Method and system for determining at least one parameter for processing a sensor output signal
By optimizing the cost function to determine the parameters of the sensor output signal, the problems of difficulty in selecting parameters and the impact of changes are solved, efficient processing of the sensor output signal and rapid acquisition of parameters are achieved, and suitable for different scenarios and locations.
Patent Information
- Application Number
- CN202410560915.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2023-05-09
- Filing Date
- 2024-05-08
- Publication Date
- 2024-11-12
AI Technical Summary
When processing sensor output signals, the selection of parameters is difficult and time-consuming, and parameter changes may affect the quality of the signal and landmark detection, especially in different scenarios, different processing parameters are required, and parameter changes may interfere with the use of the application.
A method and system are proposed to determine the parameters of the sensor output signal by optimizing the cost function, ensuring that parameter changes do not affect the result quality of the system function, using parameterized functions and machine learning models to process the signals, and storing optimization parameters in the storage device so as to Reused.
Reliable and optimal determination of sensor output signal parameters is achieved, the impact of parameter changes on processing results is reduced, and processing efficiency and usability is improved, especially in different scenarios and locations, the optimization parameters can be quickly obtained.
Smart Images

Figure CN118945499A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method and a system for determining at least one parameter for processing a sensor output signal. Background Art
[0002] Many functions used to process sensor output signals, such as image signals, are parameter-dependent. Therefore, the characteristics of the processing may be affected by certain parameters. For example, when detecting landmarks, i.e., special points or structures in the image signal, such parameter-dependent processing of the sensor output signal is performed. Often, it is difficult and time-consuming to find a set of parameters that provide adequate results for the application during the processing. Another problem is that the choice of such a parameter may also depend on boundary conditions that may change. Certain parameters determined before the change may no longer give adequate results after the change has taken place. In image processing, different scenarios, such as the processing of images taken during the day or at night, indoors or outdoors, may require different processing parameters in order to be able to detect landmarks with the desired quality. In addition to the effort of finding parameters that allow processing the signal with the desired quality, another problem is that changing the parameters during the processing may interfere in an undesirable way with the application that uses the processed signal. If a parameter is changed during image processing and this change affects the detection of landmarks, this change may also affect landmark-based localization in an undesirable way.
[0003] US2022 / 0242422A1 discloses a computer-implemented method for determining optimal operating parameter values of a model-based predictive controller of a vehicle.
[0004] US 10,668,386 B2 discloses a system for simultaneously determining settings for various parameter changes.
[0005] DE 10 2016 205 871 A1 discloses a method for locating a vehicle using a positioning module. Summary of the invention
[0006] The technical problem to be solved by the present invention is to provide a method and a system for determining at least one parameter for processing a sensor output signal, which method and system allow reliable determination of at least one parameter, wherein undesirable effects of the determination (for example through application) on further use of the processed sensor output signal are minimized.
[0007] The technical problem is solved by a method and a system for determining at least one parameter (pres) for processing a sensor output signal. Other advantageous embodiments of the invention are derived from the present application.
[0008] A method for determining at least one parameter for processing a sensor output signal is proposed. The sensor output signal may be an analog signal or a quantitative signal. The analog signal may be processed quantitatively. However, this is not mandatory. The sensor output signal may be a signal about time, i.e. a sequence of values at a plurality of time points or a continuous sequence of values. It is conceivable that the sensor output signal is a one-dimensional signal. It is also conceivable that the sensor output signal is a two-dimensional signal, such as a pixel-based image signal, or a three-dimensional signal, such as a voxel-based image signal. The processed sensor output signal is generated by performing parameter-dependent processing on the sensor output signal. Thus, the input signal for processing is the sensor output signal, wherein the processed output signal is the processed sensor output signal. For example, the processing may be performed by applying a parameterized function, wherein the sensor output signal constitutes the input signal of the function, and the function value is the processed sensor output signal. It is also conceivable that the processing is performed based on a model, wherein the model is a parameterized model. The model may be a model determined by machine learning, such as a neural network, and the parameters of the model may be determined by training the model. Of course, other forms of parameter-dependent processing of the sensor output signal may also be conceivable.
[0009] It is possible but not mandatory that, for the same input signal, the processing result may change when the parameter changes. However, it is also possible that, for the same input signal, different parameter values or combinations of parameter values lead to the same result.
[0010] Preferably, the sensor output signal is an image signal or a radar signal. However, the invention is not limited to these signals. Instead, it is conceivable to apply the proposed method to different sensor output signals.
[0011] In the first system state, the processing of the sensor output signal is used to perform at least one system function. In other words, the result of the processing is used to perform at least one system function. For example, the system function can be performed within the framework of executing a predetermined application. For example, the system function can be a positioning function that determines the system's own position based on the sensor output signal. The system function can also be a navigation function that determines a route to a predetermined target based on the eigenfunction.
[0012] In order to fully perform the system function, multiple sub-functions may be performed, one of which is the processing of the sensor output signal. If the system function is a positioning function, another sub-function may be the detection of landmarks in the processed sensor output signal. If the system function is a navigation function, another sub-function may be a positioning function.
[0013] In another system state in which the system function is not executed or not fully executed, at least one parameter is changed according to the invention and the parameter-dependent output value of the cost function is determined, wherein at least one parameter is determined as a parameter that optimizes the output value of the cost function, in particular minimizes or maximizes the output value of the cost function. The optimization can be achieved by a predetermined number of parameter changes. The optimization can also be performed within a predetermined finite time. This may mean that it is not necessarily necessary to find a global optimum to determine the parameters to be determined according to the invention. Instead, it may be sufficient to determine a local optimum.
[0014] For a system function that is not fully executed, at least one sub-function of the system function is not executed, but other sub-functions can be executed. If the system function is a positioning function, for example, a sub-function for determining the own position cannot be executed, but landmark detection for determining the own position is executed. If the system function is a navigation function, for example, a sub-function for determining the route to the destination cannot be executed, but the positioning function is executed. The system function may also be disabled, but various sub-functions of the system function are executed. Another system state may also be referred to as a shadow state.
[0015] Of course, the method can be used to determine a set of parameters and at least two parameters.
[0016] In general, it is advantageously achieved that the parameter change is carried out in a state in which the processing result is not used for executing at least one system function, so that the parameter change cannot influence the quality of the result of executing the system function. For example, if the system function is a navigation function, the parameter change cannot influence the accuracy of the navigation result, for example, route determination. In other words, a reliable and as optimal as possible determination of at least one parameter for processing the sensor output signal is advantageously produced, without influencing the quality of the result when the processed sensor output signal is further used.
[0017] In another embodiment, the sensor for generating the sensor output signal is an image sensor. Therefore, in this case, the sensor output signal is an image signal. The image sensor can generate a one-dimensional, two-dimensional or three-dimensional image signal. To this end, the image sensor can include, for example, a CCD chip or a CMOS chip. Then, the processing of the output signal can also be referred to as image processing. In this case, the processing can be filtering. The processing can be performed to detect image features, such as edges, corners, spots, ridges or other features. An example of processing is to apply a Canny edge detector to detect edges in an image. However, the processing can also be performed within the framework of other image processing methods, such as segmentation methods.
[0018] Alternatively, the sensor for generating the sensor output signal is a radar sensor. Specifically, the radar sensor may be an imaging radar sensor or a non-imaging radar sensor. Preferably, the radar sensor is a primary radar sensor, and further preferably a pulse radar sensor. The sensor output signal may represent the distance from the radar sensor to the object.
[0019] By using an image sensor to generate the sensor output signal, it is advantageously possible to determine optimal parameters for image processing reliably and without affecting the further processing of the processed image signal. The same applies to the processing of radar signals.
[0020] In another embodiment, at least one intermediate quantity is determined based on the processed sensor output signal. The intermediate quantity can also be determined by the processing. The intermediate quantity can be a quantitative quantity, such as a quantity. In addition, the cost function is selected so that an initial value of the cost function is correlated or anti-correlated with the quantity. If the initial value of the cost function is correlated with the intermediate quantity, the initial value of the cost function increases with the increase of the intermediate quantity or decreases with the decrease of the intermediate quantity, respectively. If the output value of the cost function is anti-correlated with the intermediate quantity, the output value decreases with the increase of the intermediate quantity, or the output value decreases with the increase of the intermediate quantity. This advantageously leads to the determination of the simplest initial value, and thus also to the determination of the optimal parameters. For example, the intermediate quantity can be the number of landmarks detected in the (processed) sensor output signal. This number can be changed for different parameters. For example, by appropriately selecting the parameters, the number of landmarks detected can be maximized.
[0021] Alternatively, the cost function is selected so that the deviation of the quantity from the reference quantity is correlated or anti-correlated with the initial value. Therefore, in this case, if the deviation is minimized, the cost function can be minimized. For example, the output value of the cost function can be maximized when the output value is anti-correlated with the quantity and the deviation is minimized. For example, the output value of the cost function can be minimized when the output value is correlated with the quantity and the deviation is minimized. If the sensor output signal knows in advance the number of landmarks represented / mapped by the sensor output signal (reference quantity), at least one parameter can be determined so that the deviation of the landmark detected according to the parameter corresponds as much as possible to the number of landmarks known in advance. The reference quantity can be predetermined and can be read from a storage device, for example.
[0022] This advantageously enables a simple to implement determination of the at least one parameter.
[0023] In another embodiment, signal processing is performed within the framework of a detection method. In other words, parameter-dependent processing is performed to detect a predetermined feature or characteristic in the sensor output signal. Furthermore, an initial value of the cost function is determined as a function of the number of detected elements. An exemplary embodiment of such a method has been described with respect to the detection of landmarks forming the detected elements. This advantageously leads to a simple and reliable determination of at least one parameter for detecting a feature or characteristic. In particular, no reference quantity, such as the number of elements actually to be detected, needs to be known.
[0024] In another embodiment, an output value of the cost function is inversely related to the number of detected elements, and at least one parameter is determined as a parameter that minimizes the output value of the cost function. This and the corresponding technical advantages have been explained above with reference to the example of landmarks as detected elements.
[0025] Alternatively, an output value of the cost function is related to the spread of the detected element, and at least one parameter is determined as the parameter that minimizes the output value of the cost function. In this case, a characteristic, such as an image position, and the spread of the characteristic can be determined for each detected element. For this purpose, the element can also be detected in multiple output signals.
[0026] It is of course also conceivable that the output value of the cost function is related to the number of detected elements or is inversely related to the diffusion of the detected elements, and that at least one parameter is determined as a parameter that maximizes the cost function value.
[0027] This advantageously leads to a very easy-to-implement determination of at least one parameter for the detection method.
[0028] In an alternative embodiment, the output value of the cost function is related to the false negative rate (Falsch-Negativ-Rate) of the detected elements or is anti-correlated with the true positive rate (Wahr-Positiv-Rate), and at least one parameter is determined as a parameter that minimizes the output value of the cost function. Alternatively, an output value of the cost function value is anti-correlated with the false negative rate of the detected elements, or the output value is related to the true positive rate of the detected elements, and at least one parameter is determined as a parameter that maximizes the cost function value. In this case, for each detected element, characteristics can be determined, such as image position, an interpretation rate depending on the determined characteristics, and a reference characteristic, where the reference characteristic is again known in advance (reference quantity).
[0029] For example, the characteristics of the element to be detected represented / imaged in the sensor output signal may be known in advance, so that the ratio may be determined from this reference information. This advantageously leads to an accurate determination of at least one parameter for the detection method.
[0030] In another embodiment, the detected element is a landmark. In this case, the landmark may be a feature with predetermined properties, which is recognizable in the sensor output signal, in particular due to these properties. For example, such an object may have predetermined physical or optical properties. The object may be a man-made object or a natural object. For example, environmental objects such as columns, switches, windows, road signs, traffic signs, etc. may be used as landmarks for positioning purposes.
[0031] The information about the characteristics of the element to be detected, in particular a landmark, can be present, for example, in the form of map information, wherein the map information can form previously explained reference information or reference quantities, or the reference quantities can be determined based on the map information. If the detected element is a landmark, a simple parameterization is advantageously produced based on the landmark detected in the detection method.
[0032] In another embodiment, the system function includes a positioning function as a sub-function, or the system function is a positioning function. This is to determine the own position, as explained above. In particular, the positioning function can be a landmark-based positioning function. Therefore, the sensor's own position can be determined based on the detected landmarks or their characteristics. If the sensor is installed in a vehicle, the vehicle's own position can be determined. For positioning, it may be necessary to refer to map information, wherein self-positioning can be performed by comparing the detected landmarks with the landmarks detected in the map information. In general, the determination of parameters for positioning can be achieved in an advantageous manner without affecting the currently performed positioning due to the parameter determination.
[0033] In another embodiment, the at least one parameter is stored in a position-referenced and callable manner. In other words, at least one parameter is assigned to a position, in particular a self-position. This information can be stored, for example, in a database. The storage device can be a storage device of the system that performs the parameter determination. As further explained below, the system can also include a sensor. If the system is reset at another time to a position where a parameter has been determined before according to the invention, the parameter can be called up from the storage device according to the current position and the parameter does not need to be determined again.
[0034] Other systems, such as other vehicles, can also call up at least one parameter from the storage device based on their own position and therefore do not need to perform their own parameter determination. These other systems can include the same sensors and / or perform the same system functions. This advantageously leads to improved usability and faster determination of the at least one parameter.
[0035] Alternatively or superimposedly, at least one parameter is stored with reference to a scene and can be called up. For example, a scene can be characterized by the following characteristics: time of day, spatial environment (indoor, outdoor), weather (sunny, cloudy, rainy...). Similar to the embodiment of the location, the system or another system can call up the parameter from the storage device according to the scene without having to re-determine the parameter. This also advantageously leads to improved usability and faster determination of at least one parameter.
[0036] Furthermore, alternatively or superimposedly, at least one parameter can be stored in a callable manner for other systems. This allows other systems to use already determined parameters without having to carry out their own determination. In particular, it is possible to create a database which stores at least one parameter for selected locations and / or scenarios, which optimizes the processing of the sensor output signal for said location and / or scenario as described above. The storage device for this information can be a local, decentralized or server device storage device, i.e. a cloud-based storage device. This also advantageously leads to improved usability and faster determination of at least one parameter.
[0037] In addition, a system for determining at least one parameter for processing a sensor output signal is proposed. The system comprises at least one processing device having an input interface for the sensor signal and an evaluation device for determining a system state. The processing device and the evaluation device may be computing devices. It may comprise a microcontroller or an integrated circuit, or may be designed as such. In addition, the evaluation device or the system may comprise an interface with other systems such as a vehicle communication system. In addition, the processing of the sensor output signal in a first system state is used to perform at least one system function, and in another system state in which the system function is not performed, at least one parameter is changed and a parameter-dependent output value of a cost function is determined, wherein the at least one parameter is determined as a parameter for optimizing the output value of the cost function. The system is configured to perform a method according to one of the embodiments described in the present disclosure, and has the technical advantages that have also been described.
[0038] The system may also include a sensor to generate a sensor output signal.
[0039] Furthermore, the system may comprise an interface for reading reference information, such as map information. Furthermore, the system may comprise an interface for transferring data between the system and a storage device. The system may also comprise a storage device. As described above, the storage device may be used to store at least one parameter, in particular in a position and / or scene referenced manner. The sensor may be a sensor installed in a vehicle. Therefore, a vehicle that may include the system is also described. The vehicle may be a car. The system state may be a state of a system of the vehicle. The function may then be a function performed in the vehicle. BRIEF DESCRIPTION OF THE DRAWINGS
[0040] The present invention is further explained by examples. In the accompanying drawings:
[0041] Figure 1 A schematic flow chart of the method according to the present invention is shown,
[0042] Figure 2 A schematic flow chart showing a method according to another embodiment of the present invention is shown.
[0043] Figure 3 A schematic flow chart showing a method according to another embodiment of the present invention is shown.
[0044] Figure 4 A schematic block diagram of a system according to the present invention is shown. DETAILED DESCRIPTION
[0045] In the following, the same reference numerals refer to elements having the same or similar technical properties.
[0046] Figure 1 A schematic flow chart of the method according to the invention for determining at least one parameter pres and processing a sensor output signal x is shown. In a first step S1, the method according to the invention is started. In this step S1, the current system state SZ is determined. There are at least two system states. In the first system state SZ1, the sensor output signal x is processed and the processed sensor output signal is used for the execution of a system function, such as a positioning function or a navigation function. In another system state different from the first system state SZ1, the system function is not executed or is not completely executed. This system state can also be called a shadow state.
[0047] In a second step S2, it is checked whether the current system state SZ corresponds to the first system state SZ1. If this is the case, the program will terminate. If this is not the case, a parameter change is performed in a third step to minimize the cost function C. At least one parameter p is changed, and the processed sensor output signal is determined based on the unprocessed sensor output signal x and the changed parameter p. Then, based on the processed sensor output signal y, the output value of the cost function C is determined. The at least one parameter p is determined as a parameter that minimizes or maximizes the output value of the cost function C. Such a specific parameter can also be called a result parameter pres. The parameter pres is set in a fourth step S4 so that it can be used in the first system state, in particular before the first system state is activated. In order to determine the starting value of the cost function C, it is necessary to determine an intermediate quantity y, wherein the starting value of the cost function can be determined based on the intermediate quantity. In addition, the cost function C can be selected in such a way that the initial value of the cost function C is correlated or anti-correlated with the intermediate quantity y. It is also necessary to read a reference quantity y_ref, and then determine the output value of the cost function C based on the reference quantity y_ref and the intermediate quantity. In this case, the cost function may be chosen such that its initial value is correlated or inversely correlated with the deviation of the quantity y from the reference quantity y.
[0048] For example, the signal processing can be performed within the framework of a detection method for image elements, in particular landmarks, wherein, for example, the quantity y represents the number of detected landmarks. However, alternatively, a property, such as the position of a landmark, can also be represented by the quantity y. If the target position is known, for example read as reference information, the deviation between the detected position of the landmark and the target position can be determined as an initial value for the cost function.
[0049] It is also possible that the intermediate quantity y represents the false negative rate or the true positive rate of the detected landmarks. In the first case, the cost function can be selected so that its initial value is related to the false negative rate or inversely related to the true positive rate, and then at least one parameter p is determined as a parameter pres that minimizes the initial value of the cost function.
[0050] Figure 2 FIG. 2 shows a schematic flow chart of a method according to the present invention in another embodiment. Figure 1 In contrast to the exemplary embodiment shown, the method between the third and fourth step S3 , S4 comprises an intermediate step S34 , in which the parameters determined according to the invention are stored, in particular in a manner that allows them to be retrieved by other systems.
[0051] Figure 3 FIG. 2 shows a schematic flow chart of a method according to the present invention in another embodiment. Figure 2Unlike the embodiment shown, the method between the first step S1 and the second step S2 comprises a further intermediate step S12, in which a check is made as to whether a result parameter pres has already been determined for the current own position of the system and stored in the storage device 7 (see Figure 4 In a first step S1, the characteristic position can be determined, for example, by performing a positioning function based on the processed sensor output signal. If the parameter pres has already been determined for the characteristic position, this parameter can be retrieved from the storage device 7 and set before activating the first system state SZ1. If the parameter pres for the characteristic position has not yet been determined, this can be performed Figure 1 Further steps S2, ..., S4 of the illustrated embodiment.
[0052] Figure 4 A schematic block diagram of a system 1 according to the invention is shown. The system comprises at least one processing device 2 having an input interface 3 for a (unprocessed) sensor signal x generated by a sensor 4. The sensor 4 may be an image sensor or a radar sensor. The processing device 2 has an output interface 5 for the processed sensor signal. In addition, the system comprises an evaluation device 6 capable of determining a system state SZ. In a first system state SZ1, the evaluation device 6 performs a first system function, such as a positioning function, wherein the processed sensor output signal forms an input quantity for the function. The positioning function can be performed in particular within the framework of indoor positioning of a vehicle. However, if it is detected from the evaluation device 6 that the first system state SZ1 is not activated, the evaluation device 6 can change the parameters for processing the sensor output signal x by the processing device 2 and then determine an intermediate quantity y, such as the number of detected landmarks, from the processed sensor output signal. In addition, from the quantity y determined in this way, the evaluation device 6 can determine an initial value of the cost function C. This has been explained before. The evaluation device 6 can then determine the parameter pres which minimizes or maximizes the output value of the cost function C and then set it before the first system state SZ1 is activated.
[0053] It is further possible that the evaluation device 6 stores the obtained parameters pres in a storage device 7, in particular in a position-referenced manner. For this purpose, the evaluation device 6 can determine the own position of the system, in particular the vehicle. The storage device 7 can be an onboard storage device 7 or an onboard external storage device, in particular a cloud-based storage device, wherein other vehicles or systems can also call up the obtained parameters pres.
[0054] The system can be a positioning system in a vehicle or a part of a vehicle. The system 1 can also include a storage device 7 and / or a sensor 4. The evaluation device 6 can be designed as a calculation device. The evaluation device 6 can also determine reference information that can be read from a storage device (not shown), for example. Then, the initial value of the cost function C can also be determined based on this reference information.
[0055] Reference numerals list
[0056] 1 System
[0057] 2. Processing equipment
[0058] 3 Input Interface
[0059] 4 Sensors
[0060] 5 Output Interface
[0061] 6 Evaluation device
[0062] 7 Storage Devices
[0063] X sensor output signal
[0064] p parameter
[0065] pres parameter
[0066] SZ System Status
[0067] SZ1 first system status
[0068] yref reference
[0069] y Intermediate
[0070] C Cost Function
[0071] S1 First step
[0072] S2 Second step
[0073] S3 Step 3
[0074] S4 Step 4
[0075] S12 Intermediate Step
[0076] S34 intermediate step.
Claims
1. A method for determining at least one parameter (pres) for processing a sensor output signal (x), wherein: In a first system state (SZ1), processing of a sensor output signal (x) is used to execute at least one system function, wherein, in another system state (SZ) in which the system function is not executed, at least one parameter (p) is changed and a parameter-dependent output value of a cost function (C) is determined, wherein the at least one parameter (p) is determined as a parameter (pres) for optimizing the output value of the cost function (C).
2. The method according to claim 1, characterized in that The sensor (4) for generating the sensor output signal (x) is an image sensor or a radar sensor.
3. The method according to any one of the preceding claims, characterized in that At least one intermediate quantity (y) is determined based on the processed sensor output signal, wherein the output value of the cost function (C) is correlated or inversely correlated with the intermediate quantity (y), or the output value of the cost function (C) is correlated or inversely correlated with the deviation of the intermediate quantity (y) from a reference quantity.
4. The method according to any one of the preceding claims, characterized in that Signal processing is performed within the framework of the detection method, wherein an output value of a cost function (C) is determined as a function of the number of detected elements.
5. The method according to claim 4, characterized in that The output value of the cost function (C) is inversely correlated with the number of detected elements or with the dispersion of the detected elements, and at least one parameter (p) is determined as a parameter (pres) that minimizes the output value of the cost function (C); Or the output value of the cost function (C) is inversely correlated with the number of detected elements or correlated with the dispersion of the detected elements, and at least one parameter (p) is determined as a parameter (pres) that maximizes the output value of the cost function (C).
6. The method according to claim 4, characterized in that The output value of the cost function (C) is related to the false negative rate of the detected elements or is inversely related to the true positive rate of the detected elements, and at least one parameter (p) is determined as a parameter (pres) that minimizes the output value of the cost function (C); Or the output value of the cost function (C) is inversely correlated with the false negative rate of the detected elements or correlated with the true positive rate of the detected elements, and at least one parameter (p) is determined as a parameter (pres) that maximizes the output value of the cost function (C).
7. The method according to any one of the preceding claims, characterized in that The detected elements are landmarks.
8. The method according to any one of the preceding claims, characterized in that The system function includes a positioning function as a sub-function or the system function is a positioning function.
9. The method according to any one of the preceding claims, characterized in that The at least one parameter (pres) is stored with respect to a position and / or a scene and / or is stored so as to be accessible to other systems.
10. A system for determining at least one parameter (pres) for processing a sensor output signal (x), wherein: The system (pres) comprises at least one processing device (2) having an input interface (3) for a sensor signal (x) and an evaluation device (6) for determining a system state, wherein in a first system state (SZ1) the processing of the sensor output signal (x) is used to execute at least one system function, wherein in another system state (SZ) in which the system function is not executed, at least one parameter (p) is changed and a parameter-dependent output value of a cost function (C) is determined, wherein at least one parameter (p) is determined as a parameter (pres) which optimizes the output value of the cost function (C).
