Misstep judgment method and device, electronic equipment and storage medium

Abstract

The application relates to the technical field of Internet of Vehicles, and provides a misstep judgment method and device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring current driving data, the current driving data comprising at least one of vehicle position road information, vehicle driving information, environmental meteorological information and driving related information; and judging the accelerator pedal operation of a driver based on the current driving data and a misstep judgment model corresponding to the driver, the misstep judgment model being obtained by training historical driving data of the driver. The misstep judgment method, device, electronic equipment and storage medium provided by the application can reflect the individual driving habits of the driver, are more personalized and targeted, can effectively make correct judgments according to different driving habits, and can more accurately predict whether the accelerator pedal is misstepped, thereby improving the accuracy of misstep judgment.

Description

Technical Field

[0001] This invention relates to the field of vehicle networking technology, and in particular to a method, device, electronic device, and storage medium for detecting accidental stepping. Background Technology

[0002] Cars have become an indispensable means of transportation in people's daily lives, and their level of intelligence is constantly improving. Driving safety is now a matter of life and property safety. In recent years, there have been frequent reports of drivers mistakenly pressing the accelerator instead of the brake in certain situations, causing serious accidents.

[0003] Current solutions for handling accidental accelerator pedal presses mostly rely on passive system suppression. This involves collecting driver data and comparing it to pre-defined threshold values ​​to determine if an accidental press has occurred, thus inhibiting acceleration. However, different drivers have different driving habits, and the commonly used threshold values ​​are relatively fixed, leading to inconsistent accuracy in detecting accidental presses. Summary of the Invention

[0004] This invention provides a method, device, electronic device, and storage medium for determining accidental accelerator pedal use, in order to address the shortcomings of the prior art in terms of the accuracy of determining whether a driver has accidentally pressed the accelerator pedal.

[0005] This invention provides a method for detecting accidental stepping, comprising:

[0006] Acquire current driving data, which includes at least one of vehicle location road information, vehicle driving information, environmental weather information, and driving-related information;

[0007] Based on the current driving data and the driver's mis-press judgment model, the driver's accelerator pedal operation is judged. The mis-press judgment model is trained based on the driver's historical driving data.

[0008] According to the accidental step detection method provided by the present invention, the step of obtaining the accidental step detection model includes:

[0009] Feature extraction is performed on the driver's historical driving data for each information dimension to obtain the features of each dimension, and the baseline features of the driver are determined based on the features of each dimension.

[0010] Based on the differences between the features of each dimension and the baseline features, the driver's accelerator pedal operation is predicted, and the accidental pedal judgment model is trained based on the prediction results.

[0011] According to the method for determining accidental acceleration provided by the present invention, the step of predicting the driver's accelerator pedal operation based on the difference between the features of each dimension and the baseline features includes:

[0012] Based on the information gain weights of the features in each dimension and the differences between the features in each dimension and the baseline features, the driver's accelerator pedal operation is predicted.

[0013] The information gain weight of any dimension feature is determined based on the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the feature is included, and the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the feature is not included.

[0014] According to the method for determining accidental acceleration provided by the present invention, the step of predicting the driver's accelerator pedal operation based on the difference between the features of each dimension and the baseline features includes:

[0015] The features from each dimension are fused to obtain the fused features;

[0016] Based on the differences between the features of each dimension and the baseline features, as well as the differences between the fused features and the baseline features, the driver's accelerator pedal operation is predicted.

[0017] According to the accidental step detection method provided by the present invention, determining the driver's baseline characteristics based on the various dimensional features includes:

[0018] The baseline feature is determined based on the average value of each dimension of features corresponding to the driver correctly pressing the accelerator.

[0019] According to the method for determining accidental pedal press provided by the present invention, after determining the driver's accelerator pedal operation, the method further includes:

[0020] If the judgment result indicates that the accelerator pedal operation was a mistake, a reminder message will be sent.

[0021] According to the accidental step detection method provided by the present invention, the method further includes:

[0022] The driver's identity is verified;

[0023] If the identity recognition result indicates that the driver is a new user, then a misstep judgment model corresponding to the driver is constructed based on the current driving data;

[0024] If the identity recognition result indicates that the driver is not a new user, then the wrong-stepping judgment model corresponding to the driver is loaded, and the wrong-stepping judgment model is updated based on the current driving data.

[0025] The present invention also provides a device for detecting accidental stepping, comprising:

[0026] The data acquisition unit is used to acquire current driving data, which includes at least one of vehicle location road information, vehicle driving information, and environmental weather information, as well as driving and riding related information.

[0027] The accidental pedal judgment unit is used to judge the driver's accelerator pedal operation based on the current driving data and the accidental pedal judgment model corresponding to the driver. The accidental pedal judgment model is trained based on the driver's historical driving data.

[0028] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement any of the above-described methods for determining accidental stepping.

[0029] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the accidental step detection method as described above.

[0030] The present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements any of the above-described methods for determining accidental stepping.

[0031] The method, device, electronic device, and storage medium for mis-pedal judgment provided by this invention train a mis-pedal judgment model corresponding to the driver by collecting the driver's historical driving data. The mis-pedal judgment model can reflect the driver's personal driving habits. Compared with the use of general configuration parameter thresholds in related technologies, the mis-pedal judgment model in this embodiment is more personalized and targeted, and can effectively make correct judgments for different driving habits. Therefore, it can more accurately predict whether the accelerator pedal has been accidentally pressed, and improve the accuracy of mis-pedal judgment. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0033] Figure 1 This is one of the flowcharts of the accidental step detection method provided by the present invention.

[0034] Figure 2 This is one of the flowcharts illustrating the method for obtaining the accidental step judgment model provided by the present invention.

[0035] Figure 3This is the second flowchart of the method for obtaining the accidental step judgment model provided by the present invention.

[0036] Figure 4 This is a schematic diagram of the prediction module provided by the present invention.

[0037] Figure 5 This is the second flowchart of the accidental step detection method provided by the present invention.

[0038] Figure 6 This is a schematic diagram of the accidental step detection device provided by the present invention.

[0039] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0041] In recent years, there have been frequent reports of drivers mistakenly pressing the accelerator instead of the brake in certain situations, causing serious safety accidents. Therefore, it is of great significance to use monitoring equipment and big data models to determine whether a driver has mistakenly pressed the accelerator, and to promptly alert the driver when such an error occurs, thereby reducing the probability of accidents.

[0042] Current solutions for handling accidental accelerator pedal presses mostly rely on passive system suppression. This involves collecting driver parameters and comparing them with pre-defined threshold values ​​to determine if an accidental press has occurred, thus inhibiting acceleration. However, existing technologies lack personalized driver data. Since different drivers have different driving habits, and universally applicable threshold values ​​are fixed, they cannot effectively make accurate judgments based on varying driving habits, resulting in poor accuracy in detecting accidental accelerator pedal presses.

[0043] To address the aforementioned problems, this invention proposes a method for detecting accidental accelerator pedal presses. In this method, current driving data is acquired during driving. This data includes at least one of the following: vehicle location and road information, vehicle driving information, and environmental weather information, as well as driving-related information. Based on the current driving data and a corresponding accidental accelerator pedal press detection model, the method judges whether the driver has pressed the accelerator pedal. The accidental accelerator pedal press detection model is trained based on the driver's historical driving data and reflects the driver's individual driving habits. Compared to related technologies that use universal configuration parameter thresholds, the accidental accelerator pedal press detection model in this embodiment is more personalized and targeted, effectively making correct judgments based on different driving habits. This allows for more accurate prediction of whether the accelerator pedal has been pressed accidentally, improving the accuracy of accidental accelerator pedal press detection.

[0044] This invention can be applied to scenarios where it is necessary to determine whether a driver has mistakenly pressed the accelerator instead of the brake during car driving. The executing entity of this method can be an electronic device such as a terminal device, computer, server, server cluster, or a specially designed accident detection device, or it can be an accident detection device installed in such an electronic device, which can be implemented through software, hardware, or a combination of both.

[0045] Figure 1 This is one of the flowcharts illustrating the accidental step detection method provided by the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0046] Step 110: Obtain current driving data, which includes at least one of the following: vehicle location road information, vehicle driving information, environmental weather information, and driving-related information.

[0047] Specifically, current driving data can be collected when the driver is detected to have pressed the accelerator or brake pedal heavily. Vehicle location road information refers to the vehicle's current location and the road information at that location, describing the spatial factors influencing whether the driver accidentally pressed the accelerator. The vehicle's location is obtained in real time using built-in or aftermarket positioning devices, such as GPS locators; the vehicle's location can be represented by latitude and longitude, for example.

[0048] Road information at the vehicle's location represents the vehicle's driving environment, encompassing various road-related data. Road information may include the current road type, road conditions, traffic rules, etc. The current road type may include, for example, highways, expressways, arterial roads, provincial roads, county roads, rural roads, and mountain roads; road conditions refer to the current traffic situation, such as whether there is congestion on the current road segment through the navigation map information in the vehicle's infotainment system, or whether there are obstacles within the driving range through vehicle image recognition; traffic rules refer to the current road traffic regulations, such as whether there are speed limits and speed limit thresholds, which can be obtained through navigation map information.

[0049] Vehicle driving information describes the dynamic performance or state of a vehicle during operation, and from a vehicle perspective, it describes the factors that could cause a driver to accidentally press the accelerator. Vehicle driving information may include current speed, average speed, speed range for the current time period, braking system status, and vehicle acceleration. Vehicle driving information can be read through the vehicle's OBD-II (On-Board Diagnostics) port.

[0050] Environmental meteorological information describes factors within the external environment that influence whether a driver accidentally presses the accelerator. Environmental meteorological information refers to the current state of the natural or atmospheric environment, specifically including current weather, temperature, season, time, visibility, etc.

[0051] Driver-related information describes factors that influence whether a driver accidentally presses the accelerator, from a personalized perspective. This information includes basic driver information and passenger feedback. Basic driver information may include driver identification number, gender, driving experience, and age, and can be obtained by pre-storing driver information and matching it with the driver's identity. Passenger feedback may include whether passengers provide clear voice prompts (e.g., "brake," "stop," "scream"), which can be collected by in-vehicle audio recording devices.

[0052] In this embodiment, the data used to determine the possibility of the driver accidentally stepping on the pedal includes at least one of the following: vehicle location and road information, vehicle driving information, environmental weather information, and driving-related information.

[0053] Step 120: Based on the current driving data and the driver's mis-press judgment model, judge the driver's accelerator pedal operation. The mis-press judgment model is trained based on the driver's historical driving data.

[0054] Specifically, the current driving data is input into the accidental acceleration detection model, which processes the current driving data and predicts whether the driver has accidentally pressed the accelerator based on the data processing results. Accidental acceleration refers to the driver pressing the accelerator instead of the brake according to the model's prediction.

[0055] Before executing step 120, a large amount of historical driving data of the driver can be collected, and this historical driving data can be trained using machine learning algorithms. Historical driving data can initially construct driving habit models for different drivers. For example, if user 001 is 35 years old, has 9 years of driving experience, and possesses mature driving skills, in good weather, high visibility, and pleasant temperatures at midday on an unobstructed provincial highway with a speed limit of 60 km / h, the user's habitual driving speed range is 45-55 km / h, and the user's habitual acceleration range is ±3 km / s. If the deviation between the collected feature data and the user's accidental acceleration judgment model data exceeds a threshold, it is determined that the user may have accidentally pressed the accelerator. The more feature data collected and the wider the coverage of user scenarios, the more accurately the model can predict whether the user's accelerator pedal operation is correct after training.

[0056] During training, the accidental pedal misjudgment model learns from historical driving data the driver's individual driving habits, specifically when to apply the brakes and when to apply the accelerator. The resulting accidental pedal misjudgment model reflects the driver's individual driving habits, making it more personalized and targeted.

[0057] The accidental acceleration detection model can reflect the baseline driving data when the driver correctly presses the accelerator. The greater the difference between the current driving data and the baseline driving data, the greater the probability that the driver's acceleration operation is accidental; conversely, the smaller the difference between the current driving data and the baseline driving data, the smaller the probability that the driver's acceleration operation is accidental.

[0058] The method provided in this embodiment of the invention trains a mis-press judgment model corresponding to the driver by collecting the driver's historical driving data. The mis-press judgment model can reflect the driver's personal driving habits. Compared with the use of general configuration parameter thresholds in related technologies, the mis-press judgment model in this embodiment is more personalized and targeted, and can effectively make correct judgments for different driving habits. As a result, it can more accurately predict whether the accelerator pedal has been accidentally pressed, thus improving the accuracy of mis-press judgment.

[0059] Based on any of the above embodiments Figure 2 This is one of the flowcharts illustrating the method for obtaining the accidental step detection model provided by this invention, such as... Figure 2 As shown, the steps for obtaining the accidental step detection model include:

[0060] Step 210: Extract features from the driver's historical driving data for each information dimension to obtain the features of each dimension, and determine the driver's baseline features based on the features of each dimension.

[0061] Step 220: Based on the differences between the features of each dimension and the baseline features, predict whether the driver's accelerator pedal operation is correct, and train the error judgment model based on the prediction results.

[0062] Specifically, after collecting the driver's historical driving data, the historical driving data can be cleaned, such as removing duplicates, filling in missing values, correcting obvious errors, and performing preliminary verification on the completeness and accuracy of the data to ensure that the data meets the expected format and range.

[0063] Subsequently, features were extracted from the driver's historical driving data across various information dimensions. These dimensions can include four aspects: vehicle location and road information, vehicle driving information, environmental and meteorological information, and driving-related information. The extracted features for each dimension can include vehicle location and road features, vehicle driving features, environmental and meteorological features, and driving-related features.

[0064] In some embodiments, features of each dimension can be represented by numerical encoding. For features related to driving and riding, for example, driver facial data is obtained through driver authorization and assigned different numbers to different users to distinguish them, such as 001; user gender is identified by 0 (female) and 1 (male); user driving experience and age are identified by numbers, such as 9 and 35; whether passengers have obvious voice keywords (such as "brake", "stop", "scream") prompts is collected, with 0 indicating no and 1 indicating yes. In summary, the feature vector of driving and riding related features can be represented as [001 (user number), 1 (male), 9 (driving experience), 35 (age), 0 (no voice prompts from passengers), 0 (correctly pressing the accelerator)].

[0065] For the characteristics of the vehicle's location in the road dimension, such as the car's driving position, it is represented by latitude and longitude information as (114.33, 30.35). The nature of the current road is determined, with 0 representing a highway, 1 representing a provincial road, 2 representing a provincial road, 3 representing a county road, 4 representing a rural road, and 5 representing other roads. The presence of obstacles within the driving range is identified through vehicle image recognition, with 1 indicating the presence of obvious obstacles and 0 indicating the absence of obstacles. The traffic regulations of the current road segment are obtained through the navigation map information in the vehicle's infotainment system, such as whether there is a speed limit, such as 60km / h. In summary, the vehicle's location road feature vector is represented as [001 (user ID), 114.3321 (latitude), 30.35216 (longitude), 1 (provincial road), 0 (no obstacles), 1 (speed limit), 60km / h (speed limit threshold), 0 (correct accelerator pedal pressed)].

[0066] Based on the characteristics of vehicle driving, the current vehicle speed (e.g., 55 km / h), average vehicle speed (e.g., 60 km / h), and the current time range of vehicle speed (e.g., left range 50 km / h, right range 70 km / h, indicating that the driver's driving speed range is 50~70 km / h during this period) are read through the vehicle's OBD-II (On-Board Diagnostics) port. Braking system status information is also provided, with 0 indicating braking and 1 indicating accelerator (e.g., 1). Acceleration information is obtained through vehicle acceleration sensor measurements, with positive numbers indicating acceleration and negative numbers indicating deceleration (e.g., 3 km / s). In summary, the vehicle driving characteristics are represented by the vector: [001 (user ID), 55 (current vehicle speed km / h), 60 (average vehicle speed km / h), 50 (left range of vehicle speed km / h), 70 (right range of vehicle speed km / h), 1 (braking status), 3 (acceleration km / s), 0 (correct accelerator pedal applied)].

[0067] For the characteristics of the environmental meteorological dimension, such as weather conditions, which can be represented by numbers, with negative numbers indicating severe weather and positive numbers indicating good weather, such as 2; seasonal information can be recorded by month, such as 10; temperature and time can be identified by specific numbers, such as 21 and 15; visibility is represented by data, with 0 indicating completely invisible and 100 indicating completely visible, such as 98. Therefore, the feature vector of this environmental meteorological characteristic can be represented as [001 (user ID), 2 (good weather), 10 (month), 21 (temperature), 15 (time), 98 (visibility), 0 (correctly pressing the accelerator)].

[0068] After obtaining the features of each dimension, the driver's baseline features can be determined based on these features. Here, the baseline features refer to the driver's characteristics when correctly pressing the accelerator pedal; they represent the driver's typical performance under normal operation, equivalent to the driving habit characteristics of each individual user.

[0069] In some embodiments, determining the driver's baseline characteristics based on the features of each dimension specifically includes:

[0070] The baseline features are determined based on the average values ​​of the features in each dimension corresponding to the driver correctly pressing the accelerator.

[0071] Each data record's features across all dimensions are transformed into corresponding feature vectors, encompassing various factors influencing driving behavior. The average of these feature vectors for all correct accelerator pedal applications is then calculated and denoted as the baseline feature vector, representing the driver's typical performance under normal conditions. Each baseline feature corresponds one-to-one with a driver ID, ensuring each driver has their own independent and personalized baseline feature.

[0072] Based on this, step 220 is performed. The probability of a driver accidentally pressing the accelerator in a specific situation is assessed by calculating the distance between each dimension feature and the baseline feature vector. If the distance between the feature vector of a certain data point and the baseline feature vector significantly exceeds the set threshold, it indicates that the driving behavior may deviate from the normal operating mode, and there is a risk of accidentally pressing the accelerator.

[0073] The differences between each dimension feature and the baseline feature can be measured using chi-square distance, cosine distance, Euclidean distance, etc. In some embodiments, the chi-square distance is used to quantify the degree of difference between two vectors using contingency table analysis. This is achieved by calculating the chi-square statistic. A large chi-square statistic suggests a significant difference in the variable values ​​between the two vectors. Its basic calculation formula is Equation (1).

[0074] (1)

[0075] in, yes , The dimension; It is a vector The The values ​​for each dimension, It is a vector In the Expected value (average) over the dimension; It is a vector The The values ​​for each dimension, It is a vector In the The expected value in the dimension.

[0076] Because one of the vectors being compared The average eigenvector, i.e. = = Therefore, the chi-square distance formula can be simplified to formula (2).

[0077] (2)

[0078] In some implementations, based on the differences between the features of each dimension and the baseline features, the correctness of the driver's accelerator pedal operation is predicted. Specifically, step 220 includes:

[0079] Based on the information gain weights of each dimension of features and the differences between each dimension of features and the baseline features, the system predicts whether the driver's accelerator pedal operation is correct.

[0080] The information gain weight of any dimension feature is determined based on the probability of the driver pressing the accelerator correctly and incorrectly when the feature of any dimension is included, and the probability of the driver pressing the accelerator correctly and incorrectly when the feature of any dimension is excluded.

[0081] Specifically, the chi-square distance is used to characterize the distance between the feature vector of each dimension and the baseline feature. The smaller the distance, the closer it is to the baseline feature, meaning the greater the probability that pressing the accelerator was the correct operation. When calculating the distance using the chi-square distance formula, each dimension has the same weight. However, in reality, the factors that determine whether a driver presses the accelerator are different, and different features should have different effects on the prediction results. This embodiment uses information gain to improve the chi-square distance. Before accumulation, each dimension is multiplied by its corresponding weight, i.e., the information gain of that feature. Information gain is specific to each feature; for a given feature, the difference between including it and not including it in the overall information entropy is the amount of information that feature brings to the whole. For a feature... Its information gain The calculation formula is formula (3).

[0082] (3)

[0083] in, For category (In this embodiment, the categories are correct acceleration and incorrect acceleration) entropy; It is characterized as At that time, category The conditional entropy. Assuming category have Different values , , ,…, ,but The calculation formula is formula (4).

[0084] (4)

[0085] in, for The probability of occurrence. According to the conditional probability formula, we can obtain... The calculation formula is formula (5).

[0086] (5)

[0087] in, for The probability of occurrence for The probability of it not appearing Features In the case of occurrence The probability of occurrence; Features If it does not occur The probability of occurrence.

[0088] In summary, features The formula for calculating information gain is formula (6).

[0089] (6)

[0090] For the data in this embodiment, the only sample category is "correctly pressing the accelerator". and stepping on the gas incorrectly Two kinds, , These represent the probabilities of the driver correctly pressing the accelerator and incorrectly pressing the accelerator, respectively, when considering this feature. Similarly, , These represent the probabilities of the driver correctly pressing the accelerator and the driver incorrectly pressing the accelerator, respectively, without considering this feature.

[0091] The improved formula, which incorporates information gain weights, is expressed as formula (7).

[0092] (7)

[0093] in, for Information gain of the corresponding feature subsequence.

[0094] The method provided in this invention, by combining the information gain weights of each dimension feature and the differences between each dimension feature and the baseline feature, can predict whether the driver's accelerator pedal operation is correct, thereby further improving the accuracy of the prediction.

[0095] In some embodiments, Figure 3 This is the second flowchart illustrating the method for obtaining the accidental step detection model provided by this invention, as shown below. Figure 3 As shown, step 220 predicts whether the driver's accelerator pedal operation is correct based on the differences between each dimension of features and the baseline features. Specifically, this includes:

[0096] Step 221: Fuse the features from each dimension to obtain the fused features;

[0097] Step 222: Based on the differences between each dimension feature and the baseline feature, as well as the differences between the fused feature and the baseline feature, predict whether the driver's accelerator pedal operation is correct.

[0098] Specifically, in order to further improve the accuracy of detecting accidental acceleration, features from various dimensions can be fused to obtain fused features. For example, the features of each dimension can be represented as follows: [001 (user ID), 1 (male), 9 (driving experience), 35 (age), 0 (no voice prompt for passengers), 0 (correctly pressing the accelerator)], [001 (user ID), 114.3321 (latitude), 30.35216 (longitude), 1 (provincial road), 0 (no obstacles), 1 (speed limit), 60km / h (speed limit threshold), 0 (correctly pressing the accelerator)], [001 (user ID), 55 (current vehicle speed km / h), 60 (average vehicle speed km / h), 50 (left interval value of vehicle speed range km / h), 70 (right interval value of vehicle speed range km / h), 1 (braking status), 3 (marked acceleration km / s), 0 (correctly pressing the accelerator)], [001 (user ID), 2 (good weather), 10 (month), 21 (temperature), 15 (time), 98 (visibility), 0 (correctly pressing the accelerator)].

[0099] The fused features obtained by fusing the features of the above four dimensions can be represented as: [001 (user ID), 1 (male), 9 (driving experience), 35 (age), 0 (no voice prompt for passengers), 0 (correctly pressing the accelerator), 114.3321 (latitude), 30.35216 (longitude), 1 (provincial road), 0 (no obstacles), 1 (speed limit), 60km / h (speed limit threshold), 55 (current vehicle speed), 60 (average vehicle speed), 50 (left interval of vehicle speed range), 70 (right interval of vehicle speed range), 1 (braking status), 3 (marked acceleration km / s), 2 (good weather), 10 (month), 21 (temperature), 15 (time), 78 (visibility), 0 (correctly pressing the accelerator)].

[0100] The difference between the fused features and the baseline features, as well as the difference between each dimension feature and the baseline features, is then calculated to predict whether the driver's accelerator pedal operation is correct. The difference here can be measured by chi-square distance.

[0101] In some embodiments, for the prediction module, a backpropagation (BP) neural network with 5 input layers, 5 hidden layers, and 1 output layer can be used for fuzzy computation. Figure 4 This is a schematic diagram of the prediction module provided by the present invention, as shown below. Figure 4As shown, the driving factor distance refers to the difference between the characteristics of the driving-related dimension and the baseline characteristics; the location factor distance refers to the difference between the characteristics of the vehicle location road dimension and the baseline characteristics; the vehicle factor distance refers to the difference between the characteristics of the vehicle driving dimension and the baseline characteristics; the external factor distance refers to the difference between the characteristics of the environmental and meteorological dimension and the baseline characteristics; and the comprehensive factor distance refers to the difference between the fused characteristics and the baseline characteristics.

[0102] First, the five normalized distances are used as input, with 1 for accidentally pressing the accelerator and 0 for correctly pressing it, for training. Then, these labeled data are used again as input to calculate the output value and determine the probability threshold, typically around 0.5. In practice, the five distances in real-time are used as input to calculate the probability of the output, which is compared to the threshold. If the probability is greater than the threshold, the driver is alerted to the possibility of accidentally pressing the accelerator.

[0103] Based on any of the above embodiments, the determination of the driver's accelerator pedal operation further includes:

[0104] If the judgment result indicates that the accelerator pedal operation was a mistake, a reminder message will be sent.

[0105] Specifically, considering that the existing accidental acceleration suppression solution relies on the vehicle's infotainment system to suppress acceleration, there is a possibility that the system may misjudge the situation in certain scenarios. If the system misjudges the situation and triggers the acceleration suppression command incorrectly, there is also a safety risk.

[0106] In this embodiment, if the judgment result indicates that the driver's current accelerator pedal operation is a mistake, a reminder message is sent. The reminder message can be a visual signal or an audio signal. For example, a rapid voice prompt can be used to urge the driver to actively withdraw the accelerator pedal operation, adding an extra layer of driver-initiated judgment and enhancing safety.

[0107] Based on any of the above embodiments, the method for determining accidental stepping also includes:

[0108] Driver identification;

[0109] If the identity recognition result indicates that the driver is a new user, then a misstep judgment model corresponding to the driver is constructed based on the current driving data;

[0110] If the identity recognition result indicates that the driver is not a new user, then the wrong pedal judgment model corresponding to the driver is loaded, and the wrong pedal judgment model is updated based on the current driving data.

[0111] Specifically, the system utilizes in-vehicle cameras, facial recognition technology, and other devices to collect real-time images of the driver's face or other biometric information. Other authentication methods can also be used, such as fingerprint recognition, iris recognition, or verification using physical credentials like car keys. The collected identity information is then compared with a driver database pre-stored in the system. If the driver's information already exists in the database, an identity match is performed; otherwise, the user is identified as new.

[0112] If the identification result indicates that the driver is a new user, the system needs to build a mis-press judgment model corresponding to that driver. During the initial driving process, the system records the current driving data in real time. Based on a comprehensive analysis of the driver's driving habits (such as rapid acceleration, sudden braking, etc.) and driving environment (such as weather, road conditions, etc.), a mis-press judgment model for that driver is formed.

[0113] If the identification result indicates that the driver is not a new user, the system should update the accidental pedal judgment model corresponding to that driver based on the driver's current driving data. During each driving process, the system will record new driving data in real time and regularly maintain the training set for model training to provide more personalized and accurate predictions.

[0114] Based on any of the above embodiments Figure 5 This is the second flowchart of the accidental step detection method provided by the present invention, as shown below. Figure 5 As shown, the method includes:

[0115] First, the driver's identity is confirmed using in-vehicle cockpit sensors, such as facial recognition cameras. If the identification result indicates that the driver is a new user, a mis-pedal judgment model corresponding to the driver is constructed; if the identification result indicates that the driver is not a new user, a personalized mis-pedal judgment model for the driver is loaded.

[0116] Secondly, during the driver's driving process, real-time driving data is acquired, which includes at least one of the following: vehicle location and road information, vehicle driving information, environmental weather information, and driving-related information.

[0117] Finally, based on the current driving data and the driver's mis-pressing judgment model, the driver's accelerator pedal operation is judged. If the current operation is determined to be a mis-pressing of the accelerator (the model predicts that the brake should be pressed in the current scenario, but the driver actually presses the accelerator), then the driver is given an urgent voice reminder to check if the accelerator has been pressed.

[0118] Finally, each piece of scene data can be recorded as a valid piece of information, and the training set can be maintained regularly for model training to provide more personalized and accurate predictions.

[0119] The method provided in this invention combines vehicle network big data with AI vision and voice technology. It uses facial recognition to identify the driver, matches different drivers with personalized driving models, and accesses data on traffic elements, interference elements, and other elements from external visual recognition. This data, combined with automatic positioning and road regulations, is used to compare the current driving data with the data from the driver's personalized driving model to determine whether to issue a warning message for incorrect pedal operation, thus helping the driver correct their driving. Based on data collected through machine learning algorithm training, the more mileage the driver drives and the more accurate the prediction becomes. Issuing voice reminders based on predicted data to assist the driver in correcting incorrect pedal operation adds a layer of proactive judgment by the driver, preventing secondary errors caused by misjudgment.

[0120] The accidental step detection device provided by the present invention is described below. The accidental step detection device described below can be referred to in correspondence with the accidental step detection method described above.

[0121] Based on any of the above embodiments Figure 6 This is a schematic diagram of the accidental step detection device provided by the present invention, as shown below. Figure 6 As shown, the device includes:

[0122] The data acquisition unit 610 is used to acquire current driving data, which includes at least one of vehicle location road information, vehicle driving information, environmental weather information, and driving-related information.

[0123] The accidental pedal judgment unit 620 is used to judge the driver's accelerator pedal operation based on the current driving data and the accidental pedal judgment model corresponding to the driver. The accidental pedal judgment model is trained based on the driver's historical driving data.

[0124] The accidental accelerator pedal detection device provided in this embodiment of the invention obtains an accidental accelerator pedal detection model corresponding to the driver by collecting the driver's historical driving data. The accidental accelerator pedal detection model can reflect the driver's personal driving habits. Compared with the use of general configuration parameter thresholds in related technologies, the accidental accelerator pedal detection model in this embodiment is more personalized and targeted, and can effectively make correct judgments for different driving habits. As a result, it can more accurately predict whether the accelerator pedal has been accidentally pressed, thereby improving the accuracy of accidental accelerator pedal detection.

[0125] Based on any of the above embodiments, a model acquisition unit is further included, used for:

[0126] Feature extraction is performed on the driver's historical driving data for each information dimension to obtain the features of each dimension, and the baseline features of the driver are determined based on the features of each dimension.

[0127] Based on the differences between the features of each dimension and the baseline features, the driver's accelerator pedal operation is predicted, and the accidental pedal judgment model is trained based on the prediction results.

[0128] Based on any of the above embodiments, the model acquisition unit is specifically used for:

[0129] Based on the information gain weights of the features in each dimension and the differences between the features in each dimension and the baseline features, the driver's accelerator pedal operation is predicted.

[0130] The information gain weight of any dimension feature is determined based on the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the feature is included, and the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the feature is not included.

[0131] Based on any of the above embodiments, the model acquisition unit is specifically used for:

[0132] The features from each dimension are fused to obtain the fused features;

[0133] Based on the differences between the features of each dimension and the baseline features, as well as the differences between the fused features and the baseline features, the driver's accelerator pedal operation is predicted.

[0134] Based on any of the above embodiments, the model acquisition unit is specifically used for:

[0135] The baseline feature is determined based on the average value of each dimension of features corresponding to the driver correctly pressing the accelerator.

[0136] Based on any of the above embodiments, a reminder unit is further included, used for:

[0137] If the judgment result indicates that the accelerator pedal operation was a mistake, a reminder message will be sent.

[0138] Based on any of the above embodiments, an identity recognition unit is further included, for:

[0139] The driver's identity is verified;

[0140] If the identity recognition result indicates that the driver is a new user, then a misstep judgment model corresponding to the driver is constructed based on the current driving data;

[0141] If the identity recognition result indicates that the driver is not a new user, then the accidental step judgment model corresponding to the driver is updated based on the current driving data.

[0142] Figure 7 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 7As shown, the electronic device may include a processor 710, a communications interface 720, a memory 730, and a communication bus 740, wherein the processor 710, communications interface 720, and memory 730 communicate with each other via the communication bus 740. The processor 710 can call logical instructions in the memory 730 to execute a method for determining accidental acceleration. This method includes: acquiring current driving data, which includes at least one of vehicle location road information, vehicle driving information, and environmental weather information, as well as driving-related information; and judging the driver's accelerator pedal operation based on the current driving data and an accidental acceleration judgment model corresponding to the driver, wherein the accidental acceleration judgment model is trained based on the driver's historical driving data.

[0143] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0144] On the other hand, the present invention also provides a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the accidental pedal judgment method provided by the above methods. The method includes: acquiring current driving data, which includes at least one of vehicle location road information, vehicle driving information, and environmental weather information, as well as driving-related information; judging the driver's accelerator pedal operation based on the current driving data and an accidental pedal judgment model corresponding to the driver, wherein the accidental pedal judgment model is trained based on the driver's historical driving data.

[0145] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the method for judging accidental pedal use provided by the above methods. The method includes: acquiring current driving data, the current driving data including at least one of vehicle location road information, vehicle driving information, and environmental weather information, as well as driving-related information; judging the driver's accelerator pedal operation based on the current driving data and an accidental pedal use judgment model corresponding to the driver, wherein the accidental pedal use judgment model is trained based on the driver's historical driving data.

[0146] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0147] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting accidental stepping, characterized in that, include: Acquire current driving data, which includes at least one of vehicle location road information, vehicle driving information, environmental weather information, and driving-related information; Based on the current driving data and the driver's mis-press judgment model, the driver's accelerator pedal operation is judged. The mis-press judgment model is trained based on the driver's historical driving data. The steps for obtaining the accidental step detection model include: Feature extraction is performed on the driver's historical driving data for each information dimension to obtain the features of each dimension. Based on the average value of the features of each dimension when the driver correctly presses the accelerator, the baseline features of the driver are determined. Based on the differences between the features of each dimension and the baseline features, the driver's accelerator pedal operation is predicted, and the accidental pedal judgment model is trained based on the prediction results. The step of predicting the driver's accelerator pedal operation based on the difference between the features of each dimension and the baseline features includes: The differences between each dimension feature and the baseline feature are calculated separately. Before the distance is accumulated, the difference of each dimension is multiplied by the information gain weight of the corresponding dimension feature to obtain a distance measurement result that reflects the degree of deviation between the current driving behavior and the individual's driving habits. The information gain weight of any dimension feature is determined based on the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the driver includes the dimension feature, and the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the driver does not include the dimension feature. Based on the distance measurement results, the driver's accelerator pedal operation is predicted.

2. The method for determining accidental stepping according to claim 1, characterized in that, The prediction of the driver's accelerator pedal operation based on the differences between the features of each dimension and the baseline features includes: The features from each dimension are fused to obtain the fused features; Based on the differences between the features of each dimension and the baseline features, as well as the differences between the fused features and the baseline features, the driver's accelerator pedal operation is predicted.

3. The method for determining accidental stepping according to claim 1 or 2, characterized in that, The process of judging the driver's accelerator pedal operation further includes: If the judgment result indicates that the accelerator pedal operation was a mistake, a reminder message will be sent.

4. The method for determining accidental stepping according to claim 1 or 2, characterized in that, The method further includes: The driver's identity is verified; If the identity recognition result indicates that the driver is a new user, then a misstep judgment model corresponding to the driver is constructed based on the current driving data; If the identity recognition result indicates that the driver is not a new user, then the wrong-stepping judgment model corresponding to the driver is loaded, and the wrong-stepping judgment model is updated based on the current driving data.

5. A device for detecting accidental stepping, characterized in that, include: The data acquisition unit is used to acquire current driving data, which includes at least one of vehicle location road information, vehicle driving information, and environmental weather information, as well as driving and riding related information. The accidental pedal judgment unit is used to judge the driver's accelerator pedal operation based on the current driving data and the accidental pedal judgment model corresponding to the driver. The accidental pedal judgment model is trained based on the driver's historical driving data. The steps for obtaining the accidental step detection model include: Feature extraction is performed on the driver's historical driving data for each information dimension to obtain the features of each dimension. Based on the average value of the features of each dimension when the driver correctly presses the accelerator, the baseline features of the driver are determined. Based on the differences between the features of each dimension and the baseline features, the driver's accelerator pedal operation is predicted, and the accidental pedal judgment model is trained based on the prediction results. The step of predicting the driver's accelerator pedal operation based on the difference between the features of each dimension and the baseline features includes: The differences between each dimension feature and the baseline feature are calculated separately. Before the distance is accumulated, the difference of each dimension is multiplied by the information gain weight of the corresponding dimension feature to obtain a distance measurement result that reflects the degree of deviation between the current driving behavior and the individual's driving habits. The information gain weight of any dimension feature is determined based on the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the driver includes the dimension feature, and the probability of the driver correctly pressing the accelerator and incorrectly pressing the accelerator when the driver does not include the dimension feature. Based on the distance measurement results, the driver's accelerator pedal operation is predicted.

6. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the accidental step detection method as described in any one of claims 1 to 4.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the accidental step detection method as described in any one of claims 1 to 4.

Images