Method and system for estimating depth information

Abstract

The invention relates to a method for determining depth information about image information in a vehicle (1) by means of an artificial neural network (2), comprising the following steps: - providing at least one emitter (3, 3') and at least one first and one second receiving sensor (4, 5), which are arranged spaced apart from each other (810); - emitting electromagnetic radiation (811) by means of the emitter (3, 3'); - receiving reflected components of the electromagnetic radiation emitted by the emitter (3, 3') by means of the first and second receiving sensors (4, 5) and generating first image information (B1) by means of the first receiving sensor (4) and second image information (B2) by means of the second receiving sensor (5) on the basis of the received reflected components (S12); - comparing the first and second image information (B1, B2) in order to determine at least one unevenly illuminated image region (D1, D2) in the first and second image information, which is generated by means of parallax due to the spaced-apart arrangement of the receiving sensors (4, 5) (S13); - analyzing geometric information of the at least one unevenly illuminated image region (D1, D2) and estimating depth information by means of the artificial neural network (2) on the basis of the result of the analysis of the geometric information of the at least one unevenly illuminated image region (S14).

Description

Technical Field

[0001] The present invention relates to a method and system for determining depth information of image information provided by an imaging sensor of a vehicle using an artificial neural network. Background Technology

[0002] In principle, it is known to use imaging sensors to detect the vehicle's surroundings in three dimensions. Alternatively, stereo cameras can also be used for 3D environment detection. To calculate distance information, image information from two cameras is correlated, and triangulation is used to determine the distance of image points from the vehicle.

[0003] Cameras used in stereo imaging systems are typically integrated into the front area of ​​a vehicle. In this case, the mounting location is usually the windshield area or the radiator grille. To generate sufficient brightness for image analysis at night, the vehicle's headlights are usually used.

[0004] The current problem with 3D environment detection lies in the fact that unevenly illuminated areas in the images acquired by the cameras of stereoscopic camera systems make it more difficult to determine depth information, because distance information cannot be obtained from these unevenly illuminated areas by the stereoscopic camera system. This is particularly true when shadows caused by parallax occur between the headlights and the camera due to different installation positions. Summary of the Invention

[0005] In view of this, the object of the present invention is to provide a method for determining depth information about image information, which can improve the determination of depth information.

[0006] According to a first aspect, the present invention relates to a method for determining depth information about image information in a vehicle using an artificial neural network. The neural network is preferably a convolutional neural network (CNN).

[0007] The method includes the following steps:

[0008] First, at least one transmitter and at least one first and second receiving sensor are provided. The transmitter may be adapted to emit electromagnetic radiation in the visible spectrum. Alternatively, the transmitter may emit electromagnetic radiation (for a radar transmitter) or laser radiation (for a lidar transmitter) in the infrared spectrum with a frequency range of approximately 24 GHz or approximately 77 GHz. The first and second receiving sensors are arranged spaced apart from each other. These receiving sensors are adapted to the transmitter type, i.e., they are adapted to receive the reflected component of the electromagnetic radiation emitted by at least one transmitter. Specifically, these receiving sensors may be adapted to receive electromagnetic radiation (for a radar receiver) or laser radiation (for a lidar receiver) in the visible or infrared spectrum with a frequency range of approximately 24 GHz or approximately 77 GHz.

[0009] Subsequently, electromagnetic radiation is emitted through the transmitter, and the reflected components of the emitted electromagnetic radiation are received by the first and second receiving sensors. Based on the received reflected components, the first receiving sensor generates first image information, and the second receiving sensor generates second image information.

[0010] Subsequently, the first and second image information are compared to determine at least one unevenly illuminated image region in the first and second image information, which is generated by parallax based on the spacing of the receiving sensors. If the first and second receiving sensors are not located at the projection center of the transmitter, particularly the headlight, the unevenly illuminated image region may also be generated by parallax between the respective receiving sensor and its corresponding transmitter. Therefore, in other words, at least one image region that is brighter or darker in the first image information than in the second image information is determined as an "unevenly illuminated image region".

[0011] Then, the geometric information of the at least one unevenly illuminated image region is analyzed, and based on the analysis results, depth information is estimated using an artificial neural network. Specifically, the size or extent of the unevenly illuminated image region is analyzed, as this allows the neural network to draw conclusions about the three-dimensional design of the object (e.g., the distance between a specific area of ​​the object and the vehicle is smaller than another area) or the distance between two objects within the vehicle's environmental area.

[0012] The proposed method has the technical advantage that, even in areas of uneven illumination where depth cannot be determined using triangulation, a neural network can deduce the distances between one or more objects in and / or around the unevenly illuminated image region based on the geometric information of the unevenly illuminated region. This enables more accurate and robust 3D environment detection relative to interference factors.

[0013] According to one embodiment, the unevenly illuminated image region is generated in the transition region between a first object and a second object, which are at different distances from the first and second receiving sensors. Therefore, the estimated depth information is depth difference information, which contains information about the distance difference between the first and second objects and the vehicle. This allows for better separation of foreground and background objects. Here, the foreground object is the object that is closer to the vehicle compared to the background object.

[0014] Furthermore, the unevenly illuminated image region can be associated with a single object, where the uneven illumination of the image region is generated based on the 3D design scheme of the single object. Therefore, the determination of the 3D surface shape of the object can be improved.

[0015] According to one embodiment, the transmitter is at least one headlight that emits visible light in the wavelength range of 380 nm to 800 nm, and the first and second receiving sensors are cameras, respectively. This allows the headlight located on the vehicle and the camera operating in the visible spectrum to be used as detection sensing mechanisms.

[0016] The first and second receiving sensors preferably form a stereo camera system. In this case, the image information provided by these receiving sensors is correlated with each other, and the distance of each pixel of the image information from the vehicle is determined according to the installation position of these receiving sensors. In this way, depth information about the image area detected by the two receiving sensors can be obtained.

[0017] According to one embodiment, at least two transmitters, each in the form of a vehicle headlight, are provided, and receiving sensors are respectively assigned to one headlight, such that the line of sight between the object to be detected and the headlight is substantially parallel to the line of sight between the object to be detected and the receiving sensor corresponding to the headlight. "Substantially parallel" here specifically means an angle of less than 10°. The receiving sensor can be positioned very close to the projection center of its corresponding headlight, for example, at a distance of less than 20 cm. Therefore, the illumination area of ​​the headlight and the detection area of ​​the receiving sensor are substantially the same, with virtually no parallax in the installation, thus uniformly illuminating the detection area of ​​the receiving sensor without any illumination shadows caused by the headlight corresponding to the receiving sensor.

[0018] According to one embodiment, the first and second receiving sensors are integrated into the vehicle's headlights. This makes the headlight's illumination area substantially the same as the detection area of ​​the receiving sensors. This achieves a completely or almost completely parallax-free installation.

[0019] According to one embodiment, an artificial neural network estimates depth based on the width of an image region with uneven illumination measured in the horizontal direction. The neural network is preferably trained to estimate depth information using the correlation between the width of the unevenly illuminated image region and the three-dimensional shape of the environment region represented by the image region. In this case, the horizontal width of the unevenly illuminated image region is particularly suitable for determining the depth difference between the unevenly illuminated image regions. Here, the depth difference may be related to a single contour object or to multiple objects, one of which (also called a foreground object) is located in front of another object (also called a background object).

[0020] Of course, in addition to the width of the unevenly illuminated image region measured in the horizontal direction, other geometric information and / or dimensions of the unevenly illuminated image region can be determined in order to estimate depth information. The aforementioned geometric information and / or dimensions can, in particular, be heights measured in the vertical direction or dimensions measured in an oblique direction (transverse to the horizontal direction).

[0021] According to one embodiment, the artificial neural network determines depth information in an image region detected by the first and second receiving sensors based on triangulation between image points in first and second image information and first and second receiving sensors. Preferably, the depth information is determined by the artificial neural network using triangulation. The artificial neural network also estimates depth information based on the geometry of unevenly illuminated image regions; that is, depth is determined by triangulation and the geometry of unevenly illuminated image regions is analyzed using the same neural network. By employing multiple different depth information determination mechanisms, 3D environment determination can be improved and made more robust.

[0022] According to one embodiment, the neural network compares the depth information determined by triangulation with the estimated depth information obtained by analyzing the geometric information of at least one unevenly illuminated image region, and generates adapted depth information based on the comparison. This advantageously eliminates triangulation errors, thereby obtaining more reliable depth information overall.

[0023] According to one embodiment, the artificial neural network adapts the depth information determined by triangulation based on the analysis of geometric information of at least one unevenly illuminated image region. That is, it modifies the depth information determined by triangulation based on the estimated depth information. This allows for more robust 3D environment determination.

[0024] According to one embodiment, infrared radiation, radar signals, or laser radiation are emitted by at least one transmitter. Correspondingly, at least a portion of these receiving sensors can be constituted as an infrared camera, a radar receiver, or a receiver for laser radiation. In particular, these receiving sensors are selected based on the at least one transmitter to which they correspond. Thus, when these receiving sensors correspond to an infrared transmitter, they are, for example, adapted to receive infrared radiation (IR). Specifically, transmitters and receiving sensors that do not emit light in the visible wavelength range can be used to detect environmental areas to the side or rear of the vehicle, since light in the visible wavelength range would affect other road users. This enables omnidirectional detection of at least a portion of the vehicle's surrounding area.

[0025] According to one embodiment, in order to estimate depth information regarding image information representing areas on the sides and / or rear of a vehicle, one or more transmitters and two or more receiving sensors are used to determine the image information. Multiple sensor groups are provided, each group having at least one transmitter and at least two receiving sensors, and the image information from each sensor group is merged into overall image information. This enables omnidirectional detection of at least a portion of the vehicle's surrounding environment.

[0026] According to one embodiment, the sensor array at least partially utilizes electromagnetic radiation from different frequency bands. For example, a stereo camera system can be used in the front area of ​​a vehicle, with the transmitter emitting light in the visible spectrum, while transmitters utilizing infrared or radar radiation can be used, for example, in the side areas of the vehicle.

[0027] According to another aspect, the present invention relates to a system for determining depth information about image information in a vehicle, the system comprising a computing unit performing computational operations of an artificial neural network, at least one transmitter adapted to emit electromagnetic radiation, and at least one first and second receiving sensors arranged spaced apart from each other. The first and second receiving sensors are adapted to receive reflected components of the electromagnetic radiation emitted by the transmitter. The first receiving sensor is adapted to generate first image information based on the received reflected components, and the second receiving sensor is adapted to generate second image information based on the received reflected components. The artificial neural network is adapted to:

[0028] - Compare first and second image information to determine at least one image region in the first and second image information that is unevenly illuminated, wherein the unevenly illuminated image region is generated by parallax based on the spacing layout of the receiving sensors;

[0029] - Analyze the geometric information of the at least one unevenly illuminated image region and estimate depth information based on the analysis results of the geometric information of the at least one unevenly illuminated image region.

[0030] If the first and second receiving sensors are not located at the center of the transmitter, especially the projection of the headlight, unevenly illuminated image areas may also be generated due to the parallax between the respective receiving sensor and its corresponding transmitter.

[0031] In this disclosure, "image information" refers to any information that can provide a multidimensional representation of the vehicle's environment. Specifically, it includes information provided by imaging sensors (such as cameras), radar sensors, or lidar sensors.

[0032] In this disclosure, "transmitter" refers to a transmitting unit suitable for emitting electromagnetic radiation. Examples include headlights, infrared radiators, radar transmitting units, or lidar transmitting units.

[0033] In this invention, the terms “approximately,” “substantially,” or “about” refer to a deviation of + / -10%, preferably + / -5%, from the exact value and / or a deviation in the form of a change that is not significant to the function.

[0034] Improvements, advantages, and uses of the present invention are also given in the following description and accompanying drawings of the embodiments. All features described and / or illustrated are, in principle, the subject matter of the present invention (either individually or in any combination). Attached Figure Description

[0035] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. Wherein:

[0036] Figure 1 This is an exemplary schematic diagram of a vehicle equipped with a stereo camera system adapted to detect objects located in front of the vehicle;

[0037] Figure 2 This is an exemplary schematic diagram of first image information acquired by a first detection sensor, in which two objects and an unevenly illuminated area in the transition region between the two objects can be identified;

[0038] Figure 3 This is an exemplary schematic diagram of second image information acquired by a second detection sensor, in which two objects and an unevenly illuminated area in the transition region between the two objects can be identified;

[0039] Figure 4 This is an exemplary schematic diagram of a vehicle having multiple sensor groups adapted to detect objects located in the vehicle's environmental area; and

[0040] Figure 5 This is an exemplary flowchart illustrating the steps of a method for determining depth information about image information using an artificial neural network. Detailed Implementation

[0041] Figure 1 An exemplary illustration shows a vehicle 1 equipped with a stereoscopic camera system. The stereoscopic camera system includes a first receiving sensor 4 and a second receiving sensor 5, which are, for example, image recording devices, particularly cameras. Furthermore, the vehicle 1 also has a first transmitter 3 and a second transmitter 3', which are, for example, constituted by the headlights of the vehicle 1. Therefore, the transmitters 3 and 3' are designed to emit human visible light, particularly light with wavelengths between 380 nm and 800 nm. Correspondingly, the receiving sensors 4 and 5 are designed to receive light within this wavelength range and provide image information. The first receiving sensor 4 specifically provides first image information B1, and the second receiving sensor 5 specifically provides second image information B2.

[0042] In order to analyze the image information B1 and B2 provided by the receiving sensors, the vehicle 1 has a computing unit 8, which is designed to analyze the image information B1 and B2. The computing unit 8 is specifically designed to generate depth information based on the image information B1 and B2 from at least two receiving sensors 4 and 5, so as to enable three-dimensional detection of the environment around the vehicle 1.

[0043] To analyze image information B1 and B2, an artificial neural network 2 is provided in the computing unit 8. The artificial neural network 2 is constructed and trained to calculate depth information about image information B1 and B2 using triangulation, and then checks or modifies this calculated depth information using depth information estimation. The depth information estimation determines unevenly illuminated image regions by comparing image information B1 and B2, analyzes the geometry or size of these image regions, and determines the estimated depth information based on this. The depth information calculated using triangulation can be adjusted based on this depth information.

[0044] Figure 1 A first object O1 and a second object O2 are shown located in front of vehicle 1 and can be illuminated by the headlights of vehicle 1. Receiver sensors 4 and 5 can receive a portion of the light emitted by the headlights reflected by objects O1 and O2.

[0045] Objects O1 and O2 are at different distances from vehicle 1. Furthermore, from the perspective of vehicle 1 and relative to the line of sight between objects O1 and O2 and receiving sensors 4 and 5, the second object O2 is located in front of the first object O1. The frontal view of the second object O2 towards vehicle 1 is, for example, separated from the frontal view of the first object O1 towards vehicle 1 by a distance Δd.

[0046] Due to the spacing of the transmitters 3 and 3' (here, the headlights of vehicle 1) and the receiving sensors 4 and 5, a brightness difference is generated in the first and second image information B1 and B2 due to parallax. That is, the image information B1 provided by the first receiving sensor 4 has a brightness difference in other areas compared with the second image information B2 generated by the second receiving sensor 5.

[0047] Figure 2 and Figure 3 This effect is illustrated exemplarily and schematically. Figure 2 An exemplary illustration shows first image information B1, provided by a first receiving sensor 4, which is positioned on the left side of vehicle 1 along the forward direction FR. Two unevenly illuminated image regions D1 and D2 are visible, generated such that the scenes depicted by these regions D1 and D2 are illuminated only by a single transmitter 3 and 3', respectively, and the first receiving sensor 4 views objects O1 and O2 from the front in a leftward-tilted viewing direction. Consequently, the width b (measured horizontally) of the unevenly illuminated image region D2 is greater than the width of the unevenly illuminated image region D1.

[0048] Figure 3 The second image information B2 provided by the second receiving sensor 5 is illustrated exemplarily. The second receiving sensor is located on the right side of vehicle 1 along the forward direction FR of vehicle 1. Two unevenly illuminated image regions D1 and D2 can also be identified in the second image information B2. These unevenly illuminated image regions are generated in such a way that the scene depicted by these image regions D1 and D2 is illuminated only by one transmitter 3 and 3' respectively, and the second receiving sensor 4 sees objects O1 and O2 from the front in a tilted rightward viewing direction. Therefore, the width b' (measured in the horizontal direction) of the unevenly illuminated image region D1 is greater than the width of the unevenly illuminated image region D2.

[0049] It should be noted that, due to the distance between the receiving sensors 4 and 5, a single transmitter 3 is sufficient to produce unevenly illuminated image areas D1 and D2 in the first and second image information B1 and B2. However, it is advantageous that each receiving sensor 4 and 5 corresponds to a transmitter 3 and 3', and these transmitters 3 and 3' are located in the vicinity of their respective receiving sensors 4 and 5, where "vicinity" specifically refers to a distance of less than 20 cm. The receiving sensors 4 and 5 are preferably integrated into the transmitters 3 and 3', for example, as a camera integrated into a headlight.

[0050] The neural network 2 is adapted to compare image information B1 and B2 to determine the image regions D1 and D2 that are unevenly illuminated, and to estimate depth information by analyzing the geometric differences between the unevenly illuminated image regions D1 and D2 in the first and second image information B1 and B2.

[0051] As described above, the neural network 2 is configured to determine the distance of the vehicle 1 from the detected scene regions via triangulation. These regions are visible through the first and second receiving sensors 4 and 5 and are thus visible in two image information sets B1 and B2. In this case, for example, the image information sets B1 and B2 are merged into a whole image, and depth information is calculated for the pixels of the whole image corresponding to the regions shown by the two image information sets B1 and B2.

[0052] In this case, the drawback is that it cannot target background objects (in Figure 2 and Figure 3 The region of object O1 that is invisible in both image information B1 and B2 due to parallax (in the image information B1). Figure 2 and Figure 3 Depth information is calculated by using regions D1 and D2 (which are unevenly illuminated).

[0053] However, through the estimation process of neural network 2, depth information can be estimated by comparing the geometric dimensions of the unevenly illuminated regions D1 and D2 in image information B1 and B2. Specifically, the widths of the unevenly illuminated regions D1 and D2, measured in the horizontal direction, can be used to estimate the depth information. For example, neural network 2 can determine the distance Δd between objects O1 and O2 based on the comparison of the geometric dimensions of the unevenly illuminated regions D1 and D2, i.e., how far in front of object O1 object O2 is positioned in the illustrated embodiment. The estimated depth information is thus obtained, and the depth information calculated by triangulation is corrected based on this depth information. This produces modified depth information, which is used for a three-dimensional representation of the vehicle environment.

[0054] If, for example, the distance Δd between objects O1 and O2 is calculated to be 2m at a specific pixel using triangulation, but the depth estimation based on the unevenly illuminated area only yields a distance of 1.8m between objects O1 and O2, then the depth information obtained by triangulation can be modified based on the estimated depth information, such that the modified depth information gives a distance Δd between O1 and O2 of 1.9m.

[0055] Of course, based on the comparison of the unevenly illuminated areas D1 and D2, it is also possible to determine which objects O1 and O2 these areas correspond to, thereby enabling depth estimation in areas that cannot be detected by the two receiving sensors 4 and 5.

[0056] To train neural network 2, training data in the form of image information pairs simulating the environment within a vehicle range can be used. In this case, the image information of these image information pairs represents the same scene from different directions, i.e., the scene is perceived from the detection positions of detection sensors 4, 5, 6, 6' spaced apart from each other. Furthermore, the image information of these image information pairs has image regions with non-uniform illumination, which are generated by at least one, preferably two, transmitters 3, 3'. In addition, depth information about the non-uniformly illuminated image regions is also present in these training data. This allows neural network 2 to be trained and its weighting factors to be adapted so that the depth information estimated based on the geometric information of these non-uniformly illuminated image regions approximates the actual depth information.

[0057] Figure 4 The vehicle 1 is shown, and multiple sensor groups S1-S4 are provided on the vehicle for collecting environmental information of the vehicle. Sensor group S1 is adapted to detect the environment in front of the vehicle 1, sensor group S2 is adapted to detect the environment to the right of the vehicle 1, sensor group S3 is adapted to detect the environment behind the vehicle 1, and sensor group S4 is adapted to detect the environment to the left of the vehicle 1.

[0058] Each of the sensor groups S1-S4 has at least one transmitter 6, 6', preferably at least two transmitters 6, 6', and at least two detection sensors 7, 7'.

[0059] As described above, the sensors in each sensor group S1-S4 generate three-dimensional local environmental information within their respective detection areas. The detection areas of sensor groups S1-S4 preferably overlap; therefore, the local environmental information provided by these sensor groups also overlaps. The local environmental information can be advantageously correlated to form overall environmental information, which may be, for example, a 360° omnidirectional environmental representation or a partial 360° omnidirectional environmental representation (e.g., greater than 90° but less than 360°).

[0060] Since lateral or rear illumination cannot be achieved using visible light similar to headlights, sensor groups S2 to S4 can emit electromagnetic radiation in the non-visible wavelength range, such as infrared radiation, radar radiation, or laser radiation. Therefore, transmitters 6 and 6' can be, for example, infrared transmitters, radar transmitters, or lidar transmitters. In this case, receiving sensors 7 and 7' are adapted to the radiation from corresponding transmitters 6 and 6', i.e., infrared receivers, radar receivers, or lidar receivers.

[0061] Figure 5 A flowchart illustrating the steps of a method for determining depth information about image information using an artificial neural network 2 in vehicle 1.

[0062] First, at least one transmitter and at least one first and second receiving sensors are provided (S10). In this case, the first and second receiving sensors are arranged in a spaced-apart manner.

[0063] Subsequently, electromagnetic radiation is emitted via a transmitter (S11). This can be, for example, light in the visible spectrum, light in the infrared spectrum, laser, or radar radiation.

[0064] Subsequently, the reflected components of the electromagnetic radiation emitted by the transmitter are received by the first and second receiving sensors, and based on the received reflected components, first image information is generated by the first receiving sensor, and second image information is generated by the second receiving sensor (S12).

[0065] Then, the first and second image information are compared to determine at least one image region with uneven illumination in the first and second image information (S13). In this case, the image region with uneven illumination is generated by parallax based on the spacing layout of the receiving sensors.

[0066] Subsequently, the geometric information of at least one unevenly illuminated image region is analyzed, and based on the analysis results of the geometric information of at least one unevenly illuminated image region, depth information is estimated by an artificial neural network (S14).

[0067] The invention has been described above through embodiments. Of course, many modifications and variations can be made without departing from the scope of protection defined by the patent claims.

[0068] Appendix Label Table

[0069] 1 vehicle

[0070] 2 Neural Networks

[0071] 3 First Launcher

[0072] 3' Second launcher

[0073] 4 First receiving sensor

[0074] 5 Second receiving sensor

[0075] 6, 6' transmitter

[0076] 7' Receiver sensor

[0077] 8 Computing Units

[0078] b, b' width

[0079] B1 First Image Information

[0080] B2 Second Image Information

[0081] Areas with uneven illumination (D1, D2)

[0082] Δd Spacing / Distance

[0083] O1 First object

[0084] O2 Second object

[0085] S1-S4 sensor group

Claims

1. A method for determining depth information about image information in a vehicle (1) using an artificial neural network (2), the method comprising the following steps: - Provide at least one transmitter (3, 3', 6, 6') and at least one first and one second receiving sensor (4, 5, 7, 7'), wherein the first and second receiving sensors (4, 5, 7, 7') are arranged in a manner spaced apart from each other; - Electromagnetic radiation is emitted through the transmitters (3, 3', 6, 6'); - The reflected components of the electromagnetic radiation emitted by the transmitter (3, 3', 6, 6') are received by the first and second receiving sensors (4, 5, 7, 7') and based on the received reflected components, first image information (B1) is generated by the first receiving sensor (4, 7) and second image information (B2) is generated by the second receiving sensor (5, 7'). - Compare first and second image information (B1, B2) to determine at least one unevenly illuminated image region (D1, D2) with a brightness difference in the first and second image information, the at least one unevenly illuminated image region with a brightness difference being generated by parallax based on the spacing of the receiving sensors (4, 5, 7, 7'); - Analyze the geometric information of the at least one unevenly illuminated image region (D1, D2) with brightness difference and estimate the depth information through the artificial neural network (2) based on the analysis results of the geometric information of the at least one unevenly illuminated image region with brightness difference.

2. The method according to claim 1, characterized in that, The unevenly illuminated image regions (D1, D2) with brightness differences are generated in the transition region between the first object (O1) and the second object (O2), which are at different distances from the first and second receiving sensors (4, 5, 7, 7'), and the estimated depth information is depth difference information, which contains information about the distance difference between the first and second objects (O1, O2) and the vehicle (1).

3. The method according to claim 1 or 2, characterized in that, The transmitter (3, 3') is at least one headlight that emits visible light with a wavelength range between 380 nm and 800 nm, and the first and second receiving sensors (4, 5) are cameras, respectively.

4. The method according to claim 3, characterized in that, The first and second receiving sensors (4, 5) form a stereo camera system.

5. The method according to claim 1, characterized in that, The vehicle (1) is provided with at least two transmitters (3, 3') of headlights, and receiving sensors (4, 5) are respectively corresponding to the headlights (3, 3'), such that the line of sight between the object to be detected (O1, O2) and the headlights is parallel to the line of sight between the object to be detected (O1, O2) and the receiving sensors (4, 5) corresponding to the headlights.

6. The method according to claim 1, characterized in that, The first and second receiving sensors (4, 5) are integrated into the headlights of the vehicle (1).

7. The method according to claim 1, characterized in that, The artificial neural network (2) estimates the depth based on the width (b) of the unevenly illuminated image regions (D1, D2) with brightness differences measured in the horizontal direction.

8. The method according to claim 1, characterized in that, The artificial neural network (2) determines depth information in the image region detected by the first and second receiving sensors (4, 5, 7, 7') based on the triangulation between the image points in the first and second image information (B1, B2) and the first and second receiving sensors (4, 5, 7, 7').

9. The method according to claim 8, characterized in that, The neural network (2) compares the depth information determined by triangulation with the estimated depth information obtained by analyzing the geometric information of the at least one non-uniformly illuminated image region (D1, D2) with brightness difference and generates adapted depth information based on the comparison.

10. The method according to claim 8 or 9, characterized in that, The artificial neural network (2) adapts the depth information obtained by triangulation based on the analysis of the geometric information of the at least one unevenly illuminated image region (D1, D2) with brightness difference.

11. The method according to claim 1, characterized in that, Infrared radiation, radar signals, or laser radiation are emitted through at least one transmitter (6, 6').

12. The method according to claim 11, characterized in that, At least a portion of the receiving sensors (7, 7') is an infrared camera, a radar receiver, or a receiver for laser radiation.

13. The method according to claim 1, characterized in that, In order to estimate depth information about image information representing the side region and / or rear region of the vehicle (1), one or more transmitters (3, 3', 6, 6') and two or more receiving sensors (4, 5, 7, 7') are used to determine the image information, wherein multiple sensor groups (S1, S2, S3, S4) are provided, each sensor group having at least one transmitter and at least two receiving sensors, and wherein the image information of each sensor group (S1, S2, S3, S4) is combined into overall image information.

14. The method according to claim 13, characterized in that, The sensor group (S1, S2, S3, S4) utilizes electromagnetic radiation of different frequency bands at least in part.

15. A system for determining depth information about image information in a vehicle (1), the system comprising a computing unit (8) performing computational operations of an artificial neural network (2), at least one transmitter (3, 3', 6, 6') configured to emit electromagnetic radiation, and at least one first and second receiving sensors (4, 5, 7, 7') arranged spaced apart from each other, wherein the first and second receiving sensors (4, 5, 7, 7') are configured to receive reflected components of the electromagnetic radiation emitted by the transmitter (3, 3', 6, 6'), and wherein the first receiving sensor (4, 7) is configured to generate first image information (B1) based on the received reflected components, and the second receiving sensor (5, 7') is configured to generate second image information (B2) based on the received reflected components, wherein the artificial neural network (2) is configured to: - The first and second image information (B1, B2) are compared to determine at least one unevenly illuminated image region (D1, D2) with a brightness difference in the first and second image information, wherein the unevenly illuminated image region (D1, D2) with a brightness difference is generated by parallax due to the spacing of the receiving sensors (4, 5, 7, 7'). - Analyze the geometric information of the at least one unevenly illuminated image region (D1, D2) with brightness difference and estimate depth information based on the analysis results of the geometric information of the at least one unevenly illuminated image region (D1, D2) with brightness difference.

Images