Auxiliary driving vehicle picture dragon behavior recognition method based on multi-modal time sequence feature analysis

Abstract

This invention relates to the field of intelligent driving vehicle technology, specifically disclosing a method for recognizing "dragging" behavior in assisted driving vehicles based on multimodal temporal feature analysis. First, vehicle driving data is collected and processed to calculate the temporal signal of the vehicle's lateral distance relative to the lane centerline. Second, the active state of the assisted driving system is verified. Then, the peaks and troughs of the lateral distance signal are detected collaboratively, and the extreme points are intelligently paired and optimized using temporal alternation constraints and extreme value competition rules to generate an extreme point sequence that conforms to the laws of physical motion. Finally, a joint judgment is made using multi-dimensional criteria: when the optimized extreme point sequence simultaneously satisfies the conditions of periodicity, amplitude validity, and persistence, "dragging" behavior is determined to have occurred, and quantitative parameters and severity levels can be output. This invention achieves high-precision, quantifiable detection of "dragging" behavior, providing data support for the performance evaluation and optimization of assisted driving systems.

Description

Technical Field

[0001] This invention belongs to the field of intelligent driving vehicle technology, specifically relating to a method for recognizing the dragon drawing behavior of assisted driving vehicles based on multimodal temporal feature analysis. Background Technology

[0002] With the widespread adoption of advanced driver assistance systems (ADAS), the lateral control performance of vehicles during automatic lane keeping has become a key indicator for evaluating system maturity. "Dragon-drawing" behavior, characterized by continuous, periodic left-right swaying within the lane, is a typical manifestation of imperfect control strategies and poor system stability. Therefore, there is an urgent need for a method that can accurately and reliably identify dragon-drawing behavior and quantify its characteristics for use in system development optimization, performance evaluation, and fault diagnosis.

[0003] In existing technologies, the detection of "dragon drawing" behavior mostly relies on single-point threshold judgments of simple lateral acceleration or lateral offset, which has the following drawbacks:

[0004] (1) It is impossible to effectively distinguish between normal minor driver corrections and abnormal, continuous oscillations;

[0005] (2) It is susceptible to interference from environmental factors such as road curvature and sensor noise, resulting in a high false alarm rate;

[0006] (3) The lack of comprehensive quantitative analysis of the swing amplitude, frequency and duration makes it impossible to accurately classify the severity of the dragon-drawing behavior.

[0007] Therefore, it is necessary to design a method for monitoring and evaluating the lateral stability of L2 and above driver assistance systems during lane keeping. Specifically, it is a method for identifying "dragon-drawing" behavior based on the temporal waveform characteristics of vehicle lateral motion, in order to solve the technical problems of the existing driver assistance systems having simple dragon-drawing behavior identification methods and high false alarm rates. Summary of the Invention

[0008] The purpose of this invention is to provide a method for recognizing the "dragon-drawing" behavior of assisted driving vehicles based on multimodal temporal feature analysis. By analyzing the temporal data of the lateral distance of the vehicle relative to the center line of the lane, key waveform features are extracted, and multi-level criteria are adopted to achieve high accuracy and low false alarm rate recognition and evaluation of the "dragon-drawing" behavior.

[0009] To address the aforementioned problems in the existing technology, the technical solution adopted in this invention is: a method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis, comprising the following steps:

[0010] S1. Data preprocessing: Multi-source data fusion is performed, and lane distance information, lateral offset information and driver assistance system status information provided by the vehicle are collected simultaneously. The validity of the collected data is verified and outlier is filtered.

[0011] S2. Timing window analysis and state verification: Timing window analysis uses a sliding window segmentation method; state verification is driving state verification, which determines that the analysis object is system behavior rather than human driving behavior.

[0012] S3. Waveform feature extraction and optimization: A dual-modal extreme point detection method is adopted. Within a single analysis window, peak and trough detection algorithms are used to identify local maxima and local minima in the vehicle lateral distance sequence, and then extreme point matching is performed.

[0013] S4. Multi-dimensional dragon drawing behavior determination: including periodicity determination, amplitude determination and persistence determination, followed by feature quantification and classification.

[0014] Preferably, the lateral offset information in step S1 is calculated based on the distance information between the left and right lane lines, determining the real-time lateral distance of the vehicle relative to the physical center line of the lane.

[0015] Preferably, the time-series signal is denoised using a sliding window smoothing algorithm based on polynomial fitting in step S1, which preserves effective fluctuation characteristics while suppressing high-frequency noise.

[0016] Preferably, in step S2, the sliding window segmentation method is to divide the continuous time series data into analysis windows of fixed duration, with the optimal fixed duration being 30 seconds, and to slide with a fixed step size, with the optimal fixed step size being 5 seconds, for coverage analysis of the entire process.

[0017] Preferably, in step S2, the driving status verification is performed by checking whether the advanced driver assistance function is in an effective active state during the time period before analyzing each window. The verification is based on the fact that the lane centering state is enabled and the driver is not in a lane changing or yielding state.

[0018] Preferably, in step S3, the local maximum point is a wave peak, representing the extreme position of the swing to one side of the lane; the local minimum point is a wave trough, representing the extreme position of the swing to the other side of the lane; the dual-modal extreme point detection process introduces minimum amplitude significance and minimum time interval constraints to eliminate invalid fluctuations, wherein the amplitude significance is ≥0.1 meters and the time interval is ≥2 seconds.

[0019] Preferably, in step S3, extreme point matching involves sorting all detected peaks and troughs in chronological order and applying a "temporal alternation constraint" rule to ensure that extreme points strictly follow the physical law of alternating between "peak-trough-peak-trough". For extreme points of the same type that are too close in time, such as two consecutive peaks, an "extreme point competition" mechanism is adopted to retain the more significant one, thereby optimizing the feature point sequence.

[0020] Preferably, in step S4, the periodic determination is that the optimized extreme point sequence contains at least a specified number of peaks and troughs to ensure that the vehicle completes a sufficient number of complete oscillation cycles.

[0021] The amplitude is determined by calculating half of the distance difference between adjacent wave crests and troughs as the half amplitude value of a single swing. The half amplitude value must fall within the preset effective range, which is 0.15 meters to 0.6 meters.

[0022] The continuity criterion is that the effective oscillation is continuous in time, that is, the number of effective extreme points must exceed the threshold.

[0023] Preferably, in step S4, feature quantification and grading are performed as follows: for the dragon-drawing behavior segments that pass the judgment, the average swing amplitude and average swing period are calculated, and the average swing period is used to evaluate the severity of the "dragon-drawing" behavior; and based on the average swing amplitude, the severity of the dragon-drawing behavior is graded, including: mild: 0.15m-0.3m, moderate: 0.3m-0.45m, and severe: more than 0.45m.

[0024] The beneficial effects of this invention are as follows:

[0025] The multimodal temporal feature analysis-based method for recognizing the "dragon-drawing" behavior of assisted driving vehicles designed in this invention has high accuracy. It abandons the simple single-point threshold judgment and greatly reduces false alarms and false negatives by recognizing the essential feature of "periodic oscillation" and combining it with multi-dimensional criteria such as amplitude, period, and duration for joint verification.

[0026] The multimodal temporal feature analysis-based method for recognizing the dragon-drawing behavior of assisted driving vehicles designed in this invention has strong robustness. Through strategies such as data smoothing, extreme point saliency constraints, and intelligent pairing, it effectively suppresses interference caused by sensor noise and changes in the road environment, thereby improving the adaptability of the method.

[0027] The multimodal temporal feature analysis method for recognizing the "dragon-drawing" behavior in assisted driving vehicles designed in this invention has quantifiable output features. It not only provides a binary judgment of whether the "dragon-drawing" behavior has occurred, but also outputs quantitative indicators such as the amplitude, period, and duration of the swing, and classifies the severity, providing accurate data support for in-depth evaluation and optimization of system performance. Attached Figure Description

[0028] Figure 1 The flowchart shows a method for recognizing the dragon drawing behavior of assisted driving vehicles based on multimodal temporal feature analysis. Detailed Implementation

[0029] The present invention will be further described below with reference to the accompanying drawings and reference numerals.

[0030] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0031] The terms “first,” “second,” “third,” etc., are used only to distinguish descriptions and should not be interpreted as indicating or implying relative importance.

[0032] In the description of this invention, it should also be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0033] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0034] like Figure 1 As shown, a method for recognizing "dragon-drawing" behavior in assisted driving vehicles based on multimodal temporal feature analysis treats the lateral movement trajectory of the vehicle within the lane as a temporal signal, and identifies the behavior by detecting specific waveform patterns in this signal that conform to the characteristics of "dragon-drawing" behavior. This method emphasizes the holistic, temporal, and multi-dimensional verification of features.

[0035] 1. Data Preprocessing and Feature Engineering

[0036] Multi-source data fusion: Simultaneously collect lane distance information, lateral offset information, and driver assistance system status information provided by the vehicle.

[0037] Key indicator calculation: Based on the distance between the left and right lane lines, calculate the real-time lateral distance of the vehicle relative to the physical center line of the lane, which serves as the core analysis indicator.

[0038] Data quality assurance: The original data is validated and outliers are filtered. A sliding window smoothing algorithm based on polynomial fitting is used to denoise the time series signal, preserving effective fluctuation characteristics while suppressing high-frequency noise.

[0039] 2. Timing window analysis and state verification

[0040] Sliding window segmentation: Continuous time series data is divided into analysis windows of fixed duration (30 seconds), and the windows slide in fixed steps (5 seconds) to achieve coverage analysis of the entire process.

[0041] Driving status verification: Before analyzing each window, verify whether the advanced driver assistance functions are effectively activated during that time period (lane centering is enabled and not in lane changing or avoidance mode) to ensure that the analysis object is system behavior rather than human driving behavior.

[0042] 3. Waveform Feature Extraction and Optimization

[0043] Dual-modal extreme point detection: Within a single analysis window, peak and trough detection algorithms are used to identify local maxima (peaks, representing the extreme position of swinging to one side of the lane) and local minima (troughs, representing the extreme position of swinging to the other side) in the vehicle's lateral distance sequence. The detection process incorporates minimum amplitude significance and minimum time interval constraints (amplitude significance ≥ 0.1 meters, time interval ≥ 2 seconds) to eliminate invalid fluctuations.

[0044] Intelligent extreme point pairing: All detected peaks and troughs are sorted chronologically and a "temporal alternation constraint" rule is applied to ensure that extreme points strictly follow the physical law of alternating between "peak-trough-peak-trough". For extreme points of the same type that are too close in time (such as two consecutive peaks), an "extreme point competition" mechanism is used to retain the more significant one, thereby optimizing the feature point sequence.

[0045] 4. Multi-dimensional judgment of dragon drawing behavior

[0046] Periodicity determination: The optimized extreme point sequence must contain at least a specified number of peaks and troughs to ensure that the vehicle completes a sufficient number of full swing cycles.

[0047] Amplitude determination: Half of the distance difference between adjacent wave crests and troughs is calculated as the half-amplitude value of a single oscillation. This amplitude must fall within a preset effective range (0.15 meters to 0.6 meters) to exclude minor shaking and abnormally large turns.

[0048] Continuity determination: Effective oscillations must be continuous in time, meaning the number of effective extreme points must exceed a threshold.

[0049] Feature quantification and grading: For the dragon-drawing behavior segments that pass the judgment, calculate their average swing amplitude and average swing period. The average swing period can be used to evaluate the severity of the "dragon-drawing" behavior. Based on the average swing amplitude, the severity of the dragon-drawing behavior is graded (mild: 0.15m-0.3m, moderate: 0.3m-0.45m, severe: more than 0.45m).

[0050] This invention treats the lateral movement of a vehicle as a time-series signal and employs a complete and logically rigorous waveform feature extraction, optimization, and multi-dimensional judgment system to identify the "dragon drawing" behavior, which lies in the overall architecture and strategy process of the methodology.

[0051] (1) Preprocessing mechanism based on sliding window and state verification: The continuous journey is divided into fixed duration windows, and the advanced driver assistance system is verified to be active within the window to ensure the correctness of the analysis object.

[0052] (2) "Crest-Valley" Dual-Mode Feature Collaborative Detection and Optimization Strategy: Within a single analysis window, the peaks and troughs of the lateral distance signal are detected simultaneously, and the "temporal alternation constraint" and "extreme value competition" rules are applied to intelligently filter and pair the detected feature points to generate an optimized extreme value sequence that conforms to the laws of physical motion.

[0053] (3) Multi-dimensional and multi-condition joint judgment logic: The judgment of the "dragon drawing" behavior is not based on a single indicator, but must meet three core conditions at the same time: Periodicity condition: There are at least two complete swing cycles (i.e., the number of peaks and troughs meets the standard). Amplitude condition: The swing amplitude is within a reasonable physical range (excluding both noise and emergency obstacle avoidance). Continuity condition: The effective swing is continuous in time.

[0054] (4) Quantification and grading output of “dragon drawing” behavior: not only output the binary judgment of whether “dragon drawing” behavior exists, but more importantly, output the quantitative characteristic parameters (such as average amplitude and average period) and the severity grading based on amplitude.

[0055] This invention is not limited to the above-described optional embodiments. Anyone can derive other various forms of products under the guidance of this invention. However, regardless of any changes made in their shape or structure, any technical solution that falls within the scope of the claims of this invention shall be protected by this invention.

Claims

1. A method for recognizing the "dragon-drawing" behavior of assisted driving vehicles based on multimodal temporal feature analysis, characterized in that, Includes the following steps: S1. Data preprocessing: Multi-source data fusion is performed, and lane distance information, lateral offset information and driver assistance system status information provided by the vehicle are collected simultaneously. The validity of the collected data is verified and outlier is filtered. S2. Timing window analysis and state verification: Timing window analysis uses a sliding window segmentation method; state verification is driving state verification, which determines that the analysis object is system behavior rather than human driving behavior. S3. Waveform feature extraction and optimization: A dual-modal extreme point detection method is adopted. Within a single analysis window, peak and trough detection algorithms are used to identify local maxima and local minima in the vehicle lateral distance sequence, and then extreme point matching is performed. S4. Multi-dimensional dragon drawing behavior determination: including periodicity determination, amplitude determination and persistence determination, followed by feature quantification and classification.

2. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 1, characterized in that, In step S1, the lateral offset information is calculated based on the distance information between the left and right lane lines, determining the real-time lateral distance of the vehicle relative to the physical center line of the lane.

3. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 1, characterized in that, The data collected in step S1 is processed by a sliding window smoothing algorithm based on polynomial fitting to denoise the time series signal, preserving effective fluctuation characteristics while suppressing high-frequency noise.

4. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 1, characterized in that, In step S2, the sliding window segmentation method is to divide the continuous time series data into analysis windows of fixed duration, with the optimal fixed duration being 30 seconds, and to slide with a fixed step size, with the optimal fixed step size being 5 seconds, for coverage analysis of the entire process.

5. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 1, characterized in that, In step S2, the driving status verification involves checking whether the advanced driver assistance function is effectively activated within the time period before analyzing each window. The verification criteria are that the lane centering state is enabled and the driver is not in a lane changing or yielding state.

6. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 1, characterized in that, In step S3, the local maximum point is the peak, representing the extreme position of the swing to one side of the lane; the local minimum point is the trough, representing the extreme position of the swing to the other side of the lane. The dual-modal extreme point detection process introduces minimum amplitude significance and minimum time interval constraints to eliminate invalid fluctuations, wherein the amplitude significance is ≥0.1 meters and the time interval is ≥2 seconds.

7. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 6, characterized in that, In step S3, extreme point matching involves sorting all detected peaks and troughs in chronological order and applying a "temporal alternation constraint" rule to ensure that extreme points strictly follow the physical law of alternating between "peak-trough-peak-trough". For extreme points of the same type that are too close in time, such as two consecutive peaks, an "extreme point competition" mechanism is adopted to retain the more significant one, thereby optimizing the feature point sequence.

8. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 7, characterized in that, In step S4, the periodic determination is that the optimized extreme point sequence contains at least a specified number of peaks and troughs to ensure that the vehicle completes a sufficient number of complete swing cycles. The amplitude is determined by calculating half of the distance difference between adjacent wave crests and troughs as the half amplitude value of a single swing. The half amplitude value must fall within the preset effective range, which is 0.15 meters to 0.6 meters. The continuity criterion is that the effective oscillation is continuous in time, that is, the number of effective extreme points must exceed the threshold.

9. The method for recognizing the "dragon-drawing behavior" of assisted driving vehicles based on multimodal temporal feature analysis according to claim 8, characterized in that, In step S4, feature quantification and grading are performed as follows: For the dragon-drawing behavior segments that pass the judgment, the average swing amplitude and average swing period are calculated. The average swing period is used to evaluate the intensity of the "dragon-drawing" behavior. The severity of the dragon drawing was graded based on the average swing amplitude, including: mild: 0.15m-0.3m, moderate: 0.3m-0.45m, and severe: more than 0.45m.

Images