Vehicle anti-theft detection method, vehicle and storage medium

Abstract

The invention discloses a vehicle anti-theft detection method, a vehicle and a storage medium. The method comprises the steps that in-vehicle environment sensing data of a target vehicle are acquired, and the in-vehicle environment sensing data are used for representing the change condition of the in-vehicle environment; feature extraction is carried out on the in-vehicle environment sensing data to obtain a feature extraction result, and the feature extraction result is used for representing time domain features and frequency domain features corresponding to the in-vehicle environment sensing data; the feature extraction result is analyzed and processed through a target detection model, a target prediction result is obtained, the target prediction result is used for determining whether the target vehicle has a preset type of behavior or not, and the target detection model is obtained by adopting multiple sets of sample training data to conduct machine learning training; and generating and issuing a vehicle safety strategy corresponding to the target vehicle based on the target prediction result. According to the invention, the technical problems of high detection cost and poor reliability during vehicle safety protection in the prior art are solved.

Description

Technical Field

[0001] This application relates to the field of vehicle security technology, and more specifically, to a vehicle anti-theft detection method, a vehicle, and a storage medium. Background Technology

[0002] With the rapid development of intelligent and IoT technologies, vehicle anti-theft and intrusion detection have become key areas for ensuring automotive safety. While detection methods such as seat sensors, vibration sensors, and cameras improve vehicle security to some extent, they generally suffer from high costs, high power consumption, and other economic and environmental burdens. Especially in certain scenarios, such as when windows or doors are not fully closed, security detection efficiency drops significantly, creating significant blind spots and making it difficult to comprehensively and accurately monitor the security situation inside the vehicle. Furthermore, rule-based intrusion detection algorithms can be used, but this approach lacks flexibility and versatility in intrusion judgment, making it difficult to adapt to complex and changing environments, thus affecting the overall performance of vehicle anti-theft systems. Therefore, the high detection costs and poor reliability of current vehicle security technologies severely restrict the effectiveness of vehicle anti-theft systems and pose a challenge to improving user safety and trust.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a vehicle anti-theft detection method, a vehicle, and a storage medium to at least solve the technical problems of high detection costs and poor reliability in related technologies when performing vehicle security protection.

[0005] According to one aspect of the embodiments of this application, a vehicle anti-theft detection method is provided, comprising: acquiring in-vehicle environment perception data of a target vehicle, wherein the in-vehicle environment perception data is used to characterize changes in the in-vehicle environment; performing feature extraction on the in-vehicle environment perception data to obtain feature extraction results, wherein the feature extraction results are used to represent the time-domain features and frequency-domain features corresponding to the in-vehicle environment perception data; analyzing and processing the feature extraction results using a target detection model to obtain a target prediction result, wherein the target prediction result is used to determine whether the target vehicle has a preset type of behavior, and the target detection model is obtained by machine learning training using multiple sets of sample training data; and generating and distributing a vehicle security policy corresponding to the target vehicle based on the target prediction result.

[0006] According to another aspect of the embodiments of this application, a vehicle anti-theft detection device is also provided, comprising: an acquisition module for acquiring in-vehicle environment perception data of a target vehicle, wherein the in-vehicle environment perception data is used to characterize changes in the in-vehicle environment; an extraction module for extracting features from the in-vehicle environment perception data to obtain feature extraction results, wherein the feature extraction results are used to represent the time-domain features and frequency-domain features corresponding to the in-vehicle environment perception data; an analysis module for analyzing and processing the feature extraction results using a target detection model to obtain a target prediction result, wherein the target prediction result is used to determine whether the target vehicle exhibits a preset type of behavior, and the target detection model is obtained by machine learning training using multiple sets of sample training data; and a determination module for generating and issuing a vehicle security policy corresponding to the target vehicle based on the target prediction result.

[0007] According to another aspect of the embodiments of this application, a vehicle is also provided, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods in various embodiments of this application when it runs.

[0008] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0009] According to another aspect of the embodiments of this application, a computer program product is also provided, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0010] According to another aspect of the embodiments of this application, a computer program product is also provided, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0011] According to another aspect of the embodiments of this application, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this application.

[0012] In this embodiment, by acquiring in-vehicle environmental perception data of the target vehicle, and then extracting features from the in-vehicle environmental perception data to obtain feature extraction results, a target detection model is used to analyze and process the feature extraction results to obtain target prediction results. Finally, based on the target prediction results, a vehicle security policy corresponding to the target vehicle is generated and issued. This achieves real-time, intelligent, and accurate in-vehicle environment monitoring, thereby providing an immediate and effective vehicle security policy, significantly improving the overall efficiency of the vehicle security system and user experience. By continuously perceiving and collecting data on the in-vehicle environment of the target vehicle in real time, changes in the in-vehicle environment can be captured in a timely manner, ensuring the timeliness and comprehensiveness of monitoring and providing an accurate information foundation for subsequent processing. The target detection model trained using machine learning can perform in-depth analysis of the feature extraction results and intelligently predict whether a preset type of behavior has occurred inside the vehicle. Compared with traditional rule-based algorithms, it has stronger learning ability and adaptability, can effectively cope with various complex intrusion situations, reduce false positives and false negatives, and further reduce detection costs. Based on the target prediction results, the corresponding vehicle safety strategy can be activated in real time, enabling the vehicle safety system to actively defend itself, which significantly improves the reliability of vehicle safety protection and solves the technical problems of high detection cost and poor reliability in related technologies when performing vehicle safety protection. Attached Figure Description

[0013] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0014] Figure 1 This is a flowchart of a vehicle anti-theft detection method according to an embodiment of this application;

[0015] Figure 2 This is a schematic diagram of the deployment location of an ultra-wideband sensing chip according to an embodiment of this application;

[0016] Figure 3 This is a schematic diagram showing the deployment location of another ultra-wideband sensing chip according to an embodiment of this application;

[0017] Figure 4 This is a schematic diagram illustrating the training and application process of an object detection model according to an embodiment of this application;

[0018] Figure 5 This is a structural block diagram of a vehicle anti-theft detection device according to an embodiment of this application. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] According to an embodiment of this application, a vehicle anti-theft detection method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0022] Figure 1 This is a flowchart of a vehicle anti-theft detection method according to an embodiment of this application, such as... Figure 1 As shown, the method includes the following steps:

[0023] Step S11: Obtain in-vehicle environment perception data of the target vehicle, wherein the in-vehicle environment perception data is used to characterize changes in the in-vehicle environment.

[0024] Step S12: Extract features from the in-vehicle environment perception data to obtain feature extraction results, wherein the feature extraction results are used to represent the time domain features and frequency domain features corresponding to the in-vehicle environment perception data.

[0025] Step S13: The feature extraction results are analyzed and processed using the target detection model to obtain the target prediction result. The target prediction result is used to determine whether the target vehicle has a preset type of behavior. The target detection model is obtained by machine learning training using multiple sets of sample training data.

[0026] Step S14: Generate and distribute the vehicle safety policy corresponding to the target vehicle based on the target prediction results.

[0027] The aforementioned in-vehicle environment perception data can be data about the interior environment of the target vehicle collected in real time through an integrated Ultra-Wideband (UWB) module. This in-vehicle environment perception data can meticulously reflect the dynamic changes in the in-vehicle environment, including but not limited to the movement and shape changes of objects inside the vehicle, or any unexpected activities. Due to the high resolution and low latency of UWB technology, the accuracy and timeliness of the in-vehicle environment perception data can be guaranteed, laying a solid foundation for subsequent processing.

[0028] After acquiring the in-vehicle environment perception data, the data is processed to extract key features reflecting changes in the in-vehicle environment. The feature extraction results can include time-domain features and frequency-domain features, which can more intuitively depict the state changes of the in-vehicle environment.

[0029] In the time domain, in-vehicle environmental perception data typically exists as time-series signals. The feature extraction process analyzes the temporal characteristics of this data to extract features crucial for intrusion detection. Time-domain features may include, but are not limited to, signal energy peaks, attenuation slopes, pulse widths, and signal intervals. A signal energy peak is the maximum energy value of the UWB signal at a specific time point, used to detect the movement of a person or object inside the vehicle; movement causes sudden changes in signal energy. The attenuation slope is the rate of change of signal strength over time or distance, used to identify the direction and speed of an object's movement. Pulse width is the duration of a signal pulse, effectively distinguishing different types of objects or behaviors. Signal intervals are the time intervals between signals, reflecting the stability of an object or environment. Frequency-domain features include, but are not limited to, spectral entropy and band energy distribution. Spectral entropy reflects the uncertainty of the signal's spectral distribution; high spectral entropy may indicate high signal complexity and significant dynamic changes in the in-vehicle environment. Band energy distribution represents the distribution of signal energy across different frequency ranges; different objects or behaviors produce specific frequency distribution patterns. Through meticulous feature extraction, the embodiments of this application can extract deeper information from in-vehicle environment perception data, thereby achieving accurate monitoring of changes in the in-vehicle environment and intelligent identification of abnormal behaviors.

[0030] The aforementioned object detection model is an algorithm trained through machine learning. Its training process is based on multiple sets of sample training data, which include feature extraction results of vehicles in different states, such as an empty, stationary vehicle, normal passenger behavior, and pre-defined types of abnormal behavior, such as illegal intrusion. By learning the correlation between feature extraction results and corresponding behaviors, recognition rules are formed to achieve accurate classification and prediction of new data.

[0031] After obtaining the feature extraction results, the object detection model can classify the input data based on the patterns learned during training, outputting a classification probability distribution to represent the probability that the target vehicle is in various preset behavioral states. The target prediction result output by the object detection model after analysis can be a classification label that can clearly indicate whether there is a preset type of behavior inside the target vehicle. For example, if the object detection model identifies a high signal energy peak in the time domain features and a high spectral entropy in the frequency domain features, it may mean that there is an unexpected object or person moving inside the vehicle. In this case, the object detection model will predict a high risk of illegal intrusion or other abnormal behavior.

[0032] The aforementioned object detection model is obtained by training an initial detection model through machine learning. This initial detection model includes, but is not limited to, Random Forest, Support Vector Machine (SVM), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). Random Forest is an ensemble learning method that predicts targets by constructing multiple decision trees. It is suitable for multi-class classification problems and can handle a large number of features. Support Vector Machine is a supervised learning model that distinguishes different categories of data by finding the optimal decision boundary, and is effective for problems with non-linear boundaries. Convolutional Neural Networks are suitable for processing temporal and frequency domain features, and can automatically learn the spatial and temporal relationships between features, making them effective for classifying image and signal data. Long Short-Term Memory (LSTM) is a type of recurrent neural network that can effectively process time-series data, thereby capturing the trend of feature changes over time. Through the machine learning process, the object detection model can not only make immediate predictions based on the current feature extraction results, but also perform deeper analysis by combining historical data, improving the accuracy and robustness of predictions, ultimately achieving effective monitoring of in-vehicle behavior and timely early warning of abnormal activities.

[0033] The target prediction results described above are the output of the target detection model. They indicate whether a preset type of behavior exists inside the target vehicle. This preset type of behavior could be, for example, illegal intrusion or vandalism targeting the vehicle. The target prediction results typically include one or more category labels, along with corresponding confidence or probability values, clearly indicating whether the current state inside the vehicle is normal and the specific type of abnormal behavior.

[0034] Based on the target prediction results, corresponding vehicle security policies can be dynamically generated and issued. The formulation of vehicle security policies needs to consider the behavior type, confidence level, and the vehicle's current status, such as whether it is in a parking lot or locked. Vehicle security policies include, but are not limited to: real-time warnings, remote vehicle locking, recording and uploading data, notifying the vehicle owner or security center, and linking with other security systems. When illegal intrusion is predicted inside the vehicle, an audible and visual alarm signal can be triggered immediately to warn potential intruders and alert nearby people. Through vehicle networking technology, the target vehicle can be remotely controlled, instantly locking all doors and windows to prevent further intrusion. Furthermore, the target prediction results, raw data collected by the UWB module, and feature vectors can be recorded and uploaded to a cloud server and the vehicle owner's terminal device via the vehicle networking system as a basis for subsequent investigations. This application embodiment can also send alarm information to the vehicle owner's mobile phone or other devices and simultaneously notify the vehicle security service center to take further security actions. When linking with other security systems, the in-vehicle camera can be activated to record video, the auxiliary system on the vehicle bumper can be controlled to enter a protective state, or the vehicle's tracking system can be activated to provide information on the intruder's movement path.

[0035] Based on the prediction results of the target detection model, effective vehicle safety strategies can be activated in real time, which can effectively prevent and respond to security risks such as illegal intrusion. This not only improves the response speed of security protection and reduces potential property losses, but also ensures the accuracy of measures through intelligent decision-making, avoids unnecessary false alarms, and significantly improves the efficiency and reliability of security protection.

[0036] Based on steps S11 to S14 above, by acquiring in-vehicle environmental perception data of the target vehicle, feature extraction is performed on the in-vehicle environmental perception data to obtain feature extraction results. Subsequently, the feature extraction results are analyzed and processed using a target detection model to obtain target prediction results. Finally, based on the target prediction results, a vehicle safety policy corresponding to the target vehicle is generated and issued. This achieves real-time, intelligent, and accurate in-vehicle environment monitoring, thereby providing an immediate and effective vehicle safety policy and significantly improving the overall efficiency and user experience of the vehicle safety system. By continuously perceiving and collecting data on the in-vehicle environment of the target vehicle in real time, changes in the in-vehicle environment can be captured in a timely manner, ensuring the timeliness and comprehensiveness of monitoring and providing an accurate information foundation for subsequent processing. The target detection model trained using machine learning can perform in-depth analysis of the feature extraction results and intelligently predict whether a preset type of behavior has occurred inside the vehicle. Compared with traditional rule-based algorithms, it has stronger learning ability and adaptability, can effectively cope with various complex intrusion situations, reduce false positives and false negatives, and further reduce detection costs. Based on the target prediction results, the corresponding vehicle safety strategy can be activated in real time, enabling the vehicle safety system to actively defend itself, which significantly improves the reliability of vehicle safety protection and solves the technical problems of high detection cost and poor reliability in related technologies when performing vehicle safety protection.

[0037] The vehicle anti-theft detection method in the embodiments of this application will be further described below.

[0038] In an optional embodiment, step S11, acquiring the in-vehicle environment perception data of the target vehicle includes:

[0039] Step S111: Multiple ultra-wideband sensing chips are used to transmit ultra-wideband pulse signals, wherein the multiple ultra-wideband sensing chips are arranged in pairs symmetrically along the longitudinal centerline of the target vehicle.

[0040] Step S112: Collect channel impulse response data corresponding to the ultra-wideband pulse signal, wherein the channel impulse response data includes: signal propagation time, signal amplitude, and signal phase;

[0041] Step S113: Obtain in-vehicle environment perception data based on channel impulse response data.

[0042] The aforementioned ultra-wideband (UWB) sensing chip is a hardware implementation of a wireless communication technology used for short-range, high-precision positioning and data transmission. UWB technology is characterized by transmitting data over an extremely wide spectrum; for example, UWB chips can operate in the frequency band between 3.1 GHz and 10.6 GHz. UWB chips can transmit and receive UWB signals, and due to their high resolution, low interference, strong penetration, and accurate distance measurement capabilities, they are suitable for indoor positioning and object detection. In the embodiments of this application, multiple UWB chips are used, possessing high-speed data transmission and high-precision ranging capabilities, enabling precise position tracking and motion detection of objects inside the vehicle.

[0043] The aforementioned ultra-wideband pulse signal is generated by a UWB chip and can cover a wide frequency range. Because UWB signals are transmitted at very low power over a wide frequency band, they can provide a clear signal response in complex environments, which is extremely important for identifying the movement of small objects, the presence of human bodies, and precise distance measurements between objects. In a vehicle interior environment, UWB pulse signals can penetrate non-metallic materials such as plastic and glass, achieving omnidirectional coverage within the vehicle and providing high-quality signal feedback even in low visibility or obstruction conditions.

[0044] Multiple ultra-wideband (UWB) sensing chips are arranged in pairs symmetrically along the longitudinal centerline of the target vehicle, i.e., the y0 direction. This aims to create a uniformly covered signal detection network, ensuring that all areas inside the vehicle can be effectively monitored. The symmetrical arrangement maximizes signal cross-coverage, reduces blind spots, and utilizes the symmetrical characteristics of UWB signals to optimize signal reception quality, thereby improving detection accuracy and reliability.

[0045] For example, the installation location of the UWB chip can be determined according to different vehicle models and internal structures. Figure 2 This is a schematic diagram of the deployment location of an ultra-wideband sensing chip according to an embodiment of this application, as shown below. Figure 2 As shown, for vehicles with two rows of seats, a UWB chip can be installed at position y0 in the front and rear roofs. Utilizing the penetrating and diffusion characteristics of UWB signals, it can capture signal changes in every corner of the vehicle. Even when the vehicle doors and windows are not completely closed or are blocked, it can maintain stable signal reception and environmental perception capabilities, thereby achieving real-time and accurate perception of changes in the vehicle's interior environment. Figure 3 This is a schematic diagram showing the deployment location of another ultra-wideband sensing chip according to an embodiment of this application, as shown below. Figure 3As shown, for a three-row seat vehicle, a UWB chip can be installed at position y0 on the roof of the front row, middle row, and rear row. The UWB chips are symmetrically arranged in the y0 direction of the vehicle, which can make full use of the symmetrical gain of the radar signal to ensure that the radar signal can fully cover the interior space of the vehicle and penetrate obstacles such as windows and doors, reducing blind spots.

[0046] Channel Impulse Response (CIR) data corresponding to ultra-wideband (UWB) pulse signals is collected. This data includes signal propagation time, signal amplitude, and signal phase. Specifically, after the UWB pulse signal propagates within the vehicle environment, it is received by the UWB chip. By measuring the time difference between transmission and reception, i.e., the signal propagation time, the length of the signal path can be calculated. Inside a closed or semi-closed vehicle, changes in the signal propagation path can reflect whether there are moving or deformed objects inside the vehicle. The presence of obstacles can cause changes in the signal diffraction or direct path, thus affecting the signal propagation time. Signal amplitude reflects the strength of the received signal. Inside the vehicle, obstruction or reflection by different objects or people will change the signal amplitude. When the in-vehicle environment changes, such as the approach of a person or the movement of an object, the amplitude information will show significant fluctuations. Analyzing the trend and statistical characteristics of signal amplitude changes can effectively identify whether new objects or people have appeared inside the vehicle, as well as the relative size and position of the objects. Signal phase is the phase angle on the signal waveform and is affected by the signal propagation path. The movement of objects or people causes changes in the signal phase, especially when the object is close to the UWB chip, the phase change is more significant. By continuously monitoring the signal phase, subtle motion information of objects, such as a person's breathing or micro-movements, can be captured, thereby enabling early warning of unauthorized intrusion or potential dangers inside the vehicle.

[0047] Furthermore, in-vehicle environment perception data is obtained based on the received channel impulse response data, that is, CIR data is converted into more intuitive and easy-to-understand in-vehicle environment perception data, which provides rich input features for subsequent target detection models, enabling intelligent perception of the in-vehicle environment, effectively distinguishing different types of in-vehicle behaviors, and improving the vehicle safety protection level.

[0048] Based on the above optional embodiments, by employing multiple ultra-wideband sensing chips to emit ultra-wideband pulse signals, and then collecting channel impulse response data corresponding to the ultra-wideband pulse signals, and finally obtaining in-vehicle environment perception data based on the channel impulse response data, comprehensive and high-precision perception of the in-vehicle environment is achieved. Even when the vehicle is not completely closed, the stability and accuracy of intrusion detection can be ensured, thereby further improving the reliability and real-time performance of the detection results.

[0049] In an optional embodiment, in step S12, feature extraction is performed on the in-vehicle environment perception data to obtain the feature extraction results, including:

[0050] Step S121: Preprocess the in-vehicle environment perception data to obtain the preprocessing result, wherein the preprocessing includes at least one of the following: noise reduction processing and normalization processing;

[0051] Step S122: Perform feature extraction on the preprocessing results to obtain the feature extraction results.

[0052] The preprocessing process aims to improve data quality and lay a solid foundation for subsequent feature extraction. Preprocessing includes, but is not limited to, denoising and normalization, thereby further improving the accuracy and reliability of the data. Specifically, in the in-vehicle environment, UWB signals are affected by various environmental noises, such as electromagnetic interference, temperature fluctuations, and humidity changes. Environmental noise interference reduces signal clarity and affects the accuracy of subsequent feature extraction. In this embodiment, wavelet transform, filters, and other methods can be used for denoising to effectively eliminate or reduce environmental noise, ensuring the purity of the channel impulse response data and providing a more reliable signal source for subsequent data analysis. Normalization is the process of adjusting signal amplitude, phase, and other feature values ​​to a uniform scale to eliminate the influence of differences in vehicle models, UWB chip locations, or signal strength. By converting the raw data into standardized data with the same range, normalization ensures that data from different vehicle types and time points can be compared on the same benchmark, improving the versatility and generalization ability of the target detection model.

[0053] After obtaining the preprocessing results, feature extraction is performed on them to obtain the feature extraction results. Feature extraction focuses on mining time-domain and frequency-domain features, including signal energy peaks, attenuation slopes, spectral entropy, and frequency band energy distribution. These features can more intuitively reflect the state changes and motion trajectories of objects inside the vehicle, providing rich input information for the training of the target detection model.

[0054] Based on the above optional embodiments, by preprocessing the in-vehicle environment perception data, a preprocessing result is obtained, and then feature extraction is performed on the preprocessing result to obtain the feature extraction result. This significantly enhances the data processing capability and feature recognition accuracy, laying a solid data foundation for subsequent model training and real-time detection. It can effectively reduce false alarms and false negatives, and improve the intelligence and reliability of the vehicle anti-theft intrusion detection system.

[0055] In an optional embodiment, the vehicle anti-theft detection method in this application further includes:

[0056] Obtain the training dataset, which includes: multiple sets of sample training data, and each set of sample training data includes: sample environment-aware data and sample annotation data;

[0057] The initial detection model is iteratively trained based on the training dataset to obtain the object detection model.

[0058] Specifically, the training dataset contains multiple sets of sample training data, each consisting of sample environment-aware data and sample annotation data. The sample environment-aware data can be channel impulse response data collected by the UWB module under various vehicle states and conditions, including signal propagation time, signal amplitude, and signal phase. The sample annotation data consists of manually labeled data based on the real vehicle environment states represented by the sample environment-aware data, such as labels for normal states and intrusion behaviors. The rich and diverse data in the sample training dataset is crucial for model training. By covering as many usage scenarios as possible, the trained object detection model can achieve better generalization capabilities.

[0059] Furthermore, the initial detection model is iteratively trained based on the training dataset to obtain the target detection model. Iterative training is a typical machine learning algorithm optimization method, which continuously updates the model parameters with the training dataset to gradually improve the model's prediction accuracy. In this embodiment, the initial detection model is first initially trained using a training dataset containing multiple sets of sample training data. The initial detection model can be, but is not limited to, a random forest, support vector machine, or lightweight convolutional neural network. In each training iteration, the initial detection model can adjust parameters such as weights and biases based on the errors fed back by the sample labeled data, with the aim of minimizing the gap between the predicted result and the actual label. As the number of iterations increases, the initial detection model gradually learns to extract key features from the channel impulse response data, eventually achieving the ability to accurately identify and classify the in-vehicle state, thus forming a high-performance target detection model.

[0060] Based on the above optional embodiments, by acquiring a training dataset and then iteratively training the initial detection model based on the training dataset, an object detection model can be obtained. This ensures that the object detection model has high accuracy and robustness when facing complex and ever-changing in-vehicle environments, thereby significantly improving the accuracy and processing speed of the object detection model in recognizing in-vehicle states.

[0061] In an optional embodiment, in step S13, the feature extraction results are analyzed and processed using a target detection model to obtain target prediction results, including:

[0062] In step S131, feature matching is performed based on the feature extraction results and the feature weight library of the target detection model to obtain the feature matching results;

[0063] In step S132, the target detection model is used to classify the feature matching results to obtain the initial prediction results;

[0064] In step S133, the initial prediction results are cross-validated to obtain the target prediction results.

[0065] In the feature matching stage, the current feature extraction results are compared with the feature weight library stored in the trained object detection model to find similar feature patterns, thereby drawing a preliminary judgment conclusion, i.e., the feature matching result. The feature weight library is learned during the model training phase based on historical sample data and annotation results, and can record the typical features corresponding to different types of intrusion behaviors (such as human movement and object displacement) and their importance. Feature matching can not only quickly identify the specific patterns of changes in the current in-vehicle environment, but also weight and score the matching results according to the importance of features, providing strong support for subsequent classification decisions.

[0066] Furthermore, the object detection model is used to classify the feature matching results to obtain an initial prediction. This classification decision involves the object detection model performing in-depth analysis of the feature matching results to determine whether the current in-vehicle state constitutes an intrusion. The object detection model can comprehensively consider all matching features and utilize the knowledge acquired during training to make a preliminary prediction of the current situation—the initial prediction. The object detection model, built using machine learning algorithms, can perform deep pattern recognition and analysis based on features such as signal energy peaks, attenuation slopes, spectral entropy, and frequency band energy distribution contained in the feature matching results. This accurately distinguishes between normal states and abnormal intrusions, providing real-time intelligent decision-making support for vehicle security.

[0067] Cross-validation is performed on the initial prediction results to obtain the target prediction results. Cross-validation is an additional verification step set up to ensure the reliability and accuracy of the model predictions. During the validation phase, the initial prediction results can be validated from multiple angles and levels to eliminate the possibility of random errors or false alarms. Cross-validation may include comparing the detection results of multiple UWB chips to analyze the spatial consistency of the signals; or introducing time series analysis to track the historical trends of vehicle state changes to ensure that the current prediction results are consistent with the vehicle's normal behavior patterns. When performing cross-validation, a spatial-temporal dual-dimensional validation can be introduced, i.e., spatial consistency verification through multiple UWB radar signals and temporal analysis combining historical state sequence analysis, significantly reducing the false alarm rate. Only when the initial prediction results pass strict validation conditions will a final, highly reliable target prediction result be generated, thereby avoiding false detections.

[0068] Figure 4This is a schematic diagram illustrating the training and application process of an object detection model according to an embodiment of this application, as shown below. Figure 4 As shown, the UWB chip is activated to collect CIR data of the in-vehicle environment in real time. The acquired CIR dataset is then split into training and testing datasets to ensure the initial detection model can be learned and validated using different datasets. During data preprocessing, techniques such as denoising and normalization are used to improve data quality, laying the foundation for subsequent feature extraction. In feature extraction, time-frequency domain features, including signal energy peaks, attenuation slopes, spectral entropy, and frequency band energy distribution, are extracted from the preprocessed data to construct multi-dimensional feature vectors. Machine learning algorithms, such as random forests, SVMs, or lightweight CNNs, are used to train the model on the training dataset, continuously fine-tuning the model parameters until a good classification effect is achieved. After training, the model is evaluated using the testing dataset to test its performance on unknown data. The final target detection model's parameters and feature weight library will be used for real-time detection and intrusion behavior classification. During model deployment, the in-vehicle environment perception data of the target vehicle is acquired, and then features are extracted from the in-vehicle environment perception data to obtain feature extraction results. Subsequently, the target detection model is used to analyze and process the feature extraction results to obtain target prediction results. Finally, based on the target prediction results, the corresponding vehicle safety policy for the target vehicle is generated and distributed. This achieves real-time, intelligent, and accurate in-vehicle environment monitoring, thereby providing immediate and effective vehicle safety policies and significantly improving the overall efficiency of the vehicle safety system and user experience.

[0069] Based on the above optional embodiments, feature matching is performed based on the feature extraction results and the feature weight library of the target detection model to obtain feature matching results. Then, the target detection model is used to classify the feature matching results to obtain initial prediction results. Finally, the initial prediction results are cross-validated to obtain target prediction results. This enhances the precision of data processing and model decision-making, ensuring high accuracy and intelligent response for vehicle anti-theft intrusion detection even in complex real-world scenarios, thereby significantly improving vehicle security and user experience.

[0070] In an optional embodiment, the vehicle anti-theft detection method in this application further includes:

[0071] Obtain vehicle status data for the target vehicle;

[0072] The anomaly detection threshold corresponding to the target detection model is determined based on vehicle state data, whereby the anomaly detection threshold is used to determine the prediction sensitivity of the target detection model.

[0073] The object detection model is used to classify the feature matching results, and the initial prediction results are as follows:

[0074] The object detection model and anomaly detection threshold are used to classify the feature matching results to obtain the initial prediction results.

[0075] The aforementioned vehicle status data covers a variety of status indicators closely related to vehicle safety, such as whether the doors are closed, whether the windows are open, and whether the vehicle is in motion. Through vehicle-to-everything (V2X) technology or the vehicle's own communication protocols, vehicle status data can be acquired in real time, providing real-time background information for subsequent dynamic model optimization and intrusion behavior detection.

[0076] Determining the anomaly detection threshold for the target detection model based on vehicle state data allows for flexible adjustment of the model's predictive sensitivity according to different vehicle states, adapting to more complex real-world environments. The anomaly detection threshold directly determines the target detection model's sensitivity to intrusion behavior, and vehicle state data provides the basis for dynamically adjusting the threshold. For example, when a car door is not fully closed or a window is open, the anomaly detection threshold can be increased to avoid frequent false alarms caused by external environmental factors; conversely, when the vehicle is locked, the threshold can be lowered to improve the ability to detect minor anomalies. This dynamic adjustment mechanism not only improves system response speed but also enhances the applicability and accuracy of the target detection model in different scenarios, reduces the false alarm rate, and improves the user experience.

[0077] Furthermore, by using the object detection model and the anomaly detection threshold together to classify the feature matching results, the feature matching results can be analyzed in depth, classified using the object detection model, and combined with the setting of the anomaly detection threshold, a preliminary judgment can be made on the changes in the current in-vehicle environment, generating initial prediction results.

[0078] Based on the above optional embodiments, by acquiring the vehicle status data of the target vehicle, and then determining the anomaly detection threshold corresponding to the target detection model based on the vehicle status data, the target detection model and the anomaly detection threshold are used to classify the feature matching results to obtain the initial prediction results. This can effectively identify potential intrusion behaviors, while reducing the risk of false alarms. It provides the ability to respond instantly and analyze intelligently for vehicle anti-theft security, and significantly enhances the practicality and user experience of the vehicle security protection system.

[0079] In an optional embodiment, step S14, generating a vehicle safety policy corresponding to the target vehicle based on the target prediction result includes:

[0080] The response determines that the target vehicle exhibits a preset type of behavior based on the target prediction results, sends a safety alert to the target user, and controls the target vehicle to perform safety warning operations.

[0081] Based on the target prediction results output by the model, it can intelligently determine whether preset abnormal behaviors, such as intrusion, have occurred inside the vehicle. By comparing the real-time target prediction results with predefined behavior types, it automatically identifies whether further security measures are needed. The criteria for determining preset behavior types, including but not limited to the frequency and intensity of features such as human movement and object displacement, can be derived by the target detection model through comprehensive analysis of feature matching results and anomaly detection thresholds.

[0082] Next, a security alert will be sent to the target user. When a pre-defined type of abnormal behavior is confirmed inside the target vehicle, a security alert will be immediately sent to the vehicle owner or a designated monitoring person via vehicle-to-everything (V2X) or mobile communication technology. This aims to notify the user as soon as possible so that they can react quickly to potential security threats. The security alert may include, but is not limited to, key information such as vehicle location, type of abnormal behavior, and time of occurrence, ensuring that the user is accurately aware of the current security status of the vehicle.

[0083] Simultaneously, upon detecting a preset type of behavior, the system controls the target vehicle to execute a series of preset safety warning operations, such as activating audible and visual alarms, locking the vehicle, and recording on-site images. By directly intervening in the vehicle safety protection system, this embodiment of the application can quickly take physical measures to prevent or reduce user losses, enhancing the initiative and immediacy of vehicle protection.

[0084] Based on the above optional embodiments, by responding to the determination of a preset type of behavior of the target vehicle based on the target prediction result, sending safety alarm information to the target user, and controlling the target vehicle to perform safety warning operations, the real-time response capability and user notification efficiency of the vehicle safety protection system are significantly improved.

[0085] In this embodiment, by symmetrically arranging UWB chips along the longitudinal central axis of the vehicle body, the penetrating characteristics of the broadband signals emitted by the UWB chips are utilized. Even if the windows or doors are not completely closed, blind-spot-free coverage of the in-vehicle environment can be achieved, effectively solving the problem of detection failure caused by physical obstacles in traditional sensors. The collaborative operation of multiple UWB chips can create a three-dimensional signal field in the in-vehicle space. Through signal superposition technology, the detection capability of weak targets is significantly enhanced, especially those small objects with inconspicuous reflectivity, thereby improving the detection sensitivity of the overall system.

[0086] By fully utilizing the time-frequency domain features of channel impulse response data through a target detection model, millisecond-level real-time detection is achieved on a low-power automotive chip. Compared to traditional camera monitoring solutions, the power consumption of this embodiment is effectively reduced, saving energy and optimizing the user experience, especially significantly reducing the battery burden when the vehicle is parked for extended periods. Furthermore, the target detection model supports incremental learning, flexibly adapting to differences in the internal structure of various vehicle models, demonstrating strong generalization capabilities. It is applicable to various types of vehicles without frequent model parameter adjustments, ensuring the broad applicability and long-term effectiveness of the anti-theft detection solution.

[0087] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.

[0088] According to an embodiment of this application, a vehicle anti-theft detection device is provided. It should be noted that the device can be used to perform the above-described vehicle anti-theft detection method.

[0089] Figure 5 This is a structural block diagram of a vehicle anti-theft detection device according to an embodiment of this application, such as... Figure 5 As shown, the device includes:

[0090] The acquisition module 501 is used to acquire in-vehicle environment perception data of the target vehicle, wherein the in-vehicle environment perception data is used to characterize the changes in the in-vehicle environment;

[0091] The extraction module 502 is used to extract features from the in-vehicle environment perception data and obtain feature extraction results, wherein the feature extraction results are used to represent the time domain features and frequency domain features corresponding to the in-vehicle environment perception data;

[0092] Analysis module 503 is used to analyze and process the feature extraction results using the target detection model to obtain the target prediction result. The target prediction result is used to determine whether the target vehicle has a preset type of behavior. The target detection model is obtained by machine learning training using multiple sets of sample training data.

[0093] The determination module 504 is used to generate and distribute the vehicle safety policy corresponding to the target vehicle based on the target prediction results.

[0094] Optionally, the acquisition module 501 is further configured to: transmit ultra-wideband pulse signals using multiple ultra-wideband sensing chips, wherein the multiple ultra-wideband sensing chips are arranged symmetrically in pairs along the longitudinal centerline of the target vehicle; collect channel impulse response data corresponding to the ultra-wideband pulse signals, wherein the channel impulse response data includes: signal propagation duration, signal amplitude, and signal phase; and acquire in-vehicle environmental perception data based on the channel impulse response data.

[0095] Optionally, the extraction module 502 is used to: preprocess the in-vehicle environment perception data to obtain a preprocessing result, wherein the preprocessing includes at least one of the following: noise reduction processing and normalization processing; and to extract features from the preprocessing result to obtain a feature extraction result.

[0096] Optionally, the acquisition module 501 is also used to acquire a training dataset, wherein the training dataset includes: multiple sets of sample training data, and each set of sample training data includes: sample environment perception data and sample annotation data; the vehicle anti-theft detection device also includes: a training module 505, used to iteratively train the initial detection model based on the training dataset to obtain a target detection model.

[0097] Optionally, the analysis module 503 is also used to: perform feature matching based on the feature extraction results and the feature weight library of the target detection model to obtain feature matching results; use the target detection model to make classification decisions on the feature matching results to obtain initial prediction results; and perform cross-validation on the initial prediction results to obtain target prediction results.

[0098] Optionally, the acquisition module 501 is further configured to acquire vehicle state data of the target vehicle; the determination module 504 is further configured to determine the anomaly detection threshold corresponding to the target detection model based on the vehicle state data, wherein the anomaly detection threshold is used to determine the prediction sensitivity of the target detection model; the analysis module 503 is further configured to: use the target detection model and the anomaly detection threshold to perform classification decision on the feature matching results to obtain the initial prediction result.

[0099] Optionally, the determining module 504 is also used to: respond to the determination based on the target prediction result that the target vehicle has a preset type of behavior, send a safety alarm message to the target user, and control the target vehicle to perform a safety warning operation.

[0100] Embodiments of this application also provide a vehicle, including: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods described in various embodiments of this application when it runs.

[0101] Embodiments of this application also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is running, it controls the device where the computer-readable storage medium is located to perform the methods of various embodiments of this application.

[0102] Embodiments of this application also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this application.

[0103] Embodiments of this application also provide a computer program product, including a non-volatile computer-readable storage medium for storing a computer program that, when executed by a processor, implements the methods in various embodiments of this application.

[0104] Embodiments of this application also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this application.

[0105] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0106] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0107] 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 units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0108] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0109] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or 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 this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0110] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A vehicle anti-theft detection method, characterized in that, include: Acquire in-vehicle environment perception data of the target vehicle, wherein the in-vehicle environment perception data is used to characterize changes in the in-vehicle environment; Feature extraction is performed on the in-vehicle environment perception data to obtain feature extraction results, wherein the feature extraction results are used to represent the time domain features and frequency domain features corresponding to the in-vehicle environment perception data; The feature extraction results are analyzed and processed using an object detection model to obtain object prediction results. The object prediction results are used to determine whether the target vehicle has a preset type of behavior. The object detection model is obtained by machine learning training using multiple sets of sample training data. Based on the target prediction results, a vehicle safety policy corresponding to the target vehicle is generated and issued.

2. The vehicle anti-theft detection method according to claim 1, characterized in that, Acquiring the in-vehicle environment perception data of the target vehicle includes: Multiple ultra-wideband sensing chips are used to emit ultra-wideband pulse signals, wherein the multiple ultra-wideband sensing chips are arranged in pairs symmetrically along the longitudinal centerline of the target vehicle; The channel impulse response data corresponding to the ultra-wideband pulse signal is collected, wherein the channel impulse response data includes: signal propagation duration, signal amplitude, and signal phase; The in-vehicle environment perception data is obtained based on the channel impulse response data.

3. The vehicle anti-theft detection method according to claim 1, characterized in that, Feature extraction is performed on the in-vehicle environment perception data to obtain the following feature extraction results: The in-vehicle environment perception data is preprocessed to obtain a preprocessing result, wherein the preprocessing includes at least one of the following: noise reduction processing and normalization processing; Feature extraction is performed on the preprocessing result to obtain the feature extraction result.

4. The vehicle anti-theft detection method according to claim 1, characterized in that, The vehicle anti-theft detection method also includes: Obtain a training dataset, wherein the training dataset includes: the multiple sets of sample training data, and each set of sample training data includes: sample environment awareness data and sample annotation data; The initial detection model is iteratively trained based on the training dataset to obtain the target detection model.

5. The vehicle anti-theft detection method according to claim 1, characterized in that, The target prediction result is obtained by analyzing and processing the feature extraction result using the target detection model, including: Based on the feature extraction results, feature matching is performed with the feature weight library of the target detection model to obtain feature matching results; The target detection model is used to classify the feature matching results to obtain an initial prediction result. The initial prediction results are cross-validated to obtain the target prediction results.

6. The vehicle anti-theft detection method according to claim 5, characterized in that, The vehicle anti-theft detection method also includes: Obtain the vehicle status data of the target vehicle; Based on the vehicle state data, an anomaly detection threshold corresponding to the target detection model is determined, wherein the anomaly detection threshold is used to determine the prediction sensitivity of the target detection model; The initial prediction result is obtained by classifying the feature matching results using the target detection model, including: The target detection model and the anomaly detection threshold are used to classify the feature matching results to obtain the initial prediction result.

7. The vehicle anti-theft detection method according to claim 1, characterized in that, The vehicle safety strategy generated based on the target prediction result includes: The response determines, based on the target prediction result, that the target vehicle exhibits the preset type of behavior, sends a safety alert to the target user, and controls the target vehicle to perform a safety warning operation.

8. A vehicle, characterized in that, include: Memory, which stores executable programs; A processor for running the program, wherein the program, when running, performs the method according to any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored executable program, wherein, when the executable program is executed, it controls the device on which the storage medium is located to perform the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 7.

Images