Intelligent automobile driving fingerprint authentication method and system
By extracting and optimizing the feature of the car CAN bus data, combining Bayesian network and CNN-SVDD model, the problems of low accuracy and weak adaptability of traditional intelligent vehicle driving fingerprint recognition methods are solved, and efficient and accurate driver identity authentication is achieved.
Patent Information
- Application Number
- CN202510215389.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-26
- Publication Date
- 2025-06-13
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Traditional smart car driving fingerprint recognition methods have low detection accuracy and weak adaptability, so they cannot effectively monitor the driver's identity in real time.
By extracting the original data of the car CAN bus, data preprocessing and optimizing the relationship between driving data and driver identity is analyzed using a Bayesian network classifier based on mutual information, low correlation characteristics are eliminated, driving fingerprint authentication model is constructed, and the CNN and SVDD models are combined for identification.
It realizes fast and accurate driver fingerprint authentication, improves recognition accuracy and identity recognition delay, and enhances real-time monitoring capabilities of driver identity.
Smart Images
Figure CN120145366A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of fingerprint authentication, and particularly relates to a fingerprint authentication method and system for intelligent vehicle driving. Background Art
[0002] With the continuous development of current technology, in many special vehicles, the driver needs to be authorized to drive the vehicle. For example, a cash transport vehicle requires an authorized driver to ensure the safety of a large amount of cash; in order to provide a safe and comfortable public transportation environment for passengers, only authorized drivers are allowed to drive buses, coaches, subways, etc.; vehicles of special forces should only be driven by authorized soldiers to ensure that weapons and combat equipment can reach the destination safely. Currently, identification technologies such as fingerprint recognition and iris recognition cannot monitor drivers in real time.
[0003] Regarding the existing fingerprint recognition method for intelligent vehicle driving, there are the following problems: the traditional driving has low accuracy in detecting driving fingerprints and weak adaptability, so it needs to be further improved. Summary of the Invention
[0004] Aiming at the deficiencies of the prior art, the purpose of the present invention is to provide a fingerprint authentication method and system for intelligent vehicle driving, which solves the problems in the background art.
[0005] The purpose of the present invention can be achieved by the following technical solutions:
[0006] A fingerprint authentication method for intelligent vehicle driving includes the following steps:
[0007] Extract features from the original data of the vehicle CAN bus to obtain driving data, perform data preprocessing, and divide the preprocessed data into a sample set;
[0008] Based on the preprocessed data, optimize the driving data, analyze the correlation between the driving data and the driver's identity using a Bayesian network classifier based on mutual information, and obtain a new driving data sample set by eliminating low-correlation features;
[0009] Based on the new driving data sample set, construct a driving fingerprint authentication model.
[0010] Further, extracting features from the original data of the vehicle CAN bus to obtain driving data is specifically as follows:
[0011] Collect the corresponding original data from the OBD-II port through an automotive diagnostic tool.
[0012] Further, the driving data is: extracting continuously changing data from the driving data received from the vehicle as the driving data.
[0013] Further, preprocess the driving data as follows:
[0014] Convert the hexadecimal data into decimal data. At the same time, unify the data sending frequency, select one second as a time unit, calculate the average value of each driving data within each second, and generate a t×m 0 sample set of. Use the maximum-minimum normalization method to limit the value range of all sample sets of t×m 0 between 0 and 1.
[0015] Further, based on the preprocessed data, optimize the driving data, and use the Bayesian network classifier based on mutual information to analyze the correlation between the driving data and the driver's identity, and obtain a new driving data sample set by eliminating features with low correlation, as follows:
[0016] Map the m 0 driving data of the driver to the driver's identity one by one. Calculate the correlation degree between the driving data and the driver's identity through the Bayesian classifier. Subsequently, retain the driving data with high correlation as feature data, and delete the driving data with low correlation. The sample set of t×m 0 will become a feature sample set of t×m.
[0017] Further, based on the new driving data sample set, construct a driving fingerprint authentication model as follows:
[0018] Construction of the CNN network: The input of the CNN network is a tensor, that is, "image width × image height × image depth". Therefore, first divide the input data into densely overlapping sliding windows. It should be noted that the width of the sliding window is set to m to better learn the relationship between m feature data. The length of the sliding window is a variable quantity. Finally, the input data of the CNN network forms multiple tensors of 300×m,
[0019] Before the data is input into the CNN network, a normalization operation must be performed first to transform the feature data of different scales into the same scale for training. A 21×m convolutional kernel is used to extract the features of each part. In addition, the number of channels is set to 256 and the stride is set to 1. After the first layer of convolution, a 280×1 matrix will be formed. After the operation of the C1 layer, the output of the C1 layer is used as the input of the pooling layer. The function of the pooling layer is to extract the significant features of the driver. The number of channels of the C1 layer is 256, and the number of channels remains unchanged after the calculation of the pooling layer. Immediately after the pooling layer is the work of the fully connected layer, which performs high-level reasoning on the refined data to realize the mapping between the label and the data features. In the pooling layer, the neurons in each layer are fully connected to the previous layer and the next layer. After calculating the corresponding cross-entropy loss through the activation function, the identity authentication of n drivers is finally realized;
[0020] When using the SVDD model to implement the illegal driver detection work, it is necessary to find the smallest sphere that contains almost all of the training data set. It should be noted that in the model of the training data, there will always be individual abnormal data that are far from the center of the hypersphere. If they are included in the hypersphere, a very large circle will be generated, which cannot well represent the features of other data. In the work of illegal driver detection, the sequential minimal optimization algorithm is used to optimize its value. By determining the radius r and the center 0 in the hyperdimensional space, when any other data is input, the illegal driver detection work can be directly realized according to the data;
[0021] After the driver's driving feature extraction work is carried out by the CNN, then the input of the fully connected layer is used as the input of the SVDD model. It should be noted that since the data of 128 dimensions will increase the computational burden of the SVDD model, the PCA algorithm needs to be used to reduce the feature dimension to make it easier for later processing. The PCA algorithm can alleviate the curse of dimensionality of the sample data (128 features) while also retaining the features of the driver's driving behavior data as much as possible, enabling the driving fingerprint authentication to identify the driver with higher accuracy in a shorter time. Based on the PCA algorithm theory, the 128-dimensional data of the fully connected layer is transformed into 3 dimensions and then input into the SVDD model for unauthorized driver detection work.
[0022] An intelligent vehicle driving fingerprint authentication system, comprising:
[0023] A preprocessing module, configured to extract features from the original data of the vehicle CAN bus to obtain driving data, perform data preprocessing, and divide the preprocessed data into a sample set;
[0024] An identification module, configured to optimize driving data based on the preprocessed data, analyze the correlation between the driving data and the driver identity by using a Bayesian network classifier based on mutual information, and obtain a new driving data sample set by eliminating features with low correlation;
[0025] A construction module, configured to construct a driving fingerprint authentication model based on the new driving data sample set.
[0026] The explanations of the nouns, conjunctions or adjectives involved in the above technical solutions are as follows:
[0027] Fixed connection means that after the parts or components are fixed, there is no relative movement between them. It is divided into two types: detachable connection and non-detachable connection.
[0028] (1) Detachable connection uses screws, splines, wedge pins, etc. to fix the parts together. This connection method can be disassembled during maintenance without damaging the parts. However, the specifications of the connecting parts used must be correct (such as the length of bolts, keys, and wedge pins), and they must be fastened properly.
[0029] (2) Non-detachable connection mainly refers to welding, riveting, and mortise fitting, etc. Since it needs to be forged, sawed, or oxy-cut to be disassembled during maintenance or replacement, the spare parts generally cannot be used twice. At the same time, during connection, attention should be paid to process quality, technical inspection, and remedial measures (such as correction, polishing, etc.).
[0030] Threaded connection means a detachable connection that connects the connected parts together with threaded parts (or the threaded parts of the connected parts).
[0031] Sliding connection means that two objects are in contact but not fixed, and they can slide relative to each other.
[0032] Rotational connection means that the connection between parts allows the parts to rotate relative to each other.
[0033] The beneficial effects of the present invention:
[0034] Taking the output of the fully connected layer in the CNN model as the input of the SVDD model, by using this combined mode of CNN and SVDD, fast and accurate driver fingerprint authentication is realized, with a high accuracy rate of driver fingerprint recognition and a small time delay in identifying the driver identity. Description of the Drawings
[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.
[0036] Figure 1It is the flowchart of the fingerprint authentication method according to the embodiment of the present invention;
[0037] Figure 2 It is the flowchart of the CNN-SVDD algorithm according to the embodiment of the present invention;
[0038] Figure 3 It is the combined model of CNN and SVDD according to the embodiment of the present invention. Detailed implementation manners
[0039] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.
[0040] In the description of the present invention, it should be understood that the terms "opening", "upper", "lower", "thickness", "top", "middle", "length", "inner", "perimeter", etc. indicating the orientation or positional relationship are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the components or elements referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation of the present invention.
[0041] Supplementary explanations for the English abbreviations involved in the present invention are as follows:
[0042] CNN (Convolutional Neural Network), Convolutional Neural Network
[0043] SVDD (Support Vector Domain Description, Support Vector Domain Description)
[0044] CAN (Controller Area Network, Controller Area Network)
[0045] OBD-II (On-Board Diagnostics-II, On-Board Diagnostic Interface)
[0046] PCA (Principal Component Analysis, Principal Component Analysis)
[0047] Embodiment 1:
[0048] Please refer to Figure 1 and Figure 3 The present invention provides a technical solution: an intelligent vehicle driving fingerprint authentication method, including the following steps:
[0049] Step 1: Extract features from the original data of the automotive CAN bus to obtain driving data, perform data preprocessing, and divide the preprocessed data into a sample set;
[0050] Step 2: Optimize the driving data based on the preprocessed data, analyze the correlation between the driving data and the driver's identity using a Bayesian network classifier based on mutual information, and obtain a new driving data sample set by eliminating features with low correlation;
[0051] Step 3: Build a driving fingerprint authentication model based on the new driving data sample set.
[0052] In Step 1: The original data of the automotive CAN bus is used for feature extraction in the experiment, and its corresponding original data is collected from the OBD-II port through an automotive diagnostic tool.
[0053] Among them, since there is a lot of interference data in the vehicle that has nothing to do with the driver's driving behavior, these data need to be removed to improve the accuracy of fingerprint authentication. Extract the continuously changing data from the driving data received from the vehicle as the driving data. Suppose there are m 0 pieces of driving data in the vehicle.
[0054] Among them, the preprocessing of the driving data is as follows: Convert the hexadecimal data into decimal data to facilitate data processing. Secondly, since the data sending frequencies of different IDs in the vehicle are different, it is necessary to unify the data sending frequency. Select one second as a time unit, calculate the average value of each driving data per second, and generate a t×m 0 sample set for the data of t seconds. Use the maximum-minimum normalization method to limit the value range of the entire t×m 0 sample set between 0 and 1.
[0055] In Step 2: Among the m 0 pieces of driving data collected, there may be a part of the data that cannot reflect the driver's identity characteristics. Therefore, it is necessary to analyze the correlation between multiple driving data and the driver's identity, use a Bayesian network classifier based on mutual information to achieve the correlation analysis, and by eliminating features with low correlation, the m 0 driving data sample set will be changed to a new sample set of m features, where m ≤ m 0 . m is regarded as the feature data of the driving fingerprint authentication model.
[0056] Specifically, map the m 0 pieces of driving data of the driver to the driver's identity one by one, calculate the correlation degree between the driving data and the driver's identity through the Bayesian classifier. Subsequently, retain the driving data with high correlation degree as the feature data, and delete the driving data with low correlation degree. t×m0 The sample set will become a feature sample set of t×m.
[0057] In step 3: Construction of the CNN network: The input of the CNN network is a tensor, that is, "image width × image height × image depth". Therefore, the input data needs to be first divided into densely overlapping sliding windows. It should be noted that the width of the sliding window is set to m to better learn the relationship between m feature data. The length of the sliding window is a variable quantity. Finally, the input data of the CNN network forms multiple tensors of 300×m.
[0058] Before the data is input into the CNN network, it needs to be first normalized to train the feature data of different scales into the same scale. A 21×m convolutional kernel is used to extract the features of each part. In addition, the number of channels is set to 256 and the stride is set to 1. After the first layer of convolution, a 280×1 matrix will be formed. After the operation of the C1 layer, the output of the C1 layer is used as the input of the pooling layer. The function of the pooling layer is to extract the significant features of the driver. The number of channels of the C1 layer is 256, and the number of channels remains unchanged after the calculation of the pooling layer. Immediately after the pooling layer is the work of the fully connected layer. It performs high-level inference on the refined data to realize the mapping between the label and the data features. In the pooling layer, the neurons in each layer are all connected to the previous layer and the next layer. After calculating the corresponding cross-entropy loss through the activation function, the identity authentication of n drivers is finally realized.
[0059] The SVDD model is used to implement the detection of illegal drivers. It is necessary to find the smallest sphere that contains almost all of the training data set. It should be noted that in the model of the training data, there will always be individual abnormal data that are far from the center of the hypersphere. If they are included in the hypersphere, a very large circle will be generated, which cannot well represent the features of other data. In the work of detecting illegal drivers, the sequential minimal optimization algorithm is used to optimize its value. By determining the radius r and the center 0 in the hyperdimensional space, when any other data is input, the detection of illegal drivers can be directly realized according to the data.
[0060] After the driver's driving features are extracted by the CNN, the input of the fully connected layer is then used as the input of the SVDD model. It should be noted that since the 128-dimensional data will increase the computational burden of the SVDD model, the PCA algorithm needs to be used to reduce the feature dimension to make it easier for later processing. The PCA algorithm can alleviate the curse of dimensionality of the sample data (128 features) while retaining the features of the driver's driving behavior data as much as possible, enabling the driving fingerprint authentication to identify the driver with higher accuracy in a shorter time. Based on the PCA algorithm theory, the 128-dimensional data of the fully connected layer is converted into 3 dimensions and then input into the SVDD model for unauthorized driver detection.
[0061] Embodiment 2:
[0062] This embodiment provides an intelligent vehicle driving fingerprint authentication system for implementing the intelligent vehicle driving fingerprint authentication method described in Embodiment 1, including:
[0063] A preprocessing module for extracting features from the original data of the vehicle CAN bus to obtain driving data, performing data preprocessing, and dividing the preprocessed data into a sample set;
[0064] An identification module for optimizing the driving data based on the preprocessed data, analyzing the correlation between the driving data and the driver's identity using a Bayesian network classifier based on mutual information, and obtaining a new driving data sample set by eliminating features with low correlation;
[0065] A construction module for constructing a driving fingerprint authentication model based on the new driving data sample set.
[0066] In the description of this specification, the descriptions referring to terms such as "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.
[0067] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art of this industry should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification only illustrate the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.
Claims
1. A smart car driving fingerprint authentication method, characterized in that: The following steps are involved: Extract features from the original data of the automobile CAN bus to obtain driving data, perform data preprocessing, and divide the preprocessed data into sample sets; Based on the preprocessed data, the driving data is optimized, and the correlation between the driving data and the driver's identity is analyzed using a Bayesian network classifier based on mutual information. A new driving data sample set is obtained by eliminating features with low correlation. Based on the new driving data sample set, a driving fingerprint authentication model is constructed.
2. According to claim 1, a smart car driving fingerprint authentication method is characterized in that: The driving data is obtained by extracting features from the raw data of the car CAN bus, as follows: The corresponding raw data is collected from the OBD-II port through the car diagnostic tool.
3. A smart car driving fingerprint authentication method according to claim 2, characterized in that: The driving data is: the driving data received from the car is extracted and transformed as the driving data.
4. A smart car driving fingerprint authentication method according to claim 3, characterized in that: The driving data is preprocessed as follows: Convert the hexadecimal data into decimal data. At the same time, unify the data sending frequency, select one second as a time unit, calculate the average value of each driving data within one second, generate a sample set of t×m0 for t seconds of data, and use the maximum and minimum normalization method to limit the value range of the entire sample set of t×m0 to between 0 and 1.
5. A smart car driving fingerprint authentication method according to claim 4, characterized in that: Based on the preprocessed data, the driving data is optimized, and the correlation between the driving data and the driver identity is analyzed using the Bayesian network classifier based on mutual information. A new driving data sample set is obtained by eliminating the features with low correlation, as follows: The driver's m0 driving data are mapped one by one to the driver's identity, and the correlation between the driving data and the driver's identity is calculated through the Bayesian classifier. Subsequently, the driving data with high correlation is retained as feature data, and the driving data with low correlation will be deleted. The sample set of t×m0 will become a feature sample set of t×m.
6. A smart car driving fingerprint authentication method according to claim 5, characterized in that: Based on the new driving data sample set, a driving fingerprint authentication model is constructed as follows: Construction of CNN network: The input of CNN network is a tensor, i.e. "image width × image height × image depth". Therefore, the input data should be divided into densely overlapping sliding windows. It should be noted that the width of the sliding window is set to m in order to better learn the relationship between m feature data. The length of the sliding window is a variable quantity. Finally, the input data of CNN network forms multiple 300×m tensors. Before the data is input into the CNN network, it must first be normalized to convert feature data of different scales into the same scale for training. The 21×m convolution kernel is used to extract the features of each part. In addition, the number of channels is set to 256 and the step size is set to 1. After the first layer of convolution, a 280×1 matrix will be formed. After the C1 layer operation, the output of the C1 layer is used as the input of the pooling layer. The function of the pooling layer is to extract the significant features of the driver. The number of channels in the C1 layer is 256. After the pooling layer calculation, the number of channels remains unchanged. The pooling layer is followed by the fully connected layer, which performs high-level reasoning on the streamlined data to achieve the mapping between labels and data features. In the pooling layer, the neurons in each layer are fully connected to the previous and next layers. The corresponding cross entropy loss is calculated through the activation function, and finally the identity authentication of n drivers is realized; The SVDD model is used to implement illegal driver detection. It is necessary to find the smallest sphere that contains almost all training data sets. It should be noted that in the model of training data, there will always be individual abnormal data, far away from the center of the hypersphere. If they are included in the hypersphere, a large circle will be generated, which cannot well represent the characteristics of other data. In the work of illegal driver detection, the sequence minimum optimization algorithm is used to optimize its value. By determining the radius r and the center 0 in the hyperdimensional space, when any other data is input, the illegal driver detection work can be directly implemented based on the data; After CNN extracts the driver's driving features, the input of the fully connected layer is used as the input of the SVDD model. It should be noted that since the 128-dimensional data will increase the computational burden of the SVDD model, the PCA algorithm needs to be used to reduce the feature dimension to make it easier for post-processing. The PCA algorithm can alleviate the dimensionality disaster of the sample data (128 features) while retaining the characteristics of the driver's driving behavior data as much as possible, so that the driving fingerprint authentication can identify the driver in a shorter time and with higher accuracy. Based on the PCA algorithm theory, the 128-dimensional data of the fully connected layer is converted into 3 dimensions and then input into the SVDD model for unauthorized driver detection.
7. A smart car driving fingerprint authentication system, used to implement the smart car driving fingerprint authentication method according to any one of claims 1 to 6, characterized in that: include: The preprocessing module is used to extract features from the original data of the automobile CAN bus to obtain driving data, perform data preprocessing, and divide the preprocessed data into sample sets; The identification module is used to optimize the driving data based on the preprocessed data, analyze the correlation between the driving data and the driver's identity using a Bayesian network classifier based on mutual information, and obtain a new driving data sample set by eliminating features with low correlation; A building module is used to build a driving fingerprint authentication model based on a new driving data sample set.