Methods for evaluating the functionality of a sensor system

Abstract

Method for evaluating the functionality of a sensor system for environmental sensing, wherein at least two indicators are considered by a data evaluation unit, wherein at least two indicators are statistical indicators that specify environmental objects detected by the sensor system in a given period, a first statistical indicator considers essentially all detected targets, and a second statistical indicator considers only environmental objects with a history, a confidence value for at least one statistical indicator and a confidence value for the calculated functionality of the sensor system are determined by the data evaluation unit, wherein this confidence value depends on at least one confidence value of an indicator.

Description

The invention relates to a method for evaluating the functionality of a sensor system. The method is suitable in particular for motor vehicles which are equipped with a sensor system for monitoring the environment.Vehicles having ACC (Adaptive Cruise Control) systems as a driver assistance function are state of the art. For example, S-class vehicles from Mercedes Benz are equipped with this function. The basis of an ACC system is a sensor system which detects objects in the vehicle environment and determines their distance. An important prerequisite for an ACC function is therefore also a sensor system that functions correctly. The functionality of a sensor system can be severely impaired, for example, by dirt, so that relevant objects in the vehicle environment are no longer detected. A reliable detection of a contamination or blockage of the sensor system and an evaluation of the functionality of the sensor system are therefore of great importance.DE 103 44 617 A1 discloses a method for evaluating the functionality of a sensor system for detecting the environment, wherein two indicators are viewed by a data evaluation unit. The indicators are the time profile of the object data of an object class (or their back-scattering cross sections, RCS) and the frequency or probability of occurrence of the object classes. For the object data of each object class, the statistical deviation of the back-scattering cross sections is taken into account, wherein a temporal sequence of object data of each object class is formed. On the basis of a statistically significant temporal reduction of the radar back scatter cross sections (RCS) of the objects of the relevant object classes, an error property of the sensor system can then be established. The deviations of the back scattering cross sections and the frequency or probability of occurrence are included in the calculation of an error signal F.Furthermore, EP 1 131 651 B1 discloses a method for state detection in a system for automatic longitudinal and / or lateral control in a motor vehicle, in which objects are detected in a virtual track. Detection gaps can arise, wherein the object stability can be determined from the quotient of the number of time steps in which an object was detected and the total number of time steps in which the object was viewed.It is an object of the present invention to specify an improved method for evaluating the functionality of a sensor system.This object is achieved according to the invention by a method according to the independent claims. Advantageous further developments can be found in the dependent claims.A method for evaluating the functionality of a sensor system for detecting the environment in a vehicle is specified. The sensor system is designed, for example, as a LIDAR radar sensor. At least the strength of the measurement signal and the distance of the associated object are determined from the sensor data. In a further exemplary embodiment, the sensor system is designed as a camera sensor. At least two indicators are considered for evaluating the functionality. Of these, an indicator is a statistical indicator that indicates surrounding objects detected in a predetermined period of time. A confidence value is determined for the statistical indicator. In a preferred exemplary embodiment, a confidence value can also be determined for non-statistical indicators.The confidence value estimates the credibility of the result on the basis of predetermined plausibility considerations.In addition, a confidence value for the functionality of the sensor system (calculated from the indicator / s) is determined, wherein this confidence value depends on at least one confidence value of an indicator. In a preferred embodiment, a plurality or all available confidence values of the indicators are used for determining the confidence value for the functionality of the sensor system.In a preferred embodiment of the invention, the confidence value of a statistical indicator depends on the number of events recorded and / or the distribution of the events. In a further embodiment of the invention, the determination of the confidence value is characteristic for the respective indicator.A statistical indicator relates, for example, to the sensor range. In this case, the measurement range for at least one detected object is extrapolated from the strength of a measurement signal and the distance from the object. Preferably, the following equation is used for this, wherein D_expol is the extrapolated distance, D_a is the distance to the object, a is the amplitude of the measurement signal of the object and a0 is the detection threshold. The extrapolated distances D_expolare recorded in statistics moving over time. The extrapolated sensor ranges enter the statistical indicator, for example, with their mean value and / or maximum and / or minimum and / or distribution width, etc.In a preferred embodiment of the invention, a first predetermined statistical indicator takes into account all targets that are detected. Preferably, a second statistical indicator takes into account only environmental objects to which a history can be assigned. Environmental objects have a history when they are acquired multiple times and tracked (tracked) over a period of time t.In particular, for at least one indicator, a weight is provided with which this indicator is intended to contribute to the overall result. The weight is preset in the simplest case and remains constant. It can also be provided that each indicator contributes to the end result with the same weight. In a further embodiment, it is provided that at least one weight is variable. In an advantageous embodiment of the invention, surrounding objects having a predefined weight are also taken into account in an indicator. Preferably, surrounding objects representing other road users are taken into account for evaluating the functionality with an increased weight.In a further embodiment, the weight of the individual indicators also contributes to the determination of the confidence value for the evaluation of the functionality. For example, the weighted confidence values of the indicators are summed. For better comparative results, the result is given in normalized form.In an advantageous embodiment of the invention, deposition of foreign materials on a surface is detected. The surface is generally the entry surface through which the radiation to be detected enters the sensor system and, if appropriate, also exits in the case of an active sensor. Dust particles, dirt, water droplets, ice, snow, etc. settled on the surface due to environmental influences. The surface can be examined with an additional sensor, e.g. a camera system focused on the surface or by reflection measurement. The functionality is assessed as a function of the presence of deposited foreign materials and at least one statistical indicator.In an advantageous embodiment, the predetermined period of time depends on the result of the examination for a deposition of foreign materials. In particular, when a foreign material is detected on the sensor exit surface, the predefined time period is selected to be relatively short in order to be able to react quickly to a functional impairment of the sensor system. If no deposition of a foreign material is detected, the predetermined observation period is correspondingly longer.In a further advantageous embodiment of the invention, the weight with which at least one statistical indicator enters into the result of the contamination detection is dependent on the examination result which indicates a deposition of foreign materials. For example, when a foreign material is detected on the sensor exit surface, a weight of the statistical indicator is increased. With this procedure, an incorrect alarm can be prevented if a deposit is detected but no visibility restriction of the sensor system is detectable. In a further embodiment of the invention, the weight of the stat depends. Indicators depend not only on the current detection of a foreign material, but also on the associated confidence value. That is, only when deposition is likely to occur is the statistical indicator provided with a large weight.In an advantageous embodiment of the invention, a statistical indicator in particular records the frequency of occurrence of objects per unit of time. It is assumed that objects are almost always located in the vehicle environment, which is detected by the sensor system. If no objects or only very few objects are detected per unit time, this is an indication that the sensor functionality is limited.In a further embodiment of the invention, at least one surface through which sensor radiation passes is examined for deposition of foreign materials by means of the analysis of reflected sensor radiation. The surface is generally the entry surface through which the radiation to be detected enters the sensor system and also exits in the case of an active sensor. For example, in an active sensor system, the transmission radiation is reflected to a certain extent at the exit surface. If a deposit of a foreign material is present on this surface, a higher percentage of the transmitted radiation is reflected and, in the case of increased reflection, contamination of the exit surface is detected. However, in this type of measurement it is unclear which material has deposited on the surface and how it has an effect on the sensor functionality, i.e. whether the sensor radiation is blocked by a highly absorbing material or the material can be passed almost unimpeded by the sensor radiation.A motor vehicle with a sensor system and a data evaluation unit is claimed, on which a method according to one of Claims 1-7 is stored.The invention is explained in more detail below with reference to exemplary embodiments and two figures. FIG. 1 : Method for evaluating the functionality of a sensor system FIG. 2 : conversion of a measured value into a probability value FIG. 3 shows a method for determining the contamination probability of a sensorAll features described herein can contribute to the invention individually or in any combination. A chronological sequence of the method steps is not necessarily predefined by the sequence selected here.In a first exemplary embodiment of a method for evaluating the functionality of a sensor system for detecting the environment, at least two of the indicators described below are taken into account.A statistical indicator indicates the number of detected objects per unit time. It is assumed that detectable targets are always located in the environment of the vehicle.Another statistical indicator indicates extrapolated visibility and takes into account environmental targets for which no history is available. Stationary and moving targets are included in statistics. The maximum distance at which a target can still be detected is extrapolated knowing the sensor performance (S / N), the signal amplitude and the distance of the target. An advantage of this indicator is that the maximum represents the maximum achievable range of the sensor.A further statistical indicator specifies the extrapolated visibility range and takes into account surroundings objects with history for this purpose. Environmental objects are formed from targets that have been detected and tracked (tracked) more than once. In a situation analysis, the objects are evaluated for their relevance for the respective function, these include, inter alia, the attributes: stationary / moving, speed, acceleration, position, quality, service life, coverage by other objects, lane assignment. These criteria enter into the indicator a weight for the respective object. This procedure is based on the consideration that vehicles moving, for example, when wet are partially covered by a spray cloud and thus the detection range of the sensor for stationary targets (for example traffic signs, bridges, tunnels, guardrails, edge buildings) is dependent on factors above the targets moved and covered by a spray cloud. In statistics, this results in spreading. That is to say that the maximum of the extrapolated visibility range can be, for example, 150 m, while the minimum is 60 m and the mean value is 80 m. Stationary objects are accordingly weighted differently than moving objects which are relevant for driver assistance function. The detection range of all objects is introduced into the object statistics taking into account the weighting.A further statistical indicator specifies the extrapolated visibility range and takes into account the first and last detection of objects. Environmental objects are formed from targets that have been detected and tracked (tracked) more than once. In a situation analysis, the objects are evaluated for their relevance for the respective function, including, among other things, the attributes: first or last detection, standing / moving, speed, acceleration, position, quality, service life, coverage by other objects, lane assignment. These criteria enter into the indicator a weight for the respective object. The first and last detections of the objects are introduced into the object statistics taking into account the respective weights This indicator is only available if first or last detections of objects are available.In an advantageous exemplary embodiment, the sensor system for detecting the environment is designed as a camera system. In darkness, the lights of vehicles driving ahead and on the other hand and reflective objects are primarily visible. Objects reflecting the headlight light are, for example, traffic signs or roadway boundaries. The light spots representing objects or vehicles are tracked (tracked) in a sequence of images.A preferred statistical indicator for assessing the functionality of the camera sensor system is the fade time. The time between the first detection of the object and the classification of the object, e.g. as an oncoming or preceding vehicle, is referred to as a blending time. The blending times are statistically evaluated over an observation time T_B, and if the maximum blending time is very large, i.e. greater than an (adaptive) threshold value, this is an indication of a restriction of the functionality of the sensor system. The cause of the restriction can be, for example, fogging, icing, fog, rain, snow or the like.Another statistical indicator of the functionality of a camera sensor system is a blurred image of light sources. In the image, a collection of many closely adjacent local intensity maxima is detected, which are assigned to an object or a light source and are optionally tracked (tracked) in an image sequence. Objects with a collection of light spots are classified into statistics with a positive weight, the objects with a normal light distribution with a negative weight. If object lights with a collection of many closely adjacent local intensity maxima are predominantly present, this indicates condensation or icing.Another indicator of the functionality of a camera sensor system is the distribution of lights in the image. For this purpose, the image is divided into three regions. The area A1 is arranged in the image so as to cover the area directly in front of the motor vehicle illuminated by the vehicle headlights. Depending on the installation position of a camera system in the motor vehicle looking forwardly and obliquely downwardly, this can be, for example, the lower half of the image. The image section A 2 lies above A 1 in the image and objects are primarily imaged in a middle distance range here. The region A3 is preferably located in the upper quarter of the image. An intensity I1, I2and I3is determined for the regions A1, A2and A3, respectively. Due to the headlight of the vehicle, the area A1 should be illuminated best in darkness and the intensity I1 should be increased accordingly compared to I2 (I3). In the case of contamination or condensation, the difference between I1 and I2 (I3) is smaller or not present. The boundary between I2 and I1 is adjusted cyclically in order to follow a shift of the light / dark boundary, for example, when the headlight illumination range changes. The variance of the intensities I 1, I 2, I 3 is a measure of the confidence value, among other things. A low variance intensity is assigned a high confidence value.In a further exemplary embodiment, the observation duration for a statistical indicator is controlled by the measured variable which indicates a deposition of foreign materials. The sensor system is configured as a beam sensor (e.g., radar or LIDAR). The transmitted radiation is reflected at the exit window of the sensor system, e.g. transparent pane, beam-shaping optics, etc., and detected with a receiving element. The proportion of the reflected light depends on the material of the exit window and the sensor geometry (angle of incidence of the radiation on the surface etc.) and is known for a predefined sensor system. If a foreign material is present on the exit surface, the proportion of the reflected radiation increases measurably. The radiation is identified on the basis of the characteristic distance. This is predetermined by the geometric dimensions of the sensor unit. A distance measurement is carried out, for example, according to the pulse transit time method.When deposition is detected, the observation time for a statistical indicator is reduced in order to achieve a rapid reaction to the contamination if appropriate. If no deposition of a foreign material is detected, the predetermined observation period is correspondingly longer.FIG. 1 shows a schematic sequence of a method for evaluating the functionality of a sensor system. Three indicators M 1, M 2, M 3 are determined. In this exemplary embodiment, the indicators M 1, M 2, M 3 represent the deposition of foreign materials, the extrapolated visibility range and the number of detected objects per unit time. For each indicator value Mi, a probability Pi is calculated, which takes up a value between 0 and 1. The probability Pi indicates how high the probability is that the function of the sensor system is impaired. For each value of an indicator Mi, a confidence value Ci is calculated. For example, the confidence value C1 to the current value of the indicator M1, which is representative of the deposition of foreign materials, is calculated as a function of the previous values for M1, a strong deviation from the previous value or from an average value of previous values is unusual for a deposition process. Rather, it is likely that the exit surface will always contaminate a little more over time and the value for M1 will increase continuously (and not abruptly). In a multi-beam sensor arrangement with a plurality of transmitting elements, for example, the confidence value C1 is calculated as a function of the result for all transmitting beams (e.g. mean value); here, the result for all individual beams should be the same under comparable conditions. The confidence value C2 of the current value of the indicator M2, which is representative of the extrapolated visibility range of the sensor system, is determined as a function of the number of objects that were observed for this statistical indicator in a predetermined period of time. If relatively many objects were detected in the predefined time period and used for extrapolation of the visibility range of the sensor system, it is highly likely a reliable value, so that the confidence value C 2 is also large. If only a few detected objects contribute to determining the maximum visibility range of the sensor system, the probability that an extrapolated visibility range corresponds to the actual visibility range is lower, and only poorly reflecting objects, e.g. heavily contaminated objects, have actually been detected. In this case, the confidence value C2 is also lower.The confidence value C 3 with respect to the current value of the indicator M 3, which indicates the number of objects per unit time, depends on the difference between the current value and the mean value of previously recorded measurements. It is assumed that a vehicle environment changes continuously and not abruptly. For example, when traveling on a fixed and marked roadway, the marking poles, guardrails, etc. are detected. In contrast, when traveling in the coast, an environmental object is hardly detected. In an advantageous embodiment of the invention, the number of detected objects is filtered using a low-pass filter before averaging in order to minimize the effect of outliers, which are frequently based on false detection.For each indicator M 1, M 2, M 3, a weight W 1, W 2, W 3 is also specified, with which this indicator is intended to contribute to the overall result. These weights are preset in the simplest case and remain constant. It can likewise be provided that at least one weight is variable. In an advantageous embodiment, the weight W2 for the indicator M2, which indicates the extrapolated visibility of the sensor system, depends on the value for the indicator M1, which represents a deposition of foreign materials. If deposits are detected, the value M2 for the extrapolated visibility range is incorporated into the overall result with a particularly high weight. With this procedure, an incorrect alarm can be prevented if a deposit is detected but no visibility restriction of the sensor system is detectable. In a further embodiment of the invention, the weight W2 depends not only on the current value for M1 but also on the associated confidence value C1, i.e. only if a deposit is very likely to be present is the indicator M2 provided with a large weight. A value range is provided for a confidence value Ci, which lies, for example, between 0 and 0.95. From the probability values P1, P2, P3, a total probability P is calculated, which is a measure of the functional impairment of the sensor system. In a further exemplary embodiment, the total probability P is additionally dependent on the respective weight of an indicator W 1, W 2, W 3. For example, the normalized products of probability value Pi and associated weight Wi are summed. FIG. 1 shows a further possibility of calculating the probability of a functional impairment of the sensor system P from the probabilities P 1, P 2, P 3, the associated weights W 1, W 2, W 3 and confidence values C 1, C 2, C 3. In addition, a confidence value C is calculated from the individual confidence values C 1, C 2, C 3 and the associated weights W 1, W 2, W 3 for the overall probability P.In one exemplary embodiment of the invention, a threshold value is provided for the confidence value C. The threshold value for a confidence value C, for which a value range between 0 and 1 is provided, assumes a value between 0.60 and 0.85, for example. Only when the threshold value is exceeded, i.e. the calculated total probability P indicates a correct value, is the total probability P considered. If in this case the total probability P again exceeds a predetermined threshold value, i.e. the function of the sensor system is restricted with a high probability, a measure is taken, e.g. the driver is warned or the sensor system switches off or a cleaning device is activated, etc.In FIG. 2, the probability Pi is plotted against the value of the indicator Mi. The highest value for the probability is P1, the lowest value P0. In an advantageous embodiment of the invention, P1 is chosen to be 1 and P0 is chosen to be 0. A probability of 1 indicates that the function of the sensor is highly likely to be impaired. If, for example, the maximum value for the extrapolated visibility within a predetermined time period is 30 m or 150 m, the probability of P1 or P0 is assigned to this. In an alternative embodiment of the invention, a probability value Pi(mean, min, max) for the minimum, mean value and maximum with respectively associated confidence value Ci(mean, min, max) is considered for each indicator Mi. In addition, a weight Wi(mean, max, min) for each value is included in the calculation of the contamination probability. From these values, as shown in FIG. 3, an average probability P of minimum, maximum and average is calculated. The average probability P includes the probability values P of the individual indicators, which probability values have a weight W and a confidence value C. For the average probability values P for the minimum, maximum and mean value, a standard deviation σ mean, σmin, σmaxis calculated. A minimum probability Pmin that the sensor system is contaminated by subtracting the standard deviation σmin from the average probability P'. Accordingly, as shown in FIG. 3, Pmax is calculated. The confidence values Cmean, Cmin, Cmax for the minimum, maximum and mean probability of a sensor blockage are determined with the following relation

Claims

Method for evaluating the functionality of a sensor system for detecting the environment, wherein at least two indicators are viewed by a data evaluation unit, wherein at least two indicators are statistical indicators which indicate surrounding objects detected by the sensor system in a predefined period of time, a first statistical indicator takes into account substantially all detected targets, and a second statistical indicator takes into account only surrounding objects having a history, a confidence value for at least one statistical indicator and a confidence value for the calculated functionality of the sensor system is determined by the data evaluation unit, wherein this confidence value depends on at least one confidence value of an indicator.Method according to claim 1, characterised in that the confidence value of a statistical indicator depends on the number of events recorded and / or the distribution of the events.Method according to one of the preceding claims, characterized in that predefined surrounding objects and / or indicators with a weight are included in the assessmentMethod for evaluating the functionality of a sensor system for detecting the environment according to one of the preceding claims, characterized in that at least one area through which sensor radiation passes is examined for a deposition of foreign materials on the basis of the sensor data, and at least one indicator which is used for evaluating the functionality specifies a measure for the impairment of the sensor system by deposited foreign materials.Method according to claim 4, characterised in that the predetermined period of time during which a statistical indicator is determined depends on the result of the examination for a deposition of foreign materials.Method according to one of the preceding claims, characterized in that a weight with which at least one statistical indicator enters into the result of the contamination detection depends on the examination result which indicates a deposition of foreign materials.Method according to one of the preceding claims, characterized in that at least one surface through which sensor radiation passes is examined for deposition of foreign materials by means of the analysis of reflected sensor radiation.Vehicle with a sensor system and a data evaluation unit on which a method according to one of the preceding claims is stored.

Images