speed in the translatory and rotatory axes can be determined from the position change of the vehicle.The advantage of the invention lies in the comparison of at least two recordings. The differences become particularly clear if both recordings are virtually subtracted from one another and a difference is formed. This applies both to two different flash exposures, but also to the difference formation from one exposure with flash and the subtraction by a further exposure without flash.For the comparison of two recordings, it is important in a further embodiment to create suitable preconditions. It may be necessary for the sensitivity of the images to be matched to one another. For example, if a flashed shot is compared to an unflashed shot, then the flash shot must not become automatically less sensitive because of the lightened features in the foreground. It is advantageous to select as a reference for sensitivity a range which is practically unaffected by flash, e.g. the background removed. There, practically only the natural light or daylight has an effect in both exposures. The difference between the two recordings can now be formed in such a way that the background shows no signal in the difference. The recording in the foreground then shows the influence of the flash. When comparing two recordings with flash, the background can likewise be used as a reference value or locations at which the influence of the flash is the same in both recordings.All other surfaces which have then been illuminated differently, in particular, by the flash, can be drawn off from one another. Thus, the features are clearly emphasized by different light and shadow effects with the flash.Instead of the far background, it may also be appropriate to use a closer surface with a known reflectance. Reference surfaces which are not used for the feature finding are suitable. Such an area can be used to select the desired sensitivity, particularly for flash exposures. For example, it is advantageous to regulate the flash recording in terms of sensitivity in comparison to the non-flashing recording in order to record features, e.g. objects, on a plane, e.g. a roadway, in such a way that the region to be examined on the roadway is equally bright in both recordings. Features, in particular objects, which are then illuminated particularly by the flash then contrast better from the background.Shadows in particular will then more clearly lift off in the images. This control is also suitable if two flash exposures are compared with one another.In a further embodiment, the additional contrast formation by the flash is fundamentally suitable for the detection of situation changes. For example, if:. If the camera and flash device are arranged in a known manner and monitor a known plane and a recording with flash is produced by the plane, then, depending on the arrangement of the flash device, a feature, depending on the position, may possibly also generate a shadow and is thereby detected or also by lightening the feature. The arrangement can be stationary and monitor the level in front of an engine or also be located on the vehicle and monitor the level in front of the moving vehicle. This can be used for so-called tracking, i.e. for tracking, and optionally also for target estimation. Repetition or continuation of the recordings and comparisons increases the reliability and / or accuracy of the detection.The presence of features, e.g. objects, in the vicinity of a machine or a vehicle can be advantageously determined according to claim 12 in that at least one camera is oriented more or less perpendicularly to a plane and views this plane, e.g. the roadway or the floor. By means of at least two recordings, of which at least one recording is effected with flash, this feature can be easily determined in its position or in the coordinates.This applies in particular when the feature is relatively large and comes close to the camera or the flash device and is thus lightened more than the plane. Advantageously, raised features, e.g. objects, can determine the position if the flash irradiates the scene more from the side. Then, the object or even a standing person forms a shadow on this plane. One shadow or a plurality of shadows can be formed by different flash lights in a plurality of exposures. However, the difference with a non-flashed recording also makes the position of the feature clear. The detection of features functions, of course, even if the plane is not horizontal, but is inclined vertically or arbitrarily. It is advantageous if the camera is more or less at right angles or perpendicular above the plane.As such, the method is disturbed by extraneous flashes. In a further embodiment, this is accomplished in that the separate recordings are triggered in some way, for example controlled via a central station or by synchronization on the external flasher. The extraneous flash may be a similar device, so that tuning is offered.Flash devices also often do not have a sufficiently uniform distribution of their light energy. Often, the light at the edge of the field of view is weaker than at the center. By means of a calibration and corresponding consideration in the evaluation, sufficient corrections are achieved in a further embodiment.If a vehicle with flash device and camera moves toward a stationary feature, for example, toward an object on the roadway, the distance to the object can be calculated with two recordings in succession, in which the flash device is once farther away from the object and once closer to the object. However, it is also necessary to know the distance between the two flash device positions.However, if the object is not stationary, but rather moves toward the vehicle at its own speed, the brightness increases more than in the case of a stationary vehicle. In the evaluation, the object is calculated more closely than it is actually. The faster brightness increase can thus be used advantageously for the more secure design in obstacle detection.The opposite is true if the object is moving at the same speed as the vehicle, so that the vehicle follows the object with a constant distance. The comparison of a plurality of recordings then reveals that the brightness of the object does not change. The object is therefore apparently infinitely far away and thus no collision is to be expected. The application also applies if the feature is a different feature instead of an object and if the feature moves towards a stationary arrangement of camera and flash device instead of towards a vehicle.In a further embodiment, its true distance can be determined from the luminance change by the flash light if the speed of the above feature is known. On the other hand, if the distance of the feature is known, the speed of the feature can be determined from the brightness variation. If the size of the feature is known, the speed, distance and even the orientation or movement path, e.g. of a vehicle, can be determined from the movement history.Optical systems typically have problems with precipitations such as rain and snow, especially when the camera and flash light are close together. Snow flakes and raindrops at a short distance in front of the camera or the flash device emit unusually strong radiation and may make the recording unusable. For outdoor applications, it is therefore advantageous if the camera and flash device are sufficiently far apart from one another.If the camera and flash device are directed obliquely onto a glossy or specular plane, the flash light is predominantly reflected in such a way that it is not recorded by the camera.The glossy area is then perceived more or less as a black hole. In a further embodiment, it helps the environment, insofar as it is predominantly matt, partially reflect the flash light in such a way that it is detected by the camera and therefore a glossy or specular plane is also indirectly detected. In this way, the evaluator can recognize the "black hole" and, when applied to a vehicle, reduce the speed, for example, or let the vehicle stop completely. The gloss on the roadway can be caused, for example, by moisture, water, ice or oil. Glossy objects, in particular specular objects, are also partially problematic for optical systems. If the reflection of the flash light does not fall directly into the camera, it is usually rather useless or annoyingly reflected into the room. However, the use of the flash is advantageous, in particular in this case in application with stereo technology. The contrasts with the matt environment can still be detected and used for determining the distance to the object.When a specular or glossy surface is flashed, this is often recorded in the recording as a particularly bright, but small area or flash spot. The rest of the surface, e.g., a feature, does not benefit therefrom. If necessary, the feature remains unrecognized or cannot be properly determined in distance. This is remedied by flashing from different directions and / or the use of a plurality of spaced cameras. Each flash produces a bright area, a flash spot, in each shot. In the ideal case, for example, 5 flash devices and 5 cameras with 25 bright flash spots can identify the feature, for example on a planar surface.If the positions of the flash devices and cameras with respect to one another are known, the distance from the feature can also be calculated from the distance of the flash surfaces with respect to one another. For this purpose, however, the curvature of the surface must also be known.The clothing of people is seldom glossy by a Y. Vehicles, machines, liners of immobilia and the like, however, often have a glossy or specular paint surface. Such features are perceived all the better if they are flashed and recorded from as many and as different directions as possible. From the diameter and brightness of the flash spot, one can infer not only the luster or mirror quality of the surface, but also the distance, if the properties of the flash device and the setting of the camera are known.In a further embodiment, the energy of the flash light does not have to be wasted in a useless manner. In particular at night, the disturbing external light is weaker, so that possibly less flash light energy can be used. This applies in particular when working with visible light and persons in the environment feel disturbed by the flash light.As a rule, sufficient contrasts are obtained in direct sunlight, so that, for example, features in their distance can also be determined using the stereo camera. It is interesting to apply the flash to sunlight, especially to lighten features in the backlight or generally the shadow in sunlight. As a result, the shadow can be better represented in the difference formation between the recording with flash and the recording without flash.It may be necessary to increase the lightning power in sunlight above the otherwise sufficient level in order to represent the lightning effect in sunlight as well and thus to better identify features.In order that individual, incorrect pixel values do not trigger a misassessment, it is advantageous to match the difference formation with the adjacent pixels as well. As a result, a certain amount of noise is filtered out. Furthermore, it is advantageous to assume a minimum size or minimum number of adjacent pixels for the features. Especially in the case of very rough surfaces with many small hard contrasts (pepper and salt), image evaluation is very difficult. It is then advantageous to form an average value for the brightness in corresponding regions over a larger number of adjacent pixels. The comparison of images is made substantially easier by the fact that these regions then have an average value.The pixels may be filtered in the images and / or after the comparison of the images. Too much compensation or too much smoothing (filtering) with the adjacent pixels, however, may result in a loss of contrast. Thus, it becomes more difficult to accurately detect edge or feature boundary in position. This loss of contrast generally leads, but in particular for the stereo camera, to more inaccurate distance determination. It is advantageous in a further embodiment to perform the compensation of the pixels, i.e. the filtering of the brightness values, preferably only transversely to the contrast gradient or in other words parallel at the contrast lines or edges. In the direction of the contrast gradient, the pixel values are then not smoothed. Thus, the contrast is clearly shown.In a further embodiment, threshold values or switching levels for the difference image can be defined so that an activity, for example a hazard message, can be triggered when features reach these values or levels.Not all areas in the observed environment or the picture need be interesting. In a further embodiment, uninteresting regions in the recordings can be excluded from the difference formation and / or. This measure also simplifies the evaluation.In a further embodiment, the invention offers a particular advantage in the combination of different detections of the environment. This also results in the mutual checking of the results. This checkability is advantageous in particular for safety-relevant applications. Thus, among other things, the following are suitable for combination:• - The additional contrast formation by one or more flashes. By imaging shadows on a known plane, for example, it is possible to draw conclusions about the distance to shadow-forming features by the evaluation.• - Contrasts are formed by the flash, which are necessary for the stereo method for determining the distance.• - The determination of the distance to features by flashing from shorter and from longer distances to the features.• - Use the range estimate from the lightening on characteristics by the flash if the reflectance is known.• - Range estimation when making two shots with two different camera locations. Once the camera is closer, once further away from the feature. If the camera is denser than the speed, the distance can be determined via the enlarged feature representation. The distance between the first and second camera location must be known.The evaluations can support each other in combination for the position determination of features, in particular objects or other features on a known roadway or plane.The use of the flash for detecting retroreflective features is advantageous. When two images are taken, one with flash and one without flash, the difference very clearly gives the retroreflective surface. Thus, retroreflective surfaces can be used to provide features that are particularly sensitive to protection, for example objects and / or persons.For simplifying feature detection, digitalization offers advantages. In the extreme case, a "black or white" decision can be selected for each pixel. Then, everything which becomes brighter than a certain gray value (threshold value) than white, everything which is darker, is declared as black. Images or difference images become simpler and more contrast-enhanced in the "1-bit resolution". The evaluation can be repeated and extended with other threshold values so that features can be determined better. Instead of digitalizing with one stage ("1-bit brightness resolution"), fine gradations may be better. This is particularly, but not only, true of stereo technology.For stereo technology, it is very advantageous if a photograph (e.g. left and right) is produced with flash and a further photograph with or without flash using the stereo camera. If the two images are compared, for example by forming the difference, a relatively complex image with many gray values usually results. Black and white digitalization produces a simple image of "black" and "white" areas. With a skillful selection of the threshold value between white and black, the sought features, e.g. objects or their shadows, are represented with few black-and-white transitions. The determination of the range of stereo technology is then limited to the very much smaller number of black and white transitions. The features can now be determined much more quickly in distance than was previously necessary with the many gray values.Usually, a single camera has a limited opening angle and can only capture a part of the environment. In a further embodiment, for example, when using a so-called fish eye, a plurality of cameras or pivotable or tiltable cameras, the detection can be extended significantly and possibly the entire environment can be detected with respect to the camera and / or flash device location. Rotating line cameras are also possible. In particular, the horizon is of interest for the detection of features, for example for navigation or obstacle detection. For this purpose, at least one flash device must then also be designed accordingly for this task, so that the desired regions are illuminated. This arrangement is suitable for navigation in particular with artificial landmarks, including retroreflective landmarks. Flash device and camera then need only be large enough at the vertical opening angle (height) that they still record the landmarks. Ideally, the horizontal registration can be up to 360°. If the camera and / or flash device can be pivoted, one unit can in principle be used, while in the case of the fish eye principle or when using a plurality of cameras, one may optionally have the higher outlay for optics and cameras. However, the pivotable solution has the disadvantage that a suitable drive and an accurate detection of the rotation are necessary for direction determination.The fact that only a part of the image is illuminated with the flash device can be advantageous if the flash light is useful only in certain areas and / or is disruptive in other areas, for example for persons or devices. It is also advantageous to clearly delimit the flash, so that a flash contour is reproduced in the environment. The flash can form, for example, dots, lines or patterns with a special aperture, among other things. For example, if the flash forms a horizontal strip and at least one camera is spaced above or below the flash device and aligned with the horizon, the strip is detected on nearby features at a different elevation on the scene than on remote features. The distance of the recording can thus be determined. Even if the flash device with the stripe formation is oriented more or less upward or downward, the distance can be determined.In particular in vibrating, moving or mobile applications, e.g. when the camera is located on a moving vehicle, the camera can execute unfavorable movements between the recordings, so that the recordings do not originate exactly from the same pose, in particular due to rotational movements, and are therefore not optimally comparable to one another. By ascertaining the rotational and / or translational change in the camera pose, in a further embodiment, this error can be taken into account and corrected for the comparison of the recordings, for example in the evaluation device. In the case of purely rotational movements, a displacement or rotation of the receptacles to be compared may suffice.If the camera moves over a greater distance from one recording to the next, it may be advantageous to take into account the new perspective by corresponding equalization of at least one recording.It is advantageous to design or use features in such a way that their alignment can be recognized. The pose of the feature can then already have an advantage per se, for example when aligning or setting up with respect to the feature or the environment, but also for navigating or determining the location or pose of the camera and / or flash device. Features which permit direction detection can be, inter alia: roadway markings, guardrails, in some cases also signs, walls and the like. Special features, for example artificial landmarks, make their alignment easily recognizable, in particular if they are equipped with retroreflectors and are thus particularly clearly shown with the flash light. In contrast to detection using a laser scanner, it is advantageous that the height of the feature can also be detected and used in comparison with flashed and non-flashed recording.Also particularly suitable are three-dimensional features which, depending on the viewing direction, also make their orientation recognizable. In principle, the pose of the camera and / or flash device can already be determined on one of these features. More specifically, however, the determination becomes when several known features of this type are used.In a further embodiment, the distance of a feature can be determined with a camera that is oriented more or less horizontally if the mounting or position height of the feature is known. It is important that the feature, for example, also a natural or artificial landmark lies clearly above or below the viewing or movement plane of the camera, so that, when approaching, the feature is seen with a marked angle change to this plane. Two landmarks of this type are already sufficient for self-positioning of a vehicle.A third feature and further features can possibly cancel an ambiguity, improve the measurement accuracy and / or extend the action range. The positions of the landmarks must be known. It is advantageous if the camera and flash device individually or in a plurality capture the entire horizon. Otherwise, if the camera and flash device on a vehicle are only oriented forward or forward and rearward, for example, more landmarks may need to be used or the route is very limited. Sensor fusion with odometry may be advantageous or necessary.In a further embodiment, retroreflectors can be distinguished from lamps or other self-illuminating bodies in that they illuminate only when they are illuminated. In contrast, in the case of a comparable recording, but without lightning, only the lamps are illuminated. In comparison of the recording with flash and the recording without flash, the difference becomes very rapidly clear, for example, in the difference formation.A feature can advantageously also be established by checking whether all regions or even pixels have the expected brightness. This applies when searching for features, but in particular also for detecting deviations, for example for detecting obstacles. On a gray roadway, only the information "gray" of the pixels assigned to the roadway will be expected in the recording. However, if a different brightness or color is displayed, a feature is finally recognized, if appropriate after a specific tolerance is exceeded.The expectation that this pixel region concerned is actually gray can optionally be predefined to the evaluation device from the outside, for example by an instruction, an input or by a map. However, which brightness is expected can also be learned, for example by the fact that the roadway under consideration has always been light gray up to now and furthermore only light gray tones are expected. Known deviations, such as spots on the roadway or transitions to other roadway substrates (asphalt / concrete), may then not be recognized as a disturbance or obstacle, but may even be used as references or path markings. During a learning run or repeated runs, the brightnesses to be expected can be stored location-dependently. Thus, firstly references or landmarks are obtained, and secondly location-dependent brightnesses deviating therefrom can then be interpreted as disturbances.Changes in brightness in the environment are particularly clearly caused by the changing solar radiation during the course of the day, for example also by the corresponding formation of light and shadows or also by artificial illumination in the environment. Given corresponding intelligence of the evaluation device and / or supplied information or instructions to the evaluation device, the brightness changes can be taken into account, if appropriate, in the recording. Under the effect of the flash, the plane must then, in a further embodiment, reflect the determined expected brightness, depending on the viewing angle and distance and remission. At another location, a different brightness of the roadway can then be expected, if necessary. Depending on the location, the sunlight can then also be taken into account, for example, depending on the day. The evaluation device can thereby acquire the information about the lighting conditions via a dedicated sensor, but also from an intelligent evaluation of the recording, among other things. For example, the situation-dependent brightness can also be determined on known features, e.g. known landmarks or references. It is advantageous that the illumination of the features by the flash light is independent of the normal illumination, for example by sunlight or by the room illumination.It can thus be determined that no obstacles are present on the plane if the lightening by the flash occurs as expected. The surfaces must be lightened as much as expected, taking into account the distance, reflectivity and viewing angle of the camera and the orientation of the flash device in the photograph.This advantageous detection of obstacles above the roadway naturally also applies to other features in the environment.The flash device should be oriented and designed so that the flash substantially only illuminates the area of interest. An over-going lightening would possibly be an unnecessary waste of energy or could also lead to disturbances or glare to other participants. Another disadvantage is that the flash light is unnecessarily reflected in the environment, e.g. on ceilings and walls, and can thus even disturb the flash effect in the area to be examined, e.g. because the desired shadow effect does not become too clear (scattered light falls into the shadow area) or that the distance-dependent lightening of the features no longer occurs as provided.The recognition of features is already possible by forming the difference between two recordings in comparison, i.e. two flashed recordings or one flashed and one non-flashed recording. Further advantages are the determination of the quotient formation features Q= PB / PoB, where Q indicates how much a region or pixel in the image has illuminated only by flash compared to an image without flash, PB is illuminated only by flash measured at a pixel in the image, and PoB is illuminated only by flash measured at the same pixel in the image.Q is a measure of the distance between flash device and feature, regardless of the reflectance of the surface of the feature.On bisecting the flash-feature distance, PB, and hence Q, increases four times.A quotient formation Qm=PmB / PoB, where Qmindicates by how much a region or a pixel in the recording by flash and basic illumination has illuminated compared to a recording without flash, i.e. only with basic illumination, and PmBis the total brightness from illumination by flash and basic illumination at the pixel in the recording or similar comparisons, including reciprocal values thereof, indeed lead to other quotients, but likewise show the principle.It is advantageous that features at different distances can now be easily separated by the different quotients. Pixels that have the same quotient value then optionally belong to the same features, e.g. an object, an obstacle or a person. Pixels having a different quotient value then correspondingly belong to a different feature or also to the removed background.Note that features that are adjacent to each other do not differ in their quotient value. They have the same distance from the flash device and thus form a common pixel or point cloud in the pictorial representation of the quotients. These features may possibly be distinguished by their color or brightness. In many cases, however, it is only decisive whether or not there is any feature, for example an object or an obstacle, in the space to be monitored.In the quotient formation for distance determination, each quotient stands for a corresponding distance. The quotient is dependent on the flash intensity. Therefore, it is advantageous to detect and maintain the flash characteristics constant. Flash energy, reflectance, camera sensitivity and camera dynamics limit the range of the distance determination, in particular in the case of bright basic illumination, for example due to sunnyness.By repeating the quotient formation with further recordings and thereby changing the distance between flash device and feature by a known amount, a different quotient results. The change in this quotient must correspond to the change in distance. If the quotient or distance changes, this can be checked with the change of the respective other variable.In a further embodiment, recordings can be easily compared with one another if the requirements for the recordings are also suitable. In particular, when images are automatically produced, it is usual, when illuminated with flash light, that the opening duration of the diaphragm and / or the diaphragm opening decrease compared to the non-flashed images. However, in order to detect the effect of the flash, it is then simpler in comparison if the setting for the exposure time and the aperture do not change. The far background, which is practically not illuminated by the flash, remains the same in both exposures, flashed or unflashed, while the foreground in the flashed exposure then becomes significantly brighter. If the diaphragm aperture and / or the opening duration are not the same for the recordings to be compared, at least one recording can be corrected if necessary with the evaluator or another computer, so that a correct difference can be formed.Features can be best determined when they form hard contrasts. It is therefore advantageous if hard contrasts are preferably evaluated in the recordings and / or after the comparison of the recordings. Contrasts can be absolute or relative brightness changes between adjacent pixels.If two images are compared with one another, in particular high-contrast features, for example at their edge compared with the background, will show a distinct brightness gradient. The brightness difference between adjacent pixels can be used as brightness gradient or contrast. If the difference exceeds a certain absolute and / or relative extent, sufficient contrast is present. If the search for features is limited to sufficient contrasts in the evaluation, the majority of the recording can frequently be excluded from the further evaluation. This makes the evaluation clearly simpler.A plurality of adjacent pixel comparisons may represent a contrast line that runs transversely to the gradient. This line often represents the edge, border, or boundary of a feature.Sharp, high-contrast imaging of the environment is not possible with the conventional camera technology over the entire range from the foreground to the horizon. However, this restriction can also be used to advantage in a further embodiment, in that only the desired region is imaged sharply and with high contrast, for example, by a corresponding aperture selection and the corresponding focusing. Thus, for example, to monitor a hazardous area, it is usually only necessary to sharply image the very close features, but not the background. This optionally also facilitates the evaluation or the comparison of two recordings. If no contrasts are present in any case in the case of an uninteresting background, the two recordings to be compared will also show no appreciable differences here. Only the contrasts in the region of interest are displayed, which makes the evaluation easier.A preferred positioning of camera and flash device is when features can be displayed close to the far background. The flash device then substantially lights up only the feature while the background is unaffected. Such a setting is present if, for example, the camera and flash device are relatively close to one another and are aligned with the horizon in order to monitor a room. Camera and flash device are also positioned, for example, close to the roadway, the floor or the like, so that, for example, a large feature can actually be displayed even in front of the far background and not in front of another close plane. The far horizon is illuminated only insignificantly, if at all, by the flash device and the camera is also focused on the closer space to be monitored, in the foreground, if appropriate. The receiving region can in any case also be limited in height and width to the region of interest. Such a constellation is useful, for example, for detecting or monitoring a room, for example above a floor, above a roadway or even in front of a wall or for monitoring a part of the room.Powerful LEDs are now also suitable for the flash device. They are not only stable, long-lived and efficient, but can even be switched with steep edges and can be modulated with high frequencies, e.g. a few MHz. They are therefore also used for the so-called PMD technique (photonic mixed devices) with 1 pixel or also as a pixel array. It is therefore advantageous to modulate such a flash device and to use, in addition to the cameras for the previous recordings, additionally suitable receivers for the phase measurement. Even with only one such receiver (e.g. PMD), the distance, e.g. to the roadway in front of a vehicle, can then be measured. Thus, if the height above the roadway is known, the inclination to the roadway ahead or to a reference feature can also be checked. This is again useful, for example, for verifying the recordings produced with the cameras.The phase measurement may initially allow for ambiguities in the distance determination, but in conjunction with the evaluation of the camera recordings, the ambiguities are resolved and the distance determinations are more accurate.Without any ambiguity, the distance measurement can also be carried out with the direct transit time measurement in a further embodiment. With an LED flash and individually directed time-of-flight receivers, distance measurements are possible for supporting and verifying the camera recordings, even if individual measurements are not particularly accurate. However, since several 1,000 measurements can be carried out per second, a high accuracy can still be achieved finally via the mean value of many measurements.If the LED flash device is in any case present for the normal camera recordings, the additional outlay for an individual module or an array of phase- or time-of-flight measuring receiving modules is more favorable than the use of additional phase- or time-of-flight measuring receivers with its own light source.In a further embodiment, a distance-measuring receiver is aligned with the regions of interest. A larger range can thus be gradually measured even with a 1-pixel receiver. Thus, the feature detection is supported with the cameras. The tracking can expediently be effected, for example, with electric motors.The plane in which features in the difference formation are cancelled can be varied in their distance from the cameras. As a result, only the features which are present at other distances have to be processed further in the evaluation. This can be done by changing the optical axis of at least one camera in such a way that the optical axes of the cameras meet at a closer or further distance. One of several possibilities is the rotation of at least one camera. However, a change in the pixel evaluation is also possible, so that pixels in one image are assigned to other pixels in the other image.Depending on the distance at which the cancellation is desired (cancellation plane), one recording can be shifted relative to the other recording in the difference formation. At long distances, the shift is small, and at short distances, it is correspondingly larger. Apart from the fact that in the case of complicated recordings, ambiguities in the stereo evaluation can also occur, it is advantageous to compare the recordings with increasing or decreasing displacement, if appropriate in steps. The image areas in the images for which there is an extinction at the "non-interesting" distances can be gradually excluded from the further image evaluation. Conversely, the following applies: In the recording regions in which subtractions occur at the "interesting" distances or displacements of the recordings with respect to one another, features to be observed are present, if appropriate.The advantage of this embodiment is that the interesting viewing range can be expanded or shortened depending on the situation. For example, in monitoring the space in front of a vehicle, this space may become longer or shorter depending on the travel speed. At high driving speed, a deep space will have to be detected. At low vehicle speed, not only the wide but also the middle range is of no interest for estimating the hazardous situation. Then, the cancellation plane may be placed on the middle region. Only the near area is then observed and evaluated.In order that the extinctions take place not only in a more or less flat plane, but the extinction takes place over a larger region, i.e. quasi over an extinction space, small disparity values can be neglected. Only larger disparity values, which thus belong to features that lie outside this space, are then further treated. Conversely, it is also possible with the cancellation plane or the cancellation space to preferably select the features that are of further interest.Depending on the sign and contrast transition (bright to dark or vice versa), it can also be established in the difference image whether the associated feature is located in front of or behind the cancellation plane or the cancellation space. The difference formation (comparison of two recordings by difference, difference image) also shows the advantages in stereo technology. Interesting features can be defined and displayed with little effort, depending on the distance. The use of a flash helps in contrast formation or in many cases first creates the necessary contrasts.If two images are produced, one without flash, in order that the basic illumination can be determined and one with flash, in order to highlight a nearby feature in particular or also to determine the distance, then the distance can be determined independently of the level of remission on the feature, but not independently of the basic illumination via the quotient formation. This is because this works only if the illumination is very diffuse and the feature is illuminated uniformly. In practice, however, this is often not the case.It is advantageous that a further exposure is produced with flash light. However, the flash device must have a different distance than in the first flashed shot. The basic illumination can thus be subtracted from both flash recordings, so that only the pure flash effect of both flashed recordings is still present. Thus, the distance from the feature can then also be determined by comparing the different images with the flash light and with a known distance between the first flash location and the second flash location. For example, if a person is daylight (sunlight), then some parts of the person are brightly illuminated and others are shaded. With a non-flashed recording, therefore, the basic illumination on the person can first be ascertained. With a flash exposure, initially at an unknown distance, the illumination by the flash light will then be determined in different areas.Then, if a reduced distance exposure is made with a further flash exposure, possibly by approaching the person, all parts of the person must appear much brighter in the flash light. If the distance has halved, the flash brightness will double on the subject. This method is thus independent of the remission of the feature, but also independent of the basic illumination. If, in this example, two flash devices which otherwise have the same data are arranged on a system carrier, for example on a vehicle, these two flash devices can be arranged one behind the other on the system carrier at a known distance or else the vehicle moves over a known distance in the direction of the feature. The brightness of the flash light only decreases with the square of the distance of the feature from the flash device, at least insofar as the flash device is a point-shaped radiator. In the case of non-punctiform emitters, a correspondingly different beam characteristic must be taken into account. Defects can also arise from unfavourable reflections of the flash light on other surfaces in the environment.The method also works with glossy, specular or also retroreflective surfaces. As such surfaces are approached, the reflection becomes stronger as intended.An already present illumination can also be used. With the vehicle lighting, a working headlight or a general lighting for a monitoring space, the function of the flash device can be fulfilled in a way. With the artificial illumination that is usually permanently switched on, the recording is then created, which has previously been referred to as a recording with flash.In order to then produce a recording without flash, this illumination must be switched off briefly. In order that no impairment takes place for a human observer, for example, the switch-off may have to be very short, i.e. only a few milliseconds. In principle, light-emitting diodes, for example, are well suited for such short switching times.The advantage of the invention is evident especially in the dependent claims. Features which could not be recognized up to now can finally be represented in a hard contrast by the flash and difference formation. The nearby objects are particularly clearly emphasized. This also helps in the evaluation using stereo technology. The evaluation becomes simpler and more reliable overall.In a further embodiment, the width of the disparity is used for distance determination. When two images are compared with the stereo method, when the cameras are aligned in parallel, close contrasts with a large disparity and more remote contrasts with a smaller or infinite disparity appear. If the optical axes of the cameras are tilted with respect to each other, the plane of extinction may be advanced. However, this is also possible by displacing the receptacles relative to one another. Then, in principle, the disparity can also assume negative values. Nevertheless, the distance can be determined more or less accurately from the width of the disparity. This determination can be made over several rows or columns, so that good mean values are formed.When multiple exposures of one feature are made, the flash position is changed. If, for example, the camera is aligned obliquely from above on a feature and the feature is exposed directly from above in one recording and frontally from the front in the other recording, then the lightening in the difference, optionally also with a comparison of a third recording, can be used completely without flash in order to emphasize the front side or the top side of the feature in particular. Depending on the intensity of the flash perception on the object, it is possible to determine how much this side, front or plane of the feature is aligned with the respective flash device. In the case of the flash from above, the planes which are predominantly aligned upwards become significantly brighter, and optionally also the base on which the feature stands, in the case of the flash from the front, the planes become brighter the more the relevant plane is aligned with the front flash.Exemplary embodiments are illustrated in the following drawings and are described in more detail below.The following are shown: FIG. 1 shows the detection of features in the side view. FIG. 2 shows the detection of features in the side view with a narrow flash. FIG. 3 shows features from the view of the camera. FIG. 4 shows features viewed from the narrow flash camera. FIG. 5 shows shadow generation on a feature with two flash devices. FIG. 6 shows a feature representation from the view of the camera. FIG. 7 shows a feature representation with a camera perpendicular to a plane by shadow formation. FIG. 8 shows a feature representation with a camera perpendicular over a plane by different illumination. FIG. 9 Artificial landmark FIG. 10 shows the detection of features at different distances. FIG. 11 shows filtering of the pixel values. FIG. 12 shows the detection of a feature in front of the removed background. FIG. 13 shows further recognition of a feature in front of the removed background FIG. 14 shows the recognition of a feature using stereo technology FIG. 15 shows recognition of a feature using stereo technology on a vehicle FIG. 16 shows the change of disparity and focusing FIG. 17 Distance determination of features in the case of unfavourable illumination FIG. 18 shows the principle of the cancellation planeA possible embodiment of the invention is shown in a side view in FIG. 1. A flash device (1) irradiates the scene in front of the camera (2). Flash device (1) and camera (2) are connected to one another via evaluator (3). Two features (5.1, 5.2) can be seen on the plane (4). An artificial landmark (35), which is another feature, is mounted at a known mounting height (36), e.g., on a wall not shown. The scene is illuminated by a lighting element (6), which can be artificial illumination or the sun, so that a certain background brightness is present in the scene. Flash device (1), camera (2) and evaluator (3) are fastened to a system carrier (7). The system carrier ( 7) can be a vehicle, a machine, a wall or any desired, also portable, holder. The evaluator ( 3) can furthermore also be remote. The flash is defined for the interesting scene in the area by the flash lower limit (11.1) and the flash upper limit (11.2). The camera (2) is aligned with the horizon plane of the camera (12). In this orientation, the first feature (5.1) is established in the upper and lower feature boundaries in the lower field of view of the camera (2) at a specific angle to the horizon plane of the camera (12). The features can be lightened with flash from the background removed. In comparison with a recording without flash, the illumination is determined by the evaluator (3). All features (5.1, 5.2, 35) detect their direction with respect to the camera (2). Since the height (37) (physical size) and also the mounting height (36) are certainly known from the artificial landmark (35), which may be distinguished particularly clearly from the background by a retroreflective surface, the distance up to the camera (2) can be calculated using both values or using the angles to be assigned with respect to the horizon plane of the camera (12). It is assumed here that the height of the camera (2) above the plane (4) is also known. Thus, a distance calculation is possible via the mounting height (36), but also via the known height (37) of the artificial landmark (35) itself.The features (5.1, 5.2) which are located on the ground above the plane (4) can be, for example, objects, obstacles or even persons. If it is assumed that the features (5.1, 5.2) do not lift off significantly enough from the plane (4) or the background without flash light even by illumination with luminous element (6), then they are illuminated more clearly with flash, namely the more the closer they reach the flash device (1). The lightening on the first feature (5.1) can be multiplied if a distance of one meter is reached instead of the distance to the flash device (1) of two meters, for example by the feature (5.1) approaching one meter or if the system carrier (7) is a vehicle and the entire system carrier (7) approaching the feature (5.1) closer by one meter. However, it is also possible for a second flash device (1) to be used, which is positioned one meter closer to the feature (5.1) than the first flash device (1). In any case, when the distance between flash device (1) and feature (5.1) for camera (2) is halved, the brightness on feature (5.1) becomes four times as great as a result of the flash effect. However, the predefined brightness must be taken into account by the luminous element ( 6), for example by recording without flash.If the second feature (5.2) has the same level of remission as the first feature (5.1), the brightness on the second feature (5.2) will be only one quarter as great because of the distance to the flash device (1) twice as great.It can also be seen that the illumination of the plane (4) by the flash device (1) is advantageously relatively low if the flash device (1) is mounted as far down as possible on the system carrier (7), that is to say relatively close above the plane (4). Then, relatively little light from the flash device (1) falls onto the plane (4) before the first feature (5.1), because the angle of incidence is very flat and thus only little light is reflected into the camera (2). It is most advantageous in the case of the vertical flash incidence (11.3). Here, there is most flash energy per area.Therefore, the feature (5.1) with flash is seen much brighter in the camera (2), even if it has the same reflectance as the plane (4).It can also be seen that in the situation shown, the luminous element ( 6) irradiates the first feature ( 5.1) from behind, with the result that a shadow forms in front. It is now advantageous to lighten this shadow with the flash device ( 1), so that in the difference formation of two recordings with and without flash, the lighten is detected particularly clearly by the evaluator ( 3). As a result, the first feature ( 5.1) is particularly clearly distinguished from the environment.A similar arrangement can be seen in FIG. 2. Here, however, the light emission at the flash device (1) is very much narrower and is more narrowly limited by the lower flash limit (11.1) and the upper flash limit (11.2). It can be seen that the light gap, which runs horizontally for example, impinges on the features (5.1, 5.2). It can also be seen that, owing to the double distance, only one quarter of the flash energy per area arrives at the second feature (5.2). This difference in brightness can be detected by the camera. However, it is even clearer for the distance determination that the flash strip ( 5.1 b) on the first feature ( 5.1) is seen below the light gap limits from the view of the camera ( 2). These light gap boundaries appear in the camera (2) which is aligned with the horizon plane of the camera (12) at a specific angle or on specific lines of the camera (2). Since it can be assumed that the height of the camera (2) above the plane (4) is known, the angles which are to be assigned to the light gap boundaries result-according to the rules of the geometry-in the distance to the lightning strip (5.1b) on the feature (5.1). It is assumed here that the orientation of the flash device (1) and the height of the flash device (1) above the plane (4) are also known. It is also possible to change the light transmission of the flash device (1) in such a way that only a prominent limit of the flash falls on the features or on the environment. If the transition from the flash region to the non-flash region is very high in contrast, the environment with its features can be determined with a known arrangement and orientation of flash device (1) and camera (2), also in the distance.Furthermore, however, the entire interesting environment or the interesting scene can also be illuminated overall again with a large opening at the flash device ( 1). A prerequisite for the distance measurement is that flash device ( 1) and camera ( 2) have a sufficient distance from one another.FIG. 3.1 shows the view from a camera (1) onto a plane (4). The camera ( 1) is more or less aligned with the horizon ( 15). The obstacles (5.1, 5.2, 5.4) can be seen at the distances 1m, 2m and 4m, respectively. With diffuse illumination and the same level of reflectance on all surfaces, the scene appears "gray in gray". Here, the brightness has been given the relative value "1".In FIG. 3.2, the scene has been illuminated relatively flat with a flash device (2), not shown here, which, as can be seen in FIG. 1, is mounted closely above the plane (4). In this example, a strong lightening results on the first feature (5.1) at one meter distance and a correspondingly weaker lightening results at four meters distance. The flash support has the same value as the background brightness at a distance of four meters, i.e. "1+1=2". At a distance of one meter, however, the lightning effect is already sixteen times the value, so that an overall brightness of "1+16=17" results. In this example, it is assumed in simplified form that at the same distance on the adjacent plane (4) only one tenth of the light incidence is detected, which falls on the vertical surface of the first feature (5.1).FIG. 3.3 finally shows the difference between the two recordings (FIGS. 1-2 ), namely the "pure" lightning effect. It can also be seen that the lightning effect, i.e. the lightening by the lightning, increases inversely with the distance from 4 mto 2 mto 1 mwith a ratio in the square from 1 to 4 to the value 16. If the quotient Q, i.e. the lightening only by the flash, is formed with respect to the brightness by the natural illumination, then the quotient is independent of the reflectance of the surface.In this case, nothing changes because the background brightness was assumed to be 1. If the distance has now been calibrated beforehand to Q values, i.e. it is known at which Q value which distance is present, the distance of the features (5) can be determined directly. It is also possible to use one of the features (5.1, 5.3, 5.4) as a reference if the distance is known to do so. Then, the other feature distances can be derived from this reference.The fact that the features (5.1, 5.2, 5.4) settle well from the plane (4) becomes clear in particular in the near range.FIG. 3.4 shows, especially in the first feature (5.1.), that the contrast to the adjacent plane (4) has become particularly clear as a result of the application with the flash and the later difference formation. It must be taken into account that the flash light has a higher effect in the vicinity of the plane (4) than in the more remote region on the plane (4). The flash effect on plane (4) ranged between "1.6 to 0.1.". Thus, the contrast becomes clear as compared with the first feature (5.1) having the brightness value "16" on the entire surface. Although the contrast is already lower at the lower edge ( 25) of the first feature ( 5.1), it is still much better than in the further features ( 5.2 and 5.4). The features can be clearly shown if the brightness values are checked at a threshold in the evaluation with the evaluator ( 3) and only black or white statements are permitted. If, for example, the threshold value in FIG. 3.3 is set to "10", only the first feature (5.1) is shown. The remainder is black. If the threshold is set anywhere between "2" to "4", the second feature (5.2) would also appear bright. The distance of the features with respect to the camera (2) or the flash (1) can thus be determined in a variety of ways:• - By lightening the features.• - By their imaging height, if the height of the features is known.• - By the angle or the line in the camera on which the lower edge () of the feature is represented on the plane ().The representation of the lower edge (25) is helpful and provides good values if the orientation and height of the camera (2) with respect to the plane (4) is known.The fact that in the arrangement provided with a camera, for example at a height of approximately one meter, the features (5.1, 5.2 and 5.4) cannot be separated from the plane (4) as well if the flash device (1) is likewise approximately at the same height and is not close to the plane (4) is shown in FIG. 3.5. The flash device (1) has in FIG. 3.5 the features of which are as bright as in FIG. 3.1. However, the plane (4) has just been illuminated in the near region just as the first feature (5.1). In fact, the lightening in plane (4) is weaker in the far range than shown here.FIG. 3.6 then shows the brightness values which are to be expected at the right and left edges of the features (5.1, 5.2 and 5.4) at the specified heights. It is thus evident that the lower edge ( 25) no longer optically lifts from the plane ( 4) due to the contrast. However, a distance determination to the features is still possible by using a stereo camera or by approaching or removing the first or a comparable second flash device ( 1).FIG. 4.1 shows as initial position for the further considerations firstly a scene which was recorded without flash light. The features (5.1, 5.2 and 5.4) are located on the plane (4), for example on a roadway.In Fig. 4.2, the scene has been flashed. In an arrangement according to FIG. 2, only a very narrow, horizontal light gap was opened for the flash device (1). The result is displayed in different ways on the features (5.1, 5.2 and 5.4) from the view of the camera (2). From the view of the camera (2), the flash shows effect in the form of flash strips (5.1b, 5.2b, 5.4b). Although the strips become increasingly larger in distance, they appear the same size from the perspective of the camera (2) but different degrees of brightness with the same level of remission. The line density shows that the brightness of the first flash strip (5.1b) 16x is as high as in the case of the last flash strip (5.4b). Finally, only the flash strips (5.1b, 5.2b, 5.4b) become visible in the difference between the flashed record and the non-flashed record. The remainder has been cancelled out by the difference formation. The evaluation or the recording per se is characterized by the limit of the recording (30). Corresponding pixels are to be assigned to this on the sensor plane, which are arranged in horizontal rows and columns running vertically. Since the camera (2) has a known orientation and height with respect to the plane (4) and, in this example, the orientation and height of the flash device (2) with respect to the plane (4) is also known, it now follows that the flash strips (5.1b, 5.2b, 5.4b) or their top or bottom edges are assigned to corresponding lines, for example the top and bottom image lines (17.1 and 17.2) on the sensor plane. The distance of the first lightning strip ( 5.1 b) or of the associated first feature ( 5.1) can thus be calculated. Here too it becomes clear that the evaluation becomes very simple by the use of two recordings, once with, once without flash. The features can be clearly determined in their distance. In principle, it is sufficient if only one edge of the flash strip (5.1b) is used. If the top edge were to be used as shown, the flash could eventually become wider downward and eventually illuminate the plane (4). However, it would also be possible to use the lower edge of the flash strips on the features just as well. Then, the flash would possibly become much higher in the upward direction and possibly also illuminate the features (5.1, 5.2, 5.4) in their entire height.Figures 5.1 and 5.2 show a similar scene to that previously. However, in this case, the feature (5) is located centrally on the plane (4), for example a roadway. In FIG. 5.1, the scene is irradiated with a flash from the right, so that a corresponding left shadow ( 9.1) of the feature ( 5) forms in the first recording.In FIG. 5.2, the scene is then flashed from the right for the next recording, so that a right shadow ( 9.2) is formed.FIG. 5.3 shows the result of the difference formation from the first recording to the second recording. In the difference image, only the left-hand shadow (9.1) and the right-hand shadow (9.2) are displayed. This can be done with sign or without sign consideration. The remainder is neutral and does not appear due to the difference formation. The known arrangement and orientation of the flash devices (1) and the camera (2) with respect to the plane (4) thus also allows the distance to the feature (5), e.g. an obstacle on the plane (4), to be found. At the known boundary of the recording (30), the distance and position or also the width of the feature can be determined from this via image line (17.3) and the image columns (18.1, 18.2).A feature (5), in this case a person, can be seen in FIG. 6.1 from the view of a camera (2), not shown, on a plane (4), e.g. on a travel path ahead of the horizon (15). The camera ( 2) is aligned with the horizon ( 15). However, as simple as illustrated here, it is not practical. Diffuse, low contrast surfaces and junctions make it difficult to find the interesting features in the environment.A remedy is provided by the use of the flash, as is illustrated with FIG. 6.2 and below. The flash device (1) is at approximately the same height as the camera (2) and is also aligned with the same scene. The flash light in particular illuminates the near area. This applies to the feature (5) but also to the front region of the plane (4). As the distance from the flash device (1) increases, but also transversely thereto, the brightness decreases, which is represented by the brightness boundary lines (19.1, 19.2, 19.3) with decreasing brightness.The areas illuminated by the flash device (1) are shown with greater or lesser shading in this example, corresponding to the intensity.However, it is disturbing in this orientation of the flash device (1) that not only the feature (5) but also the plane (4) are partly illuminated to the same extent, so that they cannot be directly separated from one another in the evaluation by the evaluator (3). This applies in particular when feature (5) and plane (4) have a similar degree of remission.Fig. 6.3 shows a step for eliminating problems. The effect of the flash is detected and stored by an exemplary pattern exposure of plane (4). The greater lightening in the front region of the plane ( 4) and the decreasing lightening with the distance through the brightness boundary lines ( 19.1, 19.2, 19.3) or the associated areas, which show a lower hatching with decreasing brightness, are indicated. If the floor covering or the surface of the floor covering of level (4) changes locally, it may be necessary to store several or a plurality of these position-characteristic lightenings in order to get a reference for the later, current situation.FIG. 6.4 now shows the result from the exemplary image with a feature (5) according to FIG. 6.2, taking into account the exemplary lightening at this location, as illustrated in FIG. 6.3. Both receptacles have now been removed from one another. However, only the upper part of the feature ( 5) can now be clearly seen on the level ( 4), which can now be determined relatively easily using a stereo camera or by a second flash recording from a changed flash distance in the distance. Downwards, towards plane (4), feature (5) becomes more and more ambiguous in contrast to the background.It is evident here that it would be more advantageous if necessary to let one or more flash devices ( 1) fall onto the scene from the right and / or left, so that shadows are then formed, as shown in FIGS. 5.1 to 5.3, which shadows can then in turn be subtracted from one another by comparing two recordings, for example with a right and a left flash. Then, in addition to the feature (5), the left and right shadows (9.1, 9.2) of the feature (5) would also be displayed and could be used as an indication of the position of the feature (5). A further possibility is also to mount a flash device (1) as low as possible and close to the plane (4) and to align it on the horizon (15), so that then the lightening by the flash device (1) on the plane (4) is negligible, but instead the feature (5) is lightened as before and thus rises more clearly above the plane (4).Another possibility for detecting features (5) on a plane (4) can be established according to FIG. 7.1 in that a camera (2) is aligned more or less directly to the plane (4), i.e. more or less perpendicular thereto. It can be seen that a roadway marking ( 14) is still located on the plane ( 4). The feature (5) is raised with respect to the plane (4) and thus has a certain height, while the road marking (14) is flat and has no orientation in this third dimension. Dark spots are indicated by hatching. When exposed from the bottom right by a flash (1), a left shadow (9.1) is produced.If, according to FIG. 7.2, a flash device (1) illuminates the scene from the bottom left, a right shadow (9.2) arises behind the feature.FIG. 7.3 shows the result of the difference between the two images from FIGS. 7.1 and 7.2. The surfaces that are different in both images are visible, i.e. the left and right shadows (9.1, 9.2). The application is interesting, for example, in order to provide raised, that is to say possibly disturbing, features ( 5), for example. To detect persons on a roadway. The non-disturbing flat pavement markings (14) and other perceptible contrasts which are flat will not be able to cast shadows and thus will not be recognized as obstacles.However, it should be noted that, due to the changed position of the flash device ( 1) (once left and once right), the plane ( 4) in front of the flash device ( 1) is also illuminated from left or right by more or less. This could be disruptive in the evaluation of two recordings to be compared. It is therefore recommended, similar to FIG. 6.3, to take into account and, if appropriate, calculate out the typical lightening of the plane (4) caused by the flash device (1).A similar orientation to that shown above is shown in FIG. 8.1. In addition to the raised feature (5), a flat disc (16) is also shown. First, it is assumed that the pane (16), the plane (4) and the feature (5) have the same reflectance and are diffusely illuminated. Thus, all in a photograph without flash device (1) have the same brightness. If a flash is then used and the plane ( 4) is illuminated more or less centrally perpendicularly, the plane ( 4) and the pane ( 16) in this example can show up in the flash with the same additional brightness. The values are thus "1+1". Feature (5), on the other hand, if it is only half the distance to flash device (1) due to its size, is illuminated by four times. The total brightness is then thus "1+4=5".FIG. 8.2 then shows the difference between the recording without flash and the recording with flash, i.e. the flash effect. It is clearly shown that the feature (5) is lifted off from the plane (4) although it has the same reflectance. The pane ( 16), on the other hand, cannot be separated from the plane ( 4) with this. It may also not be annoying in the case of a roadway.The example above shows how one can separate raised features (5) from plane (4), although both have the same reflectance. However, it looks different if the reflectance is different. Then the pure difference formation of two recordings will not necessarily lead to the goal.In FIG. 8.3, therefore, it is shown in the example how feature (5) and pane (16) are shown without lightning in front of the four-fold brighter plane (4). Here too, it is not possible to first of all to identify whether feature (5) or pane (16) is a raised obstacle.FIG. 8.4 then shows a photograph with flash. Feature (5) has been illuminated to the value "5" by the shorter distance from the flash device (1). The pane ( 16), on the other hand, which has the same distance as the roadway ( 4), could just like the roadway ( 4) double in brightness. However, the plane (4) has thereby achieved an absolutely higher brightness because of its fourfold higher reflectance.FIG. 8.5 then shows the values as a result of the lightening in the flash light. The two previous recordings were subtracted from one another for this purpose. However, it can be seen that feature (5) now appears as bright as plane (4) and thus would not have to be separated from one another again.Fig. 8.6 then shows the solution to the problem. In the corresponding regions with the corresponding pixels in the images, the relative lightening is formed by quotient formation Q. If the values of the regions or pixels of FIG. 8.5, i.e. of the pure lightning effect, are divided by FIG. 8.3, a clearly relative lightening is obtained for the feature (5) in comparison with the slight lightening of the pane (16) and of the plane (4). It is interesting that the slice (16) and the plane (4) show the same quotient, regardless of their reflectance. The comparatively four times higher quotient of the feature (5) can also be seen, which has only half the distance from the flash device (1) as the plane (4).If the height of the flash device (1) above the plane (4) is known, the height of features relative to the plane (4) can thus be calculated by forming the quotient.Fig. 9 shows another feature. In this case, it is an artificial landmark (35) which is partially equipped with retroreflective material (40) on at least two visible surfaces, namely the front surface (38) and a side surface (39). The surfaces can be coded differently, so that one can distinguish front surface (38) and side surface (39). However, further encodings can also be used to distinguish between a plurality of artificial landmarks ( 35), so that not only one to four but also far more information can be applied per page.The artificial landmark (35) has a visible width of the front view (27) and a visible width of the side view (28). From the relation of these two widths it can be estimated which direction the camera (2) has with respect to the artificial landmark (35). It would be advantageous if the artificial landmark ( 35) is not too narrow.Since it is already better if the artificial landmark ( 35) appears at a sufficient height ( 37) or at a sufficient and clear extent on the receptacle, of which only the receptacle boundaries ( 30) are shown here. If the artificial landmark (35) is large enough, the distance between camera (2) and artificial landmark (35) can be determined by the extent on the image. It is assumed here that the height (37) is known, because it is, for example, an artificial landmark (35). In principle, natural landmarks can be used in a similar manner, except that they are generally not encoded.However, for determining the distance, the representation of the upper or lower edge of the artificial landmark ( 35) in the image lines ( 17.4 or 17.3) can also be used. According to the laws of trigonometry, the distance of the artificial landmark ( 35) can be calculated with a known, e.g. horizontal, orientation of the camera ( 2) here. The higher the artificial landmark ( 35) appears in the image above the horizon plane of the camera ( 12) or of the corresponding image line ( 17.5), the smaller the distance to the artificial landmark ( 35).In this example, the artificial landmark (35) is mounted above the horizon plane of the camera (12). It also works accordingly if the mark is below the horizon plane of the camera (12).FIG. 10 shows the effect of reducing the distance of the camera ( 2) above a feature ( 5) in the plane ( 4) in front of the horizon ( 15). The feature (5) is shown at a distance of 1m, 2m and 3m as feature (5.1, 5.2 and 5.3). If the extent of the feature (5) in height and / or width is known, the distance to the feature (5) can easily be determined thereafter. However, it is also useful that the effect of a flash light at a distance of 1 m is approximately nine times as high as at a distance of 3 m. The determination of the distance when approaching is also made clear. The starting basis for the feature (5) is initially the mean position at 2 m. If now the system carrier (7) with flash device (1) and camera (2), not shown, approaches the feature (5) by one meter or, conversely, the feature (5) approaches the system carrier (7), the feature (5) appears twice as large in height and width at a distance of 1 m and is illuminated four times in the flash effect. In order to determine the correct brightness, it is important that a comparison is made with a photograph without flash.Regardless of the reflectance on feature (5), the flash effect at 1m will be four times as high as at 2m distance. The quotient of the flash effect is thus at half the distance "4" (flash effect 1m / flash effect 2m). The situation is different if the distance to the feature (5) increases, so that the feature (5) appears to be reduced in 3 mand the brightness has also correspondingly decreased.It is also interesting if the system carrier (7) is located on a vehicle, not shown, and this vehicle follows another vehicle at the same speed. If this other vehicle is likewise assumed as feature (5) as before, it will be established that this second vehicle is also represented with the second recording with flash with the same brightness.Thus, no difference in brightness will result in comparison with two images taken sequentially at different locations with flash and subtracted from each other. It looks as if the second vehicle did not exist on the roadway. In fact, this second vehicle does not present an obstacle to the following vehicle. The situation is harmless. The opposite is when the second vehicle approaches the first vehicle. Then it will appear much lighter in the second shot. This additional lightening by the approach of the second vehicle at its own speed makes the second vehicle appear closer than it is actually. Here, the dynamics of the lightening from the first to the second flash have a favourable effect because actually an approaching vehicle represents a higher risk than a stationary vehicle, not only because of the additional energy, but because the collision can also pass earlier.FIG. 11.1 represents symbolically the pixel values (22) which can be assigned in one picture to the associated pixels in the image rows (17) and the image columns (18.1, 18.2). A dark surface with low pixel values (22) can be seen at the bottom right, next to a light surface (at the top left) with high pixel values (22). Unfortunately, disturbances of various kinds result in not every pixel in the camera (2) being exposed as desired, but pixel values (22) varying in noise, i.e. subject to measurement technology, although the original has a cleanly separated surface for recording. If two images are then compared with one another, disturbances can result which lead to incorrect results.It is therefore advantageous to provide some compensation or mean for adjacent pixels. This can be done with each shot or after the comparison of two shots. It can be of great advantage to compensate, i.e. filter, the pixel values (22) with one another in columns or rows or also in both directions in smaller or larger groups.FIG. 11.2 then shows a column-by-column filtering. The right hand column of images (18.1) shows the surface with dark values and the left hand column of images (18.2) shows the lighter surface. It can also be seen that, in addition to further columns, two areas of different brightness must be concerned. The contrast between the two surfaces is also clear.Fig. 11.3 shows the document for recording. A dark feature (5) can be seen with a section of the image columns (18.1, 18.2). If the feature (5) were rotated somewhat, the image columns (18.1, 18, 2) would change in their course from one state to the other, i.e. from light to dark or vice versa.FIG. 12 finally shows an advantageous position and alignment of flash device ( 1) and camera ( 2). A feature (5), in this case a person, is visible in the boundaries of the image (30) on a plane (4). Camera (2) and flash device (1) are mounted on a system carrier (7) relatively close above the plane (4) and are aligned horizontally in such a way that the horizon (15) apparently ends with the lower limit of the recording (30) or the area lying underneath up to the plane (4) does not have to be viewed or evaluated further.This arrangement can then be used, for example. Obstacles above a roadway or in a hazardous area are recognized relatively elegantly and clearly. Here again, recordings with flash and without flash are advantageously used.The camera (2) is focused on the close range of interest, i.e. the background is only imperceptible or not perceived at all. Thus, no contrasts can form in the background either. If the evaluation only needs to accept sufficiently strong contrasts for the contour recognition of the features, it does not need to deal with the blurred, uninteresting background or other minor hazards.However, two recordings in comparison with and without lightning then represent the feature (5), i.e. here the person in the danger zone, very clearly. The flash creates a corresponding lightening with which the removal can be directly determined in the case of surfaces with a known degree of remission. Furthermore, the distance can also be determined by two flash devices which are arranged one behind the other, by means of the different illumination on the feature (5), or a stereo camera can be used very well here. If a first shot without flash does not yet provide contrasts and the next shot with flash also provides only partially suitable contrasts, however, the comparison or difference or quotient formation of shots without flash and with flash will more clearly show the feature (5) in its position. The evaluator (3) then has easy to determine the distance as well.Figure 13.1 illustrates a similar scene as previously Figure 12. The background is relatively dark in the lower region and relatively light in the upper region. For this purpose, the reflectance is likewise distributed in an unfavorable manner on the feature (5). Here too, the lower region is brighter than the upper region. In normal basic lighting, for example in daylight or in artificial light, the feature (5) is therefore not to be separated from the background. The darker regions are marked with the brightness value "1" and the lighter regions with the brightness value "2".Fig. 13.2 shows the change by the flash light. The background remains equally bright, but in feature (5) the brightness doubles according to the distance and the reflectance. Even now, the feature (5) would not yet be clearly separated from the background.FIG. 13.3 shows the difference between the two recordings, i.e. the pure lightning effect. The feature (5) is clearly distinguished from the background. A distance determination using stereo technology has already been possible and has been substantially simplified by the difference formation, since one can now directly concentrate on the feature contours.FIG. 13.4 now shows the further advantage of quotient formation Q. If Q is the lightening by the lightning effect divided by the basic brightness, Q is now represented as a distance-dependent value. It can be seen that the feature (5) has the same value "Q=1" both in the upper and in the lower region, i.e. despite different degrees of remission. The quotient formation also functions without stereo technology, but can advantageously be supplemented by the stereo technology.FIG. 14 shows the monitoring of a plane ( 4) by creating two images and comparing them, different camera positions being used for the images. A camera ( 2), in this case a stereo camera, can be seen, which has at least a first camera ( 2.1) and a second camera ( 2.2), which are arranged next to one another. Furthermore, a flash device ( 1) can be seen, with which potential features ( 5) above the plane ( 4) can be detected. This can be done on the one hand by an additional lightening or also by a shadow formation, which is perceived differently by the first camera ( 2.1) and by the second camera ( 2.2). It is advantageous that the position influences are first corrected for comparing the recordings from different positions. The individual contrasts ( 10.3, 10.4) on the plane ( 4) can be assigned to the corresponding pixels in the recordings. One reason for the necessary equalization lies in different camera positions. The contrasts can be any optically recognizable features which ideally lie on the plane (4). Normally, the contrasts (10.3, 10.4) on the plane (4) are no longer determined by the equalization after the difference formation, because they cancel one another out in the recordings.However, the situation is different if a raised contrast (10.1) or even a raised feature (5) above the plane (4) is detected. It can be seen that the first camera (2.1) assumes the raised contrast (10.1) at the first apparent contrast position (10.11).However, in comparison with the recordings, it will be seen with the recording of the second camera ( 2.2) that the difference is not zero at this point. The raised contrast (10.1) is assumed with the second camera (2.2) at the second apparent contrast position (10.21). The correct position is obtained by the evaluation. The height of the raised contrast ( 10.1) can be calculated from the deviation (disparity). The distance between the cameras (base), the distance to the plane (4) and the distance between the two apparent contrast positions (10.11, 10.21) result in the height of the raised contrast (10.1) above the plane (4).It can also be seen that the flash device (1) casts a shadow (9) on the plane (4) to the right next to the feature (5). This shadow ( 9) is, however, only present in the recording of the second camera ( 2.2) but not in the recording of the first camera ( 2.1). Thus, with the flash device (1) not only a different, relative lightening of the features at a smaller distance with respect to the underlying plane (4) is possible, but also a shadow formation, which likewise leads to the recognition of the raised features and enables a distance determination with the stereo camera. The necessary contrast is formed between the lightened surface of the feature (5) and the shadow (9). The width of the shadow (9) can also be used to estimate or calculate the distance between flash device (1) and feature (5).In the shadow there is also a partially covered contrast (10.3). This contrast is only visible by the second camera ( 2.2) and is thus useful in detecting the raised feature ( 5).It is possible that the feature (5) is part of the plane (4) or a known scene to be controlled. The surface of this scene is used as a reference setting for the shot equalization. Changes in features or features now become apparent reliably and uncomplicated with the difference formation with the next recordings. The flash thus helps in all applications in which features that deviate from the reference scene or the reference environment are to be determined and, if appropriate, their pose is to be determined.The problem for stereo technology is the recognition of features that have the same brightness or, when color cameras are used, also the same color as the environment. They therefore initially do not form a usable contrast. It is evident here that the use of a flash offers several advantages.In the case of the reference formation, the regions which cannot be brought into agreement or deletion, for example the sides of an object which are seen only from one side or shadows and the like, can optionally be excluded for further observation.FIG. 15 shows advantages with stereo technology. A system carrier ( 7), e.g., a vehicle, having a stereo camera ( 2) and an evaluator ( 3) and a flash device ( 1) can be seen. The plane (4) is not completely planar, but is partially inclined and, by means of a curbstone (24), it also has a bend or different heights. The plane (4) consists, for example, of a roadway and a sidewalk. In the background, a vertical plane, in this case a wall ( 13), can be seen. Two features ( 5) are shown, namely a front feature ( 5.5), e.g. an obstacle, and a rear feature ( 5.6), e.g. a person, which, however, does not represent a risk of collision due to the greater distance.Furthermore, a distance-measuring receiver (26) can be seen which is aligned obliquely forwards with respect to the plane (4). The inclination of the plane (4) can be determined with this. The range-measuring receiver (26) can use the modulation or pulses of the flash device (1). Possibly, the range-measuring receiver (26) is also able to change its orientation so that it can assume individually selected orientation or can scan a line or even an entire area as required.With the advantageous embodiment of stereo technology, which uses a stereo camera with a right and left camera, a right and left image can now be produced and processed in such a way that, when the two images are formed differently, the features which lie in the vertical virtual cancellation plane ( 21) are cancelled.For this purpose, depending on the desired distance to the virtual cancellation plane ( 21), the recorded pixels of at least the left or right camera can be laterally displaced. A similar effect can also be achieved if the optical axes of the left or right camera are displaced with respect to one another in such a way that they meet in the virtual cancellation plane ( 21).It is now advantageous that features which are located in or near the cancellation plane ( 21) resolve completely or almost completely in comparison with or in the difference between recordings of the left and right cameras. On the other hand, a front feature (5.5) with very strong disparity in shorter distance is clearly shown. Behind the virtual cancellation plane ( 21), however, the disparity increases again with corresponding distance. Features on the wall ( 13) can then optionally also be recognized again. By shifting the optical axes of the cameras or by shifting the images of left and right cameras for the difference image, the cancellation plane ( 21) can be shifted.However, it is now additionally possible to focus the cameras in such a way that they only sharply image a restricted region in front of the extinction plane ( 21), i.e. also with a strong contrast gradient. The features that are further removed are then perceived only blurred. This is of great advantage if the evaluation is only concentrated on features shown in sufficiently sharp form. This eliminates many unimportant details. The recognition of the real features in the foreground then becomes easier and faster.It is now also possible to understand the need to adapt the distance to the cancellation plane ( 21). The distance to the virtual cancellation plane (21) should be as short as possible to exclude many unimportant features. However, it should also be long enough for important features in the hazardous area to be recognized in good time and for the necessary stopping distance to be sufficient.FIGS. 16.1 to 16.4 show the advantages of shifting the extinction plane ( 21) using at least two cameras. In order to make correct use of stereo technology, three cameras must actually be used. The principle is shown below using an example, but only with two cameras which are located next to one another and are aligned more or less parallel. Contrasts are used here, as can arise in FIGS. 3 and 4 with the features (5.1, 5.2, 5.4) during the transition from these features to the background.In FIG. 16.1a, the pixel levels can be seen in a section of one line each of a left camera and a right camera (line left camera (42), line right camera (41)). Both cameras are aligned parallel to the horizon. The transition of a contrast from the left light (multiple pixels with full value) to the right dark (multiple pixels without signal) is shown. Since both cameras are aligned in parallel, the left camera shows the contrast at a different location in the image than the right camera. The distance between both contrast maps is referred to as disparity (32).FIG. 16.1b shows how the difference formation between two recordings has an effect. If the right line is subtracted from the left, a negative signal arises in the disparity range.In FIG. 16.1, the transitions in the cameras or the difference formation at a distance of 1 m are shown. FIGS. 16.2, 16.3 and 16.4 show the same contrast with the same camera arrangement at the distances 2m, 4m and infinity. Since the angles at which the contrast appears to the optical axis become smaller and smaller at longer distances, a distance can no longer be determined at infinite distance with the stereo camera. The disparity (32) is zero here. The disparity (32) for 2m distance is only half as large as 1m distance, but twice as large as 4m distance. With the width of the disparity, the distance to the feature can be determined more or less accurately.In stereo evaluation, a displacement of the images with respect to one another can now be used in order to block out an uninteresting region, for example the remote features, so that the evaluation then only has to deal with the features of interest in the front region. This application is frequently important in obstacle detection.In addition, the focusing of the cameras can then also be directed at the foreground, so that the background is already diffused and therefore no disruptive contrasts are formed for the remote features. FIGS. 16.1 c dto 16.4 c d show the horizontal pixel-by-pixel displacement of two recordings with respect to one another. This measure can also be achieved by pivoting the optical axes relative to one another, so that they meet at a distance of 4 min this example. The cancellation plane (21) was thus realized at 4m distance. Here, the difference shows no signal, i.e. the feature has disappeared as it were. It should be noted that in FIGS. 16.4c and d, the features have a quasi negative disparity. The pages have now been interchanged, i.e. the difference shows a positive signal. In FIGS. 16.1e and f to 16.4e and f, the advantage of focusing on the 1m range is then shown. In FIGS. 16.1e and f, the contrasts are still clearly depicted; the flanks are very steep. The same feature at 2m distance is shown in Figures 16.2e and f. The flanks are already flatter. The transition is distributed over several pixels in this case. At a distance of 4 m, it can be seen more clearly in FIG. 16.3e that the flanks become even flatter. Since the disparity (32) is zero in any case here, focusing does not have an effect for this distance. In FIG. 16.4e, the clear flattening of the transition can be seen. In practice, the flattening may be even flatter. The result is shown in FIG. 16.4f. The difference signal has a very low amplitude. If difference signals with such a small amplitude are neglected, only the hard contrasts with high amplitude of the features remain in the foreground.The method presupposes that strong contrasts form in the foreground. If the scene without flash is predominantly gray in gray, the advantage of lightening by flash is thus exhibited. The contours of the features then contrast to the background.Fig. 17.1 shows the advantage of flashing features from different distances. In principle, by recording without flash and recording with flash, the distance determination is possible independently of the remission degree of the feature. However, this is true only when the illumination of the feature is uniform.In the upper part of the illustration, first the hard lighting by a lighting element (6), e.g. sun, or the ceiling lighting on a feature (5), in this case a lying cylinder, can be seen. This cylinder is recorded from the side by a camera (2). The view of the camera ( 2) is visible on the right side. The lower part lies in the shadow (value "0"), the brightness increases upwards up to the value "4". The illustration below now shows the illumination with a flash device (1). In the case of quotient formation Q (lightning action due to effect of the background brightness), an unsatisfactory result would then be obtained. In the upper region, lightning effect="4" / background brightness"4"would be the quotient Q="1". In the middle part, however, "4" / "1" = "4". It can thus be seen that this procedure does not continue because daylight is not uniformly distributed everywhere. Only when a further recording is produced from half the distance, for example, and the brightness is measured will it be determined that the object is displayed four times as bright in the pure flash light with the brightness value "16". Knowing the flash locations, e.g., 2m, between flash devices, one can determine that the feature was then also 2m away to the near flash location. The lightening decreases with the square of the distance.In FIG. 17.2, the top part shows how the hard illumination by the luminous element (6) is reflected by a glossy feature (5). Here too, the camera ( 2) views the feature ( 5) (a cylindrical body) from the side. Even if some stray light is still in space, it can also be seen that only a narrow strip on the cylindrical body illuminates with high brightness "8". The other regions are significantly darker. If the surface were specular, the narrow stripe would be even narrower and even brighter. In the case of a retroreflective surface, no light would be received in this position, or almost no light would be received depending on scattered light.The lower part of FIG. 17.2 shows the illumination with a flash device (1). Here too, the values from a remote position are shown in the left column. The values vary between "0" to "8". On bisecting the flash device-feature distance, these values will multiply. It can thus be shown that a distance determination is possible by three recordings even in the case of glossy, specular or retroreflective surfaces. With negligible basic brightness, two images with flash also suffice in principle.It can be explained with FIGS. 17.1 and 17.2 how the orientation and surface of the feature ( 5) can be determined by different flash device positions. Assuming that the luminous element (6) is also a flash device (1), the upper side of the feature (5) can be placed against the lower side thereof with flash from above. It becomes clearer if a further flash device is positioned from below, close to the roadway, for a further recording, or if the camera ( 2) is given a higher position, so that it looks at the feature ( 5) obliquely from above. Then, due to the brightness curve of the first and the second recording, possibly in difference with an image without flash, the feature will also be represented as a cylinder body. In the case of a spherical feature or even in the case of a person, a receptacle with a flash from the right and from the left would be advantageous, if appropriate, so that the sides are elevated.By means of this measure, the distance can be determined not only directly by lightening the flash, optionally supported by stereo technology, but also the orientation and profiling of the surface of a feature. The determination of the orientation of simple, geometrically writable bodies is relatively simple. However, complex surfaces can also be determined. The alignment of reflecting surfaces can thus also be determined in some cases since the position of the flash device ( 1) with respect to the camera ( 2) is known.In FIGS. 18.2 to 18.3, a plane (4), in this example a roadway, is illustrated. This plane (4) has a known position with respect to a camera (2). On this plane, a pavement marker (14) is located in the upper part, including a first raised feature (5.7) and a second raised feature (5.8) below. The raised features are e.g. three-dimensional bodies in which the cover surfaces have a certain height above the plane. The top surfaces and the roadway marking ( 14) are shown shaded. The camera ( 2) is a stereo camera that is composed of left and right cameras. The result of the recording of the right camera is shown in FIG. 18.1, the result of the left camera is shown in FIG. 18.2. In Fig. 18.1, the right sides (33.2) of the raised features (5.7, 5.8) are seen. The difference between the two recordings is now shown in FIG. 18.3. The left and right cameras are tilted with respect to each other such that the cancellation plane (21) is identical to the plane (4). Alternatively, the pixel images can also be shifted transversely with respect to one another in such a way that the extinction takes place in the plane ( 4). Thus, the roadway marking ( 14) can then no longer be seen. However, the first raised feature (5.7) does not completely cancel out. The right and left edges of the first raised feature (5.7) are represented with a corresponding width or disparity (32) depending on the height of the cover surface.The situation is different for the second raised feature (5.8) which is as high. If this feature is known and is also expected in its position, e.g. a bollard or a curb, or in another example a machine part in stationary monitoring, this feature can receive a different cancellation plane ( 21), namely its top surface. Then, in this affected area, the pixels of the left or right or both camera photographs are shifted transversely to each other, so that they have a similar correspondence to the lane marking ( 14). Thus, the second raised feature (5.8) then also disappears in the difference image.In this example, it is shown that a plurality of cancellation levels can be realized in an area to be monitored. The extinction planes can also be inclined or curved, even having any desired shapes.

Claims

Method for detecting features in an environment, in which at least one camera, a flash device and an evaluation device are used, in which at least four recordings are produced with substantially the same orientation, wherein at least two different camera locations and different illuminations are used for the recordings and in which positions of the features are determined by a comparison of the at least four recordings and a mutual checking of results of different recordings is carried out by a combination of different recordings of the features, • wherein at least two recordings with flash light are produced from the features as a first recording, and by a different lightening in the at least two recordings, in which the flash device is once further away from the object and once closer to the object, the distances of the features to the flash device and / or to the camera are determined by comparing the different images with the flash light and with known distances between a first location of the flash device and a second location of the flash device, such that the distances to the features are determined by flashing once from a shorter distance and once from a longer distance to the features, and • wherein as a second detection a size of the features is used to determine their distances by producing two images with two different camera locations, wherein once the camera is closer, once further away from the features, and • wherein as a third detection an image with flash light is produced, wherein the flash device has a different distance than in the first flashed image and wherein of the images with flash light having a different distance is produced, a basic illumination which is determined from a photograph without flash light and the distances to the features are determined by quotient formation.Method according to Claim 1, characterized - in that at least two recordings of at least two camera locations are made from a scene - in that, for a comparison of the recordings, an equalization of at least one recording is carried out in such a way that, when recordings are formed with a difference, an extinction takes place at a desired distance and in that features which have a different distance from the camera positions are detected by non-extinction.Method according to Claim 1 or 2, characterized - in that the positions of the features are detected by a shadow formation of features above a plane caused by a flash of the flash device.Method according to at least one of the preceding claims, characterized - in that the positions of the features are detected by lightening features above a known plane caused by the flash of the flash device.Method according to at least one of the preceding claims, characterized - in that a plurality of recordings are made with flash light from at least one camera and - in that the distance of the camera from the features is changed by a known amount.Method according to at least one of the preceding claims, characterized - in that narrowband flash light is used.Method according to at least one of the preceding claims, characterized - in that flashing is carried out with infrared light and / or ultraviolet light.Method according to at least one of the preceding claims, characterized - in that a filter is used in front of at least one camera.Method according to at least one of the preceding claims, characterized - in that laser, ultrasound and / or radar are additionally used for further feature detection.Method according to at least one of the preceding claims, characterized - in that at least one camera is triggered in such a way that it uses interruptions and light phases of an extraneous light.Method according to at least one of the preceding claims, characterized - in that at least one recording is adjusted in terms of brightness for the comparison of recordings.Method according to at least one of the preceding claims, characterized - in that the at least one camera is mounted predominantly perpendicularly above a plane and raised features are detected above the plane.Method according to at least one of the preceding claims, characterized - in that errors in a distribution of the flash light are corrected.Method according to at least one of the preceding claims, characterized - in that pixels in the recordings and / or after the comparison of the recordings with adjacent pixels are evaluated and / or filtered.Method according to at least one of the preceding claims, characterized - in that recordings or comparisons of recordings are digitized.Method according to at least one of the preceding claims, characterized - in that retroreflectors are distinguished from luminaires by comparing recordings with and without flash light.Method according to at least one of the preceding claims, characterized - in that the flash light is directed substantially only onto a region to be examined.Method according to at least one of the preceding claims, characterized - in that the camera and the flash device are oriented such that they can capture features close to a remote background.Method according to one of the preceding claims, characterized - in that flashing is carried out with modulated light and the distance to features is determined with phase-measuring receivers.Method according to at least one of the preceding claims, characterized - in that the distance to features is determined from a propagation time of flash pulses and reflections with propagation time-measuring receivers.Method according to at least one of the preceding claims, characterized - in that the distance of an extinction plane is changed.

Images