Method, computing unit and computer program for calculating a humidity parameter

Abstract

The invention relates to a method (400) for calculating a humidity parameter, wherein an image (300) of a carriageway (100, 220) is recorded by means of a camera (210), the image (300) is divided into a plurality of regions (330) and for each of the regions (330) a texture frequency value in a first direction transverse to the carriageway (100, 220) and / or in a second direction parallel to the carriageway (100, 220) is determined, and a humidity parameter is calculated on the basis of the determined texture frequency values. A computing unit for carrying out such a method (400) and a computer program are likewise subject matter of the invention.

Description

Technical Field

[0001] The present invention relates to a method for calculating a humidity parameter, a calculation unit for performing the method, and a computer program. Background Technology

[0002] In modern vehicles, especially those equipped with assistance systems for autonomous or self-driving driving, cameras are often installed, typically oriented towards or opposite to the direction of travel. Such cameras can be used to determine whether a lane on which vehicles are moving is wet. For example, image data generated by the camera can be evaluated, and it can be determined whether a gibbous cloud (water droplet cloud) is formed behind a vehicle traveling ahead. If so, the lane is automatically identified as wet.

[0003] If there are no vehicles traveling ahead, a similar method can be used to identify wet lanes by using a camera pointed behind the vehicle in the opposite direction of travel, which detects water droplets generated by the vehicle itself.

[0004] These methods work particularly well on very wet driveways, such as when there are puddles in the tire area.

[0005] In principle, such methods are designed to improve driving safety, for example, by setting a distance assist system based on the expected braking distance; or by operating the vehicle's drivetrain in such a way that the drive torque is matched to the expected frictional resistance.

[0006] Even if there are no puddles on the roadway, but there is residual moisture, particularly in the tire area, that is drying, the braking distance may still be longer compared to a completely dry roadway. Moisture outside the area the tires normally travel over can also be problematic, for example, during overtaking or evasive maneuvers where the vehicle will at least temporarily leave the area. Summary of the Invention

[0007] According to the present invention, a method for calculating humidity parameters, having the features of the independent patent claims, and a computing unit and computer program for performing said method are provided. Advantageous designs are the subject matter of the dependent claims and the description thereof below.

[0008] According to the present invention, a method for calculating humidity parameters is proposed, wherein an image of a driving lane is recorded by means of a camera, a texture frequency or texture frequency value is determined for a region of the driving lane in a first direction transverse to the driving lane and / or a second direction parallel to the driving lane based on the image, and a humidity parameter is calculated based on the determined texture frequency. Such a texture frequency specifically characterizes the amount of texture transformation (particularly dry / wet) per unit length. For example, texture changes can be identified based on contrast transformation.

[0009] Preferably, the texture is determined based on the color values, brightness values, and / or intensity values ​​of the image. For example, wet areas are typically darker and / or more intensely colored than dry areas of the same substrate. Therefore, this texture determination allows for differentiation between wet and dry areas.

[0010] The area mentioned here specifically includes stripes in a first direction transverse to the lane and / or a second direction parallel to the lane. Preferably, the texture frequency along the strip direction is determined. Therefore, it is possible to distinguish the characteristic pattern caused by humidity on the lane from other patterns with high reliability.

[0011] Preferably, for each texture frequency value determined in one direction of the region, a frequency value is calculated that describes the relative frequency of the texture frequency value along the corresponding other direction in several regions. Thus, texture frequencies not caused by moisture can have a smaller impact on the magnitude of the humidity parameter.

[0012] Advantageously, each texture frequency value is considered in a weighted manner using its corresponding frequency value when calculating the humidity parameter. This minimizes the error variability of the method, because only texture frequency values ​​appearing in many evaluated areas will have a significant impact on the humidity parameter.

[0013] Advantageously, when calculating the humidity parameter, a first texture frequency range with a positive correlation coefficient and / or a second texture frequency range with a negative correlation coefficient are considered. For example, texture frequencies indicating plant shadows, i.e., texture frequencies within a frequency range corresponding to leaf size, are therefore unlikely to cause an increase in the humidity parameter. Simultaneously, these frequency ranges can be used to ensure that texture frequencies present in the vicinity of these frequency ranges, for example, those possibly caused by the shadow of a tree trunk, are correctly distributed and similarly do not lead to an increase in the humidity parameter. Thus, the humidity parameter calculated in this manner more reliably reflects the actual humidity conditions.

[0014] Advantageously, vehicle functions can be controlled based on the calculated humidity parameters. Therefore, the calculated road humidity can be taken into account to improve overall driving safety.

[0015] Preferably, the vehicle functions include one or more of braking force adaptation, emergency braking assist, centrifugal protection, transmission control, engine control, warning output, and emergency call. These vehicle functions are particularly suitable for improving road traffic safety or minimizing damage caused by wet roads.

[0016] As mentioned earlier, frequency calculations are performed region-by-region. For this purpose, the image of the driving lane is divided into multiple regions, such as strips, which are advantageously arranged parallel to each other and perpendicular to the direction of travel. Texture frequencies are then calculated individually for each of these regions.

[0017] To this end, for example, one could count the image segments that terminate within the region, each segment having similar or identical brightness, color depth, or hue within that segment and different brightness, color depth, or hue between segments. The number of these segments could then be divided by a length unit characterizing the region, such as the length or width of the region in question, to indicate the texture frequency per length unit, for example, in meters. -1 Texture frequency in units of meters.

[0018] In a similar manner, texture frequencies parallel to the driving lane can be calculated. For this purpose, the image of the driving lane is advantageously divided into strips that are parallel to each other and parallel to the driving lane.

[0019] The present invention utilizes the understanding that characteristic patterns can be identified in both wet and drying driveways, and these characteristic patterns are generated, in particular, by using the driveways.

[0020] If this disclosure refers to a driveway being identified as wet, damp, or drying, this should be understood as calculating a high humidity parameter. In some designs of the invention, a distinction can be made between wet, damp, and / or drying driveways. In such cases, the following order of humidity parameter magnitude applies: wet > damp > drying. In other words, the humidity parameter is positively correlated with the proportion of the driveway area that is wetted.

[0021] For example, when a lane is drying, tire tracks tend to dry earlier, while the areas between tire tracks and the areas on either side of the lane edges remain wet for a longer period. This results in a pattern where bright and dark areas alternate in the direction transverse to the lane. Here, wet areas are generally darker than dry areas, especially on asphalt lanes. If areas of the lane where tires have generally rolled are dry, while areas outside the tire tracks remain wet or damp, then five lane areas are arranged parallel to each other in the direction of travel. Therefore, when the lane width is between 2.75 meters and 3.75 meters, the frequency transverse to the direction of travel is 0.5 meters. -1 up to 0.75m -1 Therefore, it is possible to infer, for example, that a road is drying out based on such frequency values.

[0022] Advantageously, the humidity parameter is not calculated based on the texture frequency of a single region, particularly (here, horizontal) image stripe. Instead, multiple regions are evaluated, and a particular texture frequency or range is considered in calculating the humidity parameter only if it occurs at a certain frequency, i.e., at least in a predetermined minimum number of evaluated regions. Thus, random fluctuations in the image data or moisture-independent textures of the roadway are not considered in the calculation and therefore do not distort the results. Consequently, the calculation of the humidity parameter becomes more robust, for example, to damaged areas in the roadway, which, while potentially causing corresponding texture frequencies in various regions, do not allow for valid conclusions about the presence of water in the roadway.

[0023] Texture should be determined at a resolution of at least 25 cm to reliably identify drying lanes. However, a resolution up to 2.5 cm may also be advantageous to identify, for example, the shadows cast by trees. If the frequency in the range of 10 to 20 per meter is above a threshold, the difference in brightness can be attributed to the shadows cast by vegetation.

[0024] Preferably, the identification is performed only when sunlight is present, i.e., only when the total brightness, measured by, for example, a camera or a brightness sensor, exceeds a certain brightness threshold. If the plant shadow itself is identified in this way, this can also be considered in the case of texture frequencies in a lower frequency range, where the texture frequencies are used in principle for calculating the humidity parameter. However, shadows such as those of tree trunks in areas adjacent to regions with such high frequencies may erroneously lead to the calculation of high humidity parameters, while these shadows can also result in low texture frequency values ​​in image analysis. Therefore, it is advantageous to identify the detection of plant-induced texture frequency accumulation as plant-induced texture frequency accumulation without causing an increase in the humidity parameter.

[0025] Unevenness and ruts in the roadway can create puddles and ditches that dry significantly later than raised areas of the road surface. In the case of asphalt roads, unevenness is often caused by heavy vehicles that remain stationary for extended periods in high temperatures or transfer large forces to the roadway in the longitudinal and / or lateral directions. For this reason, unevenness and the resulting wet spots on the road surface typically have a dimension or spacing similar to the wheelbase or track width of the corresponding vehicle. The wheelbase of heavy-duty trucks is typically between 1m and 2m, which corresponds to 0.5m parallel to the roadway width. -1 Up to 1m -1 The corresponding texture frequency range.

[0026] Furthermore, wheel speed sensor systems, inertial sensor systems, tire pressure sensor systems, and level sensor systems can be used to detect such uneven ground. Therefore, if the frequency between 0.5 and 1 per meter after converting such sensor signals to the frequency domain also exceeds the frequency threshold in the interpreted sense, then the road section identified as wet or drying can be confirmed or its credibility can be verified.

[0027] In the case of concrete driveways, another regular accumulation of moisture or humidity should be identified. These concrete driveways must be cut at regular intervals, for example, 4m to 12m, to allow for expansion due to temperature differences without concern for material spalling (so-called bulging). Over time, the resulting slabs sink more strongly in some areas than others, thus creating puddles whose dimensions are strongly correlated with the slab size. If the frequency of the texture is above a threshold between 0.08 and 0.25 per meter, particularly in the direction parallel to the driveway, the road is identified as wet or drying, and thus a high humidity parameter is determined.

[0028] If, after converting the signals from the wheel speed sensor system, inertial sensor system, tire pressure sensor system, or level sensor system to the frequency domain, the frequency between 0.08 and 0.25 per meter also exceeds the frequency threshold, then the road section identified as wet or drying is confirmed or its credibility is verified in a manner similar to the aforementioned uneven roadway situation.

[0029] When the road is very wet, tracks will form at the vehicles. These tracks at the vehicles ahead can be identified using a front-facing camera, and the tracks at your own vehicle can be identified using a reversing camera. These tracks have defined widths and defined spacings corresponding to the tire width or the vehicle's spurweite. For vehicles ahead, the spurweite is typically a few centimeters smaller than the vehicle width. Your own vehicle's spurweite is constant and known, and can be, for example, 1.6m to 2m. If the frequency of texture in the lateral direction within a frequency range of 0.5 to 0.7 per meter is higher than a threshold, the road can be identified as very wet accordingly. However, as mentioned above, even when the lane is drying, a cumulative texture frequency can be identified in this area. However, even when the road is very wet, tracks that almost exactly correspond to the width of the tire's rolling surface should be expected. Your own vehicle's tire width is constant and known, and can be, for example, 0.15m to 0.3m. If the frequency of texture lateral to the lane within a frequency range of 3 to 7 per meter is higher than a threshold, the road is identified as very wet. Especially when using a camera mounted at the rear of the vehicle to record images, the frequency range can be set very precisely based on the width of the tires, resulting in a significant improvement in accuracy when calculating the humidity parameter.

[0030] Texture frequency values ​​alone are insufficient to distinguish, for example, dry tire tracks between puddles in ruts and still-wet sections of a roadway, as they can have nearly identical spatial dimensions. However, wet areas typically have a darker and / or more intense color compared to dry areas on the same surface. Therefore, it is particularly preferable to consider the relative color values, relative brightness values, and / or relative intensity values ​​of image segments within the region and / or the contrast values ​​between these image segments when calculating the humidity parameter. This allows for the differentiation between drying roadways, wet roadways, and very wet roadways, which can positively impact driving safety.

[0031] For example, a corresponding threshold can be selected based on the average illumination intensity. Similarly, it is conceivable to select one or more thresholds based on the current road surface. The current road surface can be identified, for example, by means of measured ultrasonic noise levels, since the ultrasonic noise level on sound-damping asphalt is lower than that on ordinary asphalt. It is also conceivable to use the visual frequency frequencies described herein, particularly those in the very high frequency range, to determine the current road surface, since different road surfaces have, for example, different aggregate particle groups, and therefore different texture frequencies should be observed.

[0032] In addition, to identify each road condition, thresholds and frequency ranges can be specified individually for each road segment in the map material and transmitted to the vehicle, for example, via a mobile radio connection.

[0033] Since the described humidity-induced patterns each have a distinct preferred direction, either transverse or parallel to the roadway, it can be inferred from the isotropic frequency distribution that the detected texture is not caused by humidity or moisture in the roadway. Therefore, if very similar texture frequency distributions are calculated both parallel and transverse to the roadway, these texture frequency distributions are preferably not used to calculate the humidity parameter, or used in such a way that they do not increase the humidity parameter.

[0034] The computing unit according to the invention, such as the control device of a motor vehicle, is particularly configured in terms of programming technology to execute the method according to the invention.

[0035] It is also advantageous to implement the method according to the invention in the form of a computer program or computer program product having program code for performing all method steps, as this results in particularly low costs, especially when the control device performing the execution is also used for other tasks and therefore exists anyway. Suitable data carriers for providing the computer program are, in particular, magnetic, optical, and electrical memories, such as hard disks, flash memory, EEPROM, DVDs, etc. The program can also be downloaded via computer networks (Internet, intranet, etc.).

[0036] Other advantages and designs of the present invention will become apparent from the description and drawings. Attached Figure Description

[0037] The invention is illustrated schematically with reference to the embodiments in the accompanying drawings, and the invention will now be described with reference to the accompanying drawings.

[0038] Figure 1 The road is shown in a diagram.

[0039] Figure 2 A vehicle is shown that is configured to perform the method according to the invention.

[0040] Figure 3 The images are schematically shown within the advantageous design scope of the method according to the invention.

[0041] Figure 4 An advantageous design of the method according to the invention is illustrated in the form of a flowchart. Detailed Implementation

[0042] Figure 1 The road 100 shown has lane markings 130 and textured patterns 110 and 120 caused by moisture.

[0043] In a first scenario where road 100 is drying, the first lane area 110, where the tires of vehicles frequently using road 100 move, has dried, while the second lane area 120, which is rarely driven over, remains wet. Thus, on an asphalt roadway, the still-wet second area 120 typically appears a darker color compared to the brighter, dried first area 110. In an advantageous design of the method according to the invention, lane markings 130 can be used, for example, to determine lane width and road direction. Therefore, even when the road direction is curved, texture analysis can be performed based on the recorded image data, as will be further explained below.

[0044] In the second scenario, road 100 has ruts 110 in the first lane area 110, while the second lane area 120, which is rarely traversed by vehicle tires, has a raised asphalt surface. Therefore, after a rainfall event, water accumulates in the ruts 110, causing them to remain submerged for a longer period compared to the raised lane area 120. Consequently, in the second scenario, the ruts 110 are darker than the second area 120 when road 100 is partially wet.

[0045] like Figure 3 As shown, if an image 300 recording road 100 or an image 300 recording the lane 220 of road 100 is shown, then strip patterns 110, 120 extending substantially parallel to the road can be identified as textures in the corresponding image data. To infer the wetness of road 100, image texture is evaluated, preferably in an automated manner. For this purpose, for example, an image 300 of road 100 with a certain length, for example 25m, can be divided into narrow strips, for example, transverse strips 330, with a certain width, for example 25cm. These transverse strips are then analyzed individually. Thus, in the example shown here, one hundred image strips 330, each 25cm wide, are obtained, wherein these image strips have a length corresponding to the widths 310, 320 of the lane 220, particularly the widths between lane markings 130. For example, the lane widths 310, 320 can be calculated based on the position of the lane markings 130 within the image 300. For this purpose, it can be approximated, for example, that the entire road segment recorded in the corresponding image 300 is flat. Therefore, the length of the image strip 330 in the foreground (Vordergrund) can be calculated using pre-adjusted parameters, while the length of the image strip 330 in the background (Hintergrund) is calculated using perspective compression (perspektivische Stauchung), thereby making... Figure 3 The lengths 310 and 320 shown are the same. Another possibility is that the road width in the recorded road segment is constant. If this is the case, the assumption that the road segment is flat can be abandoned, because the distance of the road segment can then be inferred from the length of the image strip 330 using perspective compression. Therefore, the lane width 310 can be calculated for one image strip in the foreground, while the lengths of the other image strips 330 can be used to determine the unevenness in the road segment.

[0046] Texture analysis may include, for example, evaluating the brightness distribution over the length of each individual stripe. Here, as described above with respect to the first scenario, in the case of road 100 drying, it is necessary to distinguish the five lane areas between lane markings 130: three wet areas 120 with darker colors and two dry areas 110 located where tires frequently roll over the lanes, leading to increased evaporation of precipitation in the affected areas, thus making tire tracks 110 brighter in this first scenario. Therefore, the texture frequency for each individual stripe 330 can be calculated based on the number of stripes of varying degrees of darkness and the lengths 310, 320 of the stripes 330. A humidity parameter is then calculated based on the texture frequency. The reliability of moisture identification can be improved by performing a statistical analysis of the evaluated stripes 330 as a whole, for example, by pre-defining a minimum value for a typical texture frequency that may be caused by moisture; above this minimum value, the road is identified as wet or drying.

[0047] As already mentioned, specific textures can differ not only in their frequency but also in their contrast behavior, thus allowing contrast behavior to be included as an additional parameter in the calculation of the humidity parameter. For example, in the first scenario described above, tire track 110 is brighter than the second region 120, while in the second scenario, tire track 110 is darker than the second region 120. This contrast reversal can be used to distinguish puddles in ruts from drying tire tracks. This is particularly advantageous because such puddles can lead to aquaplaning, whereas in the presence of dry strips on the roadway, the slight decrease in the coefficient of friction only needs to be considered when driving over the second region, thus this distinction can potentially mean a considerable added value for driving safety.

[0048] It is also possible that the image data does not originate from a single image 300, but rather, for example, individual images are recorded for each individual strip 330. For this purpose, as... Figure 2 As shown, an image sensor 210, for example in the form of a 2D camera, can be installed at the front of the vehicle 200. Specifically, as... Figure 3As shown, the camera can be pointed forward in the direction of travel to record images 300 of the entire road segment, or it can be pointed downwards to record images of a single strip 330. In the case of a single image, statistical evaluation can be implemented, for example, in the form of an overflow memory, where the most recent image replaces the oldest image. Such an overflow memory may also be advantageous for recording images comprising multiple image strips 330, thereby allowing the statistical evaluation to be based on a larger data base and thus achieving lower susceptibility to interference.

[0049] It can also be stipulated that data be weighted using data age (Alter), thereby giving greater influence to the most recent texture frequencies compared to older data. This allows the calculation of the humidity parameter to be dynamically adapted to current road conditions without immediately discarding relevant data from the recent past. This can be advantageous, for example, in situations of drastic local changes in humidity, such as when driving from an earlier, sun-dry road section into a damp forest. Thus, new conditions are quickly considered when the most recent data receives a higher weight.

[0050] exist Figure 4 In the diagram, the advantageous design of the method is shown in a highly simplified form as a flowchart, and is generally represented by 400.

[0051] In step S1, for example, with Figure 3 The image shown is recorded in the form of image 300.

[0052] In step S2, the image recorded in step S1 is analyzed as described above.

[0053] In step S3, a humidity parameter is calculated based on the analysis results. This humidity parameter is ultimately used to control vehicle functions in step S4. For example, in cases of high humidity, the emergency braking assist system can be adjusted to react more sensitively or earlier, and thus trigger more quickly. Simultaneously, if the driving lane has a lower coefficient of friction due to moisture compared to dry conditions, the maximum applicable braking force can be reduced, for example, to prevent vehicle 200 from skidding. This improves driving safety by effectively shortening the braking distance and maintaining maneuverability during emergency braking.

Claims

1. A method (400) for calculating humidity parameters, wherein an image (300) of a roadway (100, 220) is recorded by means of a camera (210), the image (300) is divided into a plurality of regions (330), and for each of the regions (330) a texture frequency value is determined in a first direction transverse to the roadway (100, 220) and / or in a second direction parallel to the roadway (100, 220), and a humidity parameter is calculated based on the determined texture frequency value, wherein the texture frequency value characterizes the number of texture transformations per unit length, wherein, For each texture frequency value determined in one direction of region (330), a frequency value is calculated, which describes the relative frequency of the texture frequency value along the corresponding other direction in several regions (330), wherein each texture frequency value is considered in a weighted manner using the associated frequency value when calculating the humidity parameter.

2. The method (400) according to claim 1, wherein, When calculating the humidity parameter, a first texture frequency range with a positive correlation coefficient and / or a second texture frequency range with a negative correlation coefficient are considered.

3. The method (400) according to claim 1 or 2, wherein, The texture frequency value is determined using the color value, brightness value, and / or intensity value of the image (300).

4. The method (400) according to claim 1 or 2, wherein, Additionally, when calculating the humidity parameter, the relative color value, relative brightness value and / or relative intensity value and / or contrast value between the image segments within the region (330) are taken into consideration.

5. The method (400) according to claim 1 or 2, wherein, Vehicle functions are controlled based on the calculated humidity parameters.

6. The method (400) according to claim 5, wherein, The vehicle functions include one or more of the following: brake force adaptation, emergency brake assist, centrifugal protection, transmission control, engine control, warning output, and emergency call.

7. A computing device configured to perform all the method steps of the method (400) according to any one of claims 1 to 6.

8. A computer program product comprising instructions that, when executed on a computing device, cause the computing device to perform all the method steps of the method (400) according to any one of claims 1 to 6.

9. A machine-readable storage medium having a computer program product according to claim 8 stored on it.

Images