Vehicle door control method and system based on gesture recognition, and vehicle
By collecting point cloud data through vehicle-mounted radar sensors and performing time-series change analysis, the type of hand gesture is identified, solving the problem of low accuracy of gesture recognition in adverse weather conditions and achieving high precision and intelligence in vehicle door control.
Patent Information
- Application Number
- PCT/CN2026/087814
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-10-18
- Filing Date
- 2026-03-31
- Publication Date
- 2026-11-12
AI Technical Summary
Existing gesture recognition-based door control solutions suffer from decreased recognition accuracy and high false trigger rate in adverse weather conditions such as low light, strong light reflection, or rain, snow, fog, and haze, making them unable to flexibly adapt to the convenient door opening needs of different user scenarios.
Point cloud data is collected using vehicle-mounted radar sensors to construct a point cloud data queue. By performing time-series change analysis on the point cloud data queue, speed, distance, and angle change features are extracted to identify gesture action types, and the door opening method is controlled according to the action type.
Achieving high-precision gesture recognition in complex environments avoids accidental triggering, improves the intelligence level of door control and user experience, and ensures the accuracy of door opening.
Smart Images

Figure CN2026087814_12112026_PF_FP_ABST
Abstract
Description
Gesture recognition-based door control methods, systems, and vehicles
[0001] Cross-reference
[0002] This disclosure claims priority to Chinese Patent Application No. 202511491456.7, filed on October 18, 2025, entitled "Vehicle door control method, system and vehicle based on gesture recognition", the entire contents of which are incorporated herein by reference. Technical Field
[0003] This disclosure relates to the field of intelligent vehicle control technology, and more specifically, to a door control method, system, and vehicle based on gesture recognition. Background Technology
[0004] With the development of automotive intelligence, automatic door opening and closing technology has gradually become an important direction for improving user experience and safety. Automatic door opening and closing technology achieves automatic opening and closing of car doors through the coordinated work of sensors, actuators, and control systems. Existing automatic door opening and closing technologies mainly rely on gesture recognition to achieve door opening and closing. However, existing gesture recognition-based door control solutions cannot flexibly adapt to the convenient door opening needs of different user scenarios. Current technologies using surround-view cameras for gesture recognition suffer from poor environmental adaptability. In low light, strong light reflection, or adverse weather conditions such as rain, snow, fog, and haze, the recognition accuracy of surround-view cameras decreases significantly, and they are easily interfered with by the environment or other gestures, resulting in a high false trigger rate.
[0005] There is currently no good solution to the above problems. Summary of the Invention
[0006] This disclosure provides a gesture recognition-based door control method, system, and vehicle to at least solve the technical problem of low gesture recognition accuracy leading to poor door control accuracy in related technologies.
[0007] According to one aspect of the present disclosure, a vehicle door control method based on gesture recognition is provided. The method includes: responding to a closed vehicle door and the presence of a target object in a preset spatial area; acquiring point cloud data collected by an onboard radar sensor within a preset time period, wherein the point cloud data is motion feature data of the target object; constructing a point cloud data queue using the point cloud data; performing temporal change analysis on the point cloud data in the point cloud data queue to obtain temporal change features of the target object's hand movements; determining the gesture movement type of the target object based on the temporal change features; determining the vehicle door opening method corresponding to the gesture movement type; and controlling the opening of the vehicle door according to the vehicle door opening method.
[0008] Optionally, temporal change analysis is performed on the point cloud data in the point cloud data queue to obtain the temporal change characteristics of the target object's hand movements, including: performing temporal change characteristic analysis on the point cloud data in the point cloud data queue to determine the velocity change characteristics; performing temporal change characteristic analysis on the point cloud data in the point cloud data queue to determine the distance change characteristics; performing temporal change characteristic analysis on the point cloud data in the point cloud data queue to determine the angle change characteristics; and determining the temporal change characteristics based on the velocity change characteristics, distance change characteristics, and angle change characteristics.
[0009] Optionally, velocity temporal variation characteristic analysis is performed on the point cloud data in the point cloud data queue to determine the velocity variation characteristics, including: drawing a velocity temporal variation curve based on the point cloud data queue; determining the number of velocity peaks, the number of velocity troughs, and their relative positions based on the velocity temporal variation curve, wherein the relative position is the position of the velocity peak relative to the velocity trough; and determining the velocity variation characteristics based on the number of velocity peaks, the number of velocity troughs, and their relative positions.
[0010] Optionally, distance temporal variation feature analysis is performed on the point cloud data in the point cloud data queue to determine the distance variation features, including: drawing a distance temporal variation curve based on the point cloud data queue; and determining the distance variation features based on the distance temporal variation curve.
[0011] Optionally, the angle time-series change characteristics of the point cloud data in the point cloud data queue are analyzed to determine the angle change characteristics, including: plotting the pitch angle time-series change curve and the horizontal angle time-series change curve based on the point cloud data queue; and determining the angle change characteristics based on the pitch angle time-series change curve and the horizontal angle time-series change curve.
[0012] Optionally, the gesture recognition-based door control method further includes: determining the velocity change and distance change corresponding to two adjacent point cloud data points in the point cloud data queue; determining the number of first point cloud data points and the number of second point cloud data points in the point cloud data queue, wherein the pitch angle corresponding to the first point cloud data point is greater than a preset pitch angle threshold, and the horizontal angle corresponding to the second point cloud data point is greater than a preset horizontal angle threshold; in response to the velocity change corresponding to any two adjacent point cloud data points being less than or equal to a preset velocity change threshold, the distance change corresponding to any two adjacent point cloud data points being less than or equal to a preset distance change threshold, the number of first point cloud data points satisfying a first quantity condition, and the number of second point cloud data points satisfying a second quantity condition, recognizing the determined gesture action type as a valid action, and controlling the door to open according to the door opening method corresponding to the gesture action type.
[0013] Optionally, the gesture recognition-based door control method further includes: in response to any speed change corresponding to two adjacent point cloud data points being greater than a preset speed change threshold, or any distance change corresponding to two adjacent point cloud data points being greater than a preset distance change threshold, or the number of the first point cloud data points not meeting a first quantity condition, or the number of the second point cloud data points not meeting a second quantity condition, identifying the determined gesture action type as a mis-touch action, and controlling the door to remain closed.
[0014] Optionally, a point cloud data queue is constructed using point cloud data, including: for each frame of point cloud data included in the point cloud data, obtaining the velocity corresponding to multiple point cloud data points in each frame of point cloud data, and filtering out target point cloud data points, wherein the target point cloud data point is the point cloud data point with the maximum velocity in the current frame of point cloud data; and constructing a point cloud data queue based on multiple target point cloud data points.
[0015] According to another aspect of the embodiments of this disclosure, a door control system based on gesture recognition is also provided, including: a gesture detection module and a door control module; wherein, the gesture detection module is communicatively connected to the door control module and is configured to determine the gesture action type of the target object and the door opening method corresponding to the gesture action type according to the door control method based on gesture recognition in any of the above claims; the door control module includes a door controller and is configured to control the opening of the door according to the door opening method.
[0016] According to another aspect of the present disclosure, a vehicle is also provided, including the above-described gesture recognition-based door control system.
[0017] According to another aspect of the embodiments of the present disclosure, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the computer-readable storage medium is located to perform the methods of the various embodiments of the present disclosure.
[0018] According to another aspect of the embodiments of this disclosure, 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 disclosure.
[0019] According to another aspect of the embodiments of this disclosure, 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 disclosure.
[0020] According to another aspect of the embodiments of this disclosure, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of this disclosure.
[0021] This disclosure provides a vehicle door control method based on gesture recognition, comprising: responding to a closed vehicle door and the presence of a target object in a preset space area; acquiring point cloud data collected by an onboard radar sensor within a preset time period, wherein the point cloud data is motion feature data of the target object; constructing a point cloud data queue using the point cloud data; performing temporal change analysis on the point cloud data in the point cloud data queue to obtain the temporal change features of the target object's hand movements; determining the gesture movement type of the target object based on the temporal change features; determining the door opening method corresponding to the gesture movement type; and controlling the door opening according to the door opening method. This disclosure firstly, by responding to a closed vehicle door and monitoring a target object within a preset space area, enables timely and accurate initiation of the gesture recognition process, avoiding invalid or meaningless data processing and ensuring effective utilization of system resources. Secondly, the construction of a point cloud data queue using the point cloud data collected by the onboard radar sensor within a preset time period not only collects the raw data of the target object's hand movements but also preserves the continuity and dynamism of the movements through a time series approach, laying the foundation for subsequent accurate analysis of hand movement features. Furthermore, by performing time-series variation analysis on the point cloud data queue, the time-series variation features of multiple key parameters are extracted. The extraction and matching of these features can significantly distinguish different hand gestures, achieving high-precision gesture recognition even in complex environments or backgrounds. Then, based on the time-series variation features, the gesture type is determined, enabling intelligent recognition and classification of numerous gestures, including but not limited to grasping, waving, and drawing circles, thereby achieving a response to multi-mode gesture commands. Finally, based on the determined gesture type, the corresponding door opening method can be accurately matched, avoiding door control errors caused by gesture recognition mistakes, directly improving the intelligence level of door control and user experience. In summary, this disclosure achieves the technical effect of improving gesture recognition accuracy and thus enabling precise door control, thereby solving the technical problem of low gesture recognition accuracy leading to poor door control accuracy in related technologies. Attached Figure Description
[0022] The accompanying drawings, which are included to provide a further understanding of this disclosure and form part of this disclosure, illustrate exemplary embodiments of the present disclosure and are used to explain the disclosure, but do not constitute an undue limitation of the disclosure. In the drawings:
[0023] Figure 1 is a flowchart of a door control method based on gesture recognition according to an embodiment of the present disclosure;
[0024] Figure 2 is an example diagram of the speed timing variation curve according to an embodiment of the present disclosure;
[0025] Figure 3 is an example diagram of the distance time-series variation curve according to an embodiment of the present disclosure;
[0026] Figure 4 is an example diagram of the pitch angle timing variation curve according to an embodiment of the present disclosure;
[0027] Figure 5 is an example diagram of the time-series variation curve of the horizontal angle according to an embodiment of the present disclosure;
[0028] Figure 6 is a flowchart illustrating a gesture recognition method according to an embodiment of the present disclosure;
[0029] Figure 7 is a flowchart illustrating the determination of the validity of gesture recognition results according to an embodiment of the present disclosure;
[0030] Figure 8 is a structural block diagram of a gesture recognition-based door control system according to an embodiment of the present disclosure. Detailed Implementation
[0031] To enable those skilled in the art to better understand the present disclosure, the technical solutions of the present disclosure will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present disclosure, and not all embodiments. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present disclosure.
[0032] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this disclosure 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 disclosure 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.
[0033] According to an embodiment of the present invention, an embodiment of a door control method based on gesture recognition 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. Furthermore, 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.
[0034] This invention provides a vehicle door control method based on gesture recognition. Figure 1 is a flowchart of a vehicle door control method based on gesture recognition according to an embodiment of this disclosure. As shown in Figure 1, the vehicle door control method based on gesture recognition includes the following steps.
[0035] Step S11: In response to the door being closed and a target object existing in a preset space area, acquire point cloud data collected by the vehicle radar sensor within a preset time period, wherein the point cloud data is the motion feature data of the target object.
[0036] Step S12: Construct a point cloud data queue using the point cloud data.
[0037] Step S13: Perform temporal change analysis on the point cloud data in the point cloud data queue to obtain the temporal change characteristics of the target object's hand movements.
[0038] Step S14: Determine the type of gesture action of the target object based on the temporal change characteristics.
[0039] Step S15: Determine the door opening method corresponding to the gesture action type, and control the door opening according to the door opening method.
[0040] Optionally, the aforementioned vehicle-mounted radar is a millimeter-wave radar. The millimeter-wave radar operates in the 76GHz-81GHz frequency band, and the modulation waveform is a continuous wave whose frequency varies with time, with a data period of 10ms-100ms.
[0041] The aforementioned preset space area is a pre-defined effective interaction area, and is set to limit the effective area for gesture recognition. In an optional embodiment, the preset space area is the area near the door handle or a certain area on the side of the vehicle.
[0042] Vehicle-mounted radar sensors, with their high-precision positioning capabilities and non-contact characteristics, can accurately detect and track target objects within a preset spatial area.
[0043] Optionally, the status of the car door is monitored in real time. Once a signal indicating that the door is closed is detected, monitoring of a preset space area begins. When the target's hand enters the preset space area and is captured by the radar sensor, the point cloud data collection process is immediately initiated, laying the foundation for the next step of gesture recognition.
[0044] The aforementioned point cloud data refers to the set of points in the environmental information captured by vehicle-mounted radar sensors, with each point having its specific coordinate information.
[0045] The coordinate information of each point in a preset spatial area within a preset time period is collected by the vehicle-mounted radar sensor. The coordinate information includes distance, horizontal angle and pitch angle, and also includes velocity information to reflect the movement of the target. The above coordinate information constitutes point cloud data.
[0046] The aforementioned point cloud data refers to the set of points in the environmental information captured by vehicle-mounted radar sensors, with each point having specific coordinate information.
[0047] The construction of a point cloud data queue is essentially to arrange point cloud data that meet preset conditions in a chronological order from multiple consecutive frames of point cloud data, so as to provide a data sequence within a continuous time window for subsequent temporal change analysis.
[0048] Optionally, the vehicle-mounted millimeter-wave radar sensor continuously acquires point cloud data of the target object's hand within a preset time period, and constructs a point cloud data queue based on the aforementioned point cloud data. The length of the point cloud data queue typically covers about one second of data, which contains sufficient information without being too lengthy, facilitating rapid processing and response.
[0049] The aforementioned temporal change analysis refers to the dynamic parsing of continuous point cloud data contained in the point cloud data queue. The purpose of temporal change analysis is to capture the movement trajectory and movement pattern of the target object's hand. The temporal change analysis process involves the extraction and comparison of key parameters.
[0050] Optionally, specific hand gesture features can be extracted by comparing and analyzing the changes in parameters such as velocity, distance (between points), horizontal angle, and pitch angle of each point cloud data point in the point cloud data queue over time. For example, waving hands up and down can be identified by the periodic changes in the velocity and pitch angle of the point cloud data points, while pressing actions are characterized by the single peak change in the velocity of the point cloud data points and the rapid decrease in distance.
[0051] Optionally, the temporal change features obtained through temporal change analysis are compared and matched with various pre-stored gesture templates to identify the specific gesture type performed by the target object. For example, it can distinguish between various gesture types such as "grab," "wave," and "press," which provides a clear instruction basis for subsequent door control.
[0052] Optionally, each gesture template is extracted by a trained model based on a large amount of sample data, representing a typical set of features for a specific gesture.
[0053] The aforementioned door opening methods refer to door opening and closing mechanisms triggered by user gesture commands, including various opening and closing methods such as push opening and scissor opening.
[0054] The aforementioned door opening control is based on a defined door opening method, driving the vehicle's electric motor or hydraulic system to perform the door opening operation.
[0055] Optionally, depending on the type of gesture, the corresponding door opening method (such as side-opening door, side-sliding door or other customized opening mode) can be selected, and then a command can be sent to the door control system to achieve the specific door action.
[0056] For example, the "waving up and down" gesture will open the sliding door. The "pull" gesture will open the side-opening door.
[0057] In this embodiment of the disclosure, a door control method based on gesture recognition is provided, including the following execution steps.
[0058] In response to a closed vehicle door and the presence of a target object in a preset spatial area, point cloud data collected by the vehicle-mounted radar sensor within a preset time period is acquired. This point cloud data represents the motion characteristic data of the target object. A point cloud data queue is constructed using this data. Temporal change analysis is performed on the point cloud data in the queue to obtain the temporal change characteristics of the target object's hand movements. Based on these temporal change characteristics, the type of the target object's gesture is determined. The corresponding door opening method is then determined, and the door is opened according to this method. This disclosure firstly enables timely and accurate initiation of the gesture recognition process by responding to a closed door and monitoring a target object within a preset spatial area, avoiding invalid or meaningless data processing and ensuring effective utilization of system resources. Secondly, by constructing a point cloud data queue using the point cloud data collected by the vehicle-mounted radar sensor within a preset time period, the sensor not only collects the raw data of the target object's hand movements but also preserves the continuity and dynamism of the movements through a time series analysis, laying the foundation for subsequent accurate analysis of hand movement characteristics. Furthermore, by performing temporal variation analysis on the point cloud data queue, the temporal variation features of multiple key parameters are extracted. The extraction and matching of these temporal variation features can significantly distinguish different hand gestures, achieving high-precision gesture recognition even in complex environments or backgrounds. Then, based on the temporal variation features, the gesture type is determined, enabling intelligent recognition and classification of a wide variety of gestures (including but not limited to grasping, waving, and drawing circles), thereby achieving a response to multi-mode gesture commands. Finally, based on the determined gesture type, the corresponding door opening method can be accurately matched, avoiding door control errors caused by gesture recognition mistakes, directly improving the intelligence level of door control and user experience. In summary, this disclosure achieves the technical effect of improving gesture recognition accuracy and thus enabling precise door control, thereby solving the technical problem of poor door control accuracy due to low gesture recognition accuracy in related technologies.
[0059] The door lock control system in the embodiments of this disclosure will be further described below.
[0060] In one optional embodiment, temporal change analysis is performed on the point cloud data in the point cloud data queue to obtain the temporal change characteristics of the target object's hand movements, including the following execution steps.
[0061] Velocity temporal variation characteristic analysis was performed on the point cloud data in the point cloud data queue to determine velocity variation characteristics. Distance temporal variation characteristic analysis was performed on the point cloud data in the point cloud data queue to determine distance variation characteristics. Angle temporal variation characteristic analysis was performed on the point cloud data in the point cloud data queue to determine angle variation characteristics. Based on the velocity variation characteristics, distance variation characteristics, and angle variation characteristics, the temporal variation characteristics were determined.
[0062] Specifically, in one optional embodiment, velocity temporal variation characteristic analysis is performed on the point cloud data in the point cloud data queue to determine the velocity variation characteristics, including the following execution steps.
[0063] Based on a point cloud data queue, a velocity temporal variation curve is plotted. According to the velocity temporal variation curve, the number of velocity peaks, the number of velocity troughs, and their relative positions are determined, where the relative position is the position of the velocity peak relative to the velocity trough. The velocity variation characteristics are then determined based on the number of velocity peaks, velocity troughs, and their relative positions.
[0064] Velocity temporal variation feature analysis identifies key steps in specific hand gestures by tracking changes in the velocity of a target object's hand within consecutive time frames in point cloud data. This analysis process reveals the dynamic characteristics of hand movements (such as acceleration, deceleration, and stopping).
[0065] The aforementioned velocity temporal variation curve can be obtained by plotting the velocity information of point cloud data points in the point cloud data queue into a time series graph. The peaks, troughs, and slope changes of the velocity temporal variation curve reflect the instantaneous velocity changes of hand movements, and the information of the velocity temporal variation curve is a direct basis for gesture recognition.
[0066] On a velocity time-series curve, peaks represent local maximum values of the target object's hand movement velocity, while troughs represent local minimum values. By analyzing the peak information (including quantity, location, and amplitude) and trough information (including quantity, location, and amplitude) in the velocity changes, specific hand gesture patterns, such as waving up and down or moving left and right, can be identified.
[0067] Optionally, mathematical methods can be used to analyze the number of peaks and troughs, and the relative positions of peaks and troughs in the velocity time-series curve. For example, for the "waving hands up and down" motion, it is necessary to find at least two peaks and troughs and verify whether they alternate in the order of "peak-trough-peak" and whether the peak velocity is significantly greater than the trough velocity, thereby determining the velocity change characteristics.
[0068] In one optional embodiment, distance temporal variation feature analysis is performed on the point cloud data in the point cloud data queue to determine the distance variation features, including the following steps: Based on the point cloud data queue, a distance temporal variation curve is plotted. Based on the distance temporal variation curve, the distance variation features are determined.
[0069] Distance time-series variation feature analysis focuses on the changing pattern of the distance between the target object's hand and the radar sensor over time. Distance time-series variation features can be used to identify whether a user is approaching or moving away from a vehicle, and are an important basis for judging hand gestures such as "grabbing," "pushing," or "pulling."
[0070] The aforementioned distance-time variation curve refers to a curve plotted by recording changes in the distance to a target object in point cloud data, showing how distance changes over time. The overall trend and local details of the curve are crucial for gesture recognition.
[0071] Different hand gestures can cause specific patterns of increase or decrease in the distance between the target and the radar sensor. For example, when making a "circling" gesture, the distance changes periodically, while when making a "push" gesture, the distance gradually decreases.
[0072] Optionally, based on the point cloud data queue, the distance between the target object and the radar sensor is determined, and a distance-time variation curve is plotted. Then, the overall trend of the distance-time variation curve is identified, such as whether a motion pattern of moving from far to near and then away again appears, and local details. Through the above identification operations, distance change characteristics are determined, providing additional recognition basis for gesture types.
[0073] In one optional embodiment, angular temporal variation feature analysis is performed on the point cloud data in the point cloud data queue to determine the angular variation features, including the following steps: Based on the point cloud data queue, pitch angle temporal variation curves and horizontal angle temporal variation curves are plotted. Based on the pitch angle temporal variation curves and horizontal angle temporal variation curves, the angular variation features are determined.
[0074] The analysis of angular temporal variation characteristics involves the changes in horizontal and pitch angles over time. By observing the rotation or tilt angle of the hand in space, complex gestures such as "drawing circles" or "crossing hands" can be identified.
[0075] Optionally, the aforementioned time-series curves of horizontal angle variation and pitch angle variation depict the angular changes of the hand in the horizontal and vertical directions, respectively, which are crucial for recognizing the directionality and spatial characteristics of gestures.
[0076] In certain gestures, the hand remains relatively stable in some directions in space, while undergoing significant changes in others. For example, in a pressing motion, the hand changes minimally in the horizontal plane, but as the hand moves towards the sensor, the distance and pitch angle change significantly.
[0077] Optionally, based on the point cloud data queue, curves showing the change of the hand's horizontal angle and pitch angle over time are plotted. The stability and volatility of these two curves are then evaluated.
[0078] Temporal variation characteristics are a comprehensive reflection of changes in speed, distance, and angle, providing spatiotemporal dynamic information about hand gestures and forming the basis for accurate gesture recognition. For example, by combining periodic changes in speed, the trend of distance moving from far to near and then away again, and specific fluctuations in angle, the complete characteristics of a "waving" gesture can be identified.
[0079] Figure 2 is an example diagram of the speed-time variation curve according to an embodiment of the present disclosure. As shown in Figure 2, the speed-time curve exhibits the characteristic of two or more obvious peaks and troughs periodically switching. The above characteristics correspond to the reciprocating motion process of the hand waving upward (accelerating to the peak), decelerating at the peak (trough), and waving downward (accelerating to the peak again), which is manifested as a periodic fluctuation of "speed first increases, then decreases, then increases again".
[0080] Figure 3 is an example diagram of the distance-time change curve according to an embodiment of the present disclosure. As shown in Figure 3, the distance-time curve generally shows a trend of moving from far to near and then away again, representing the complete interaction process of the user approaching the vehicle, completing the gesture action, and then leaving.
[0081] Figure 4 is an example diagram of the pitch angle time-series variation curve according to an embodiment of the present disclosure. As shown in Figure 4, the pitch angle-time curve exhibits periodic fluctuations with significantly increased amplitude. This is a decisive characteristic for judging up-and-down hand-waving movements, directly reflecting the large-amplitude movement of the hand in the vertical direction.
[0082] Figure 5 is an example diagram of the horizontal angle time-series variation curve according to an embodiment of the present disclosure. As shown in Figure 5, the horizontal angle-time curve changes only slightly and remains relatively stable. This indicates that the action does not move significantly in the horizontal direction, thereby effectively eliminating interference from actions such as circling or waving at an angle.
[0083] In one alternative embodiment, the "grabbing action" is characterized by a sudden and significant decrease in distance accompanied by a single-peak change in speed.
[0084] The "circle drawing motion" is manifested as a coordinated periodic change in the horizontal and pitch angles.
[0085] The characteristics of the "pushing action" are as follows: the speed change is a single-peak characteristic (accelerating first and then decelerating), the distance change is continuously decreasing (the user pushes forward, so the distance sensor gets closer and closer), the horizontal angle changes are small (first change characteristic), and the pitch angle changes are small (first change characteristic).
[0086] The characteristics of the "pull action" are as follows: the speed change is a single-peak characteristic (accelerating first and then decelerating), the distance change is continuously increasing (as the hand pulls back, the distance to the sensor becomes farther and farther), the horizontal angle change is small (first change characteristic), and the pitch angle change is small (first change characteristic).
[0087] The characteristics of the "pressing action" are as follows: the speed change is characterized by a single peak, and then the speed returns to zero (and is maintained after pressing); the distance change is characterized by a rapid decrease at first, and then maintained at a relatively close distance; the horizontal angle change is small (first change characteristic); and the pitch angle change is small.
[0088] The characteristics of the "raising hand" action are as follows: the speed change is a single-peak characteristic (accelerating and then decelerating during the raising of the hand); the distance change is a characteristic of first decreasing and then stabilizing (the distance decreases first and then remains constant as the hand is extended forward and raised upward at the same time); the horizontal angle changes very little (the first change characteristic); and the pitch angle increases rapidly (raising the hand) and then tends to stabilize.
[0089] The characteristics of the "cross-waving" are as follows: the speed change is characterized by two peaks (corresponding to waving to the left and right respectively), the distance change is relatively stable (the distance changes little when waving the hand in the horizontal direction), the horizontal angle first moves to the left (decreases) and then to the right (increases) or first moves to the right (increases) and then to the left (decreases), and the pitch angle changes very little (first change characteristic).
[0090] In an optional embodiment, the gesture recognition-based door control method further includes the following steps: Determining the velocity change and distance change corresponding to two adjacent point cloud data points in the point cloud data queue. Determining the number of first point cloud data points and the number of second point cloud data points in the point cloud data queue, wherein the pitch angle corresponding to the first point cloud data point is greater than a preset pitch angle threshold, and the horizontal angle corresponding to the second point cloud data point is greater than a preset horizontal angle threshold. Responding to the following conditions: the velocity change corresponding to any two adjacent point cloud data points is less than or equal to a preset velocity change threshold; the distance change corresponding to any two adjacent point cloud data points is less than or equal to a preset distance change threshold; the number of first point cloud data points satisfies a first quantity condition; and the number of second point cloud data points satisfies a second quantity condition, the determined gesture action type is recognized as a valid action, and the door is opened according to the door opening method corresponding to the gesture action type.
[0091] The aforementioned velocity change refers to the difference in velocity values between two adjacent point cloud data points in the point cloud data queue over time. Velocity change reflects the increase or decrease in hand speed over a short period and is crucial for recognizing rapidly changing hand gestures (such as waving).
[0092] The aforementioned distance changes reflect minute variations in the distance between the target object and the radar sensor. By comparing the distance values of adjacent point cloud data points, the continuity and directionality of hand movement trajectories can be understood, which helps to distinguish hand gestures such as "push" and "pull".
[0093] Optionally, the point cloud data queue is traversed, and for each point cloud data point, it is compared with its neighboring point cloud data points to calculate the velocity difference and distance difference between the point cloud data point and its neighboring point cloud data points. This process can use difference operations, that is, the velocity or distance value of the current point is subtracted from the corresponding value of the previous point.
[0094] By statistically analyzing the number of points in point cloud data that meet specific angular conditions, the spatial characteristics of hand gestures can be further analyzed. Changes in pitch and horizontal angles play a decisive role in recognizing the vertical and horizontal movements of hand gestures, and are especially crucial for distinguishing similar hand gestures such as waving up and down versus waving left and right.
[0095] The aforementioned preset pitch angle threshold and preset horizontal angle threshold are set to pre-defined thresholds to distinguish different hand gestures.
[0096] In one optional embodiment, the number of first point cloud data points with pitch angles greater than a preset pitch angle threshold and second point cloud data points with horizontal angles greater than a preset horizontal angle threshold in the point cloud data queue are counted, and this count is matched with a specific pattern of gesture recognition.
[0097] Optionally, each data point in the point cloud data queue is filtered to identify points with pitch angles greater than a preset pitch angle threshold (i.e., the first point cloud data point) and points with horizontal angles greater than a preset horizontal angle threshold (i.e., the second point cloud data point). Then, the number of points in the entire queue that meet the above conditions is counted, and the validity of the filtered point cloud data points and their conformity to the preset pattern are determined by using a threshold condition based on the number of points.
[0098] In one alternative embodiment, to improve the accuracy of angle detection, signal processing techniques unique to millimeter-wave radar, such as Doppler frequency shift, or angle compensation algorithms can be used for detection to ensure that the pitch and horizontal angle changes can be accurately measured and analyzed even during vehicle movement or hand movement.
[0099] Based on the first two steps, a series of threshold conditions are set to comprehensively evaluate the gesture. Only when all key parameters (speed change, distance change, pitch angle and horizontal angle change) meet the preset conditions will the recognized gesture be regarded as a valid action, thereby triggering the door opening control command corresponding to the gesture.
[0100] In one optional embodiment, the speed change, distance change, and number of angle change points obtained in the first two steps are compared with preset threshold conditions. If all conditions are met, the recognized gesture action type is determined to be a valid action, and then a control command is issued according to the door opening method (such as side-opening door, side-sliding door, etc.) that matches the gesture type to execute the door opening or closing action.
[0101] In one optional embodiment, a dynamic threshold adjustment mechanism is introduced when determining the threshold. That is, the threshold is adjusted in real time according to environmental conditions, user behavior history or vehicle status to adapt to the gesture recognition needs in different scenarios and improve the adaptability and robustness of the system.
[0102] This disclosed method for vehicle door control based on gesture recognition achieves effective recognition and differentiation of various gesture actions through comprehensive analysis of speed, distance, pitch angle, and horizontal angle in a point cloud data queue, as well as strict judgment of preset thresholds. This significantly improves the accuracy and reliability of gesture control. This method not only addresses the challenges of gesture recognition in complex environments and reduces misoperations, but also ensures stable performance under different lighting conditions, weather conditions, and user behavior patterns. Furthermore, by quantifying the effectiveness of gesture actions, it ensures that door opening control is only executed when the gesture features fully meet preset conditions, effectively avoiding unexpected gesture triggers and enhancing system security and the naturalness of user interaction.
[0103] In one optional embodiment, the gesture recognition-based door control method further includes the following execution steps.
[0104] In response to the speed change of any two adjacent point cloud data points being greater than a preset speed change threshold, or the distance change of any two adjacent point cloud data points being greater than a preset distance change threshold, or the number of the first point cloud data points not meeting the first quantity condition, or the number of the second point cloud data points not meeting the second quantity condition, the determined gesture action type is identified as a mis-touch action, and the door is controlled to remain closed.
[0105] During gesture recognition, the system compares the velocity changes of adjacent frames in the point cloud data queue with a preset velocity change threshold to determine whether the hand movement exceeds the expected dynamic range. If any velocity change exceeds the threshold, it means the gesture is too violent or irregular, and the system identifies the gesture as a mis-touch action.
[0106] Setting the preset speed change threshold requires comprehensive consideration of the average speed change and standard deviation of different gestures, as well as possible noise and environmental interference factors.
[0107] By calculating the velocity difference between adjacent points in the point cloud data queue, the system can monitor the dynamic speed of hand movements in real time, which helps to determine whether the gesture movements are smooth and standardized.
[0108] During the analysis of the point cloud data queue, the velocity change of adjacent points is calculated and then compared with a preset velocity change threshold. If any velocity change exceeds the threshold range, the system will immediately mark the current gesture recognition as a potential accidental touch and proceed to the subsequent accidental touch handling process.
[0109] The system determines whether a hand gesture meets preset conditions by monitoring the change in distance between the target object and the radar in the point cloud data queue. An excessively large change in distance indicates that the hand movement trajectory deviates from the normal recognition range or that other interference factors exist; in this case, the system will classify the gesture as a false trigger.
[0110] The preset distance change threshold is a threshold set based on the typical distance changes of gesture movements. The preset distance change threshold is set to determine whether the continuity and directionality of the gesture movement meet expectations.
[0111] By comparing the distances between adjacent points in the point cloud data queue, the subtle changes in the distance between the hand and the radar sensor are analyzed to determine the effectiveness and standardization of hand gestures.
[0112] When processing the point cloud data queue, the system calculates the distance change between each data point and the previous data point and compares it with a preset distance change threshold. Once any distance change exceeds the preset distance change threshold, it is considered that there is a risk of false triggering in the current recognition process, requiring further evaluation and processing.
[0113] Furthermore, if the number of first point cloud data points with a pitch angle greater than a preset pitch angle threshold does not meet the first quantity condition, it means that the gesture execution is not standardized or there is multi-target interference, and the system will judge it as a mis-touch action.
[0114] The first quantity condition mentioned above is a set quantity threshold. The first quantity condition is set to evaluate whether the number of pitch angle change point cloud data points is reasonable, so as to determine whether the vertical motion characteristics of the gesture meet expectations.
[0115] By counting the number of data points in the point cloud data queue whose pitch angle is greater than a preset pitch angle threshold, the system can determine whether the hand has performed a specific action in the vertical direction.
[0116] Optionally, the point cloud data queue is traversed, and the number of point cloud data points with a pitch angle greater than a preset pitch angle threshold is counted. If it is found that the number of points in the entire queue that meet the condition is not within the range of the first quantity condition, the current gesture recognition behavior is regarded as a mis-touch, and the subsequent mis-touch action handling mechanism is triggered.
[0117] The system counts the number of second point cloud data points in the gesture execution area whose horizontal angle is greater than a preset horizontal angle threshold. If the number of second point cloud data points does not meet the second condition, it indicates that the gesture movement is discontinuous or irregular in the horizontal direction, and the system therefore identifies the gesture as a mis-touch action.
[0118] The second quantity condition mentioned above is set as another quantity requirement for evaluating whether the number of point cloud data points for the change in horizontal angle meets the expected normal hand gesture movement.
[0119] By counting the number of data points in the point cloud data queue whose horizontal angle is greater than a preset horizontal angle threshold, the system can analyze the characteristics of hand movements in the horizontal direction, ensuring the continuity and standardization of hand gestures.
[0120] Taking grabbing and waving as examples, both distance and speed will vary during a waving motion. Some false alarms with negligible variations can be filtered out. Up and down waving results in larger pitch angles, while grabbing motions show smaller pitch and horizontal angle changes. Judging the pitch and horizontal angles separately can filter out some false alarms. The system judges the specific location of the gesture; only when a large number of targets are present at that location is it considered that someone is preparing to perform a gesture; otherwise, it is considered a false alarm. Specifically, in multi-person scenarios, when the system detects multiple targets simultaneously performing waving actions within the valid gesture area, it will be considered a conflict scenario and the door will be refused, ensuring that only clear single-person gesture intentions are responded to. By comprehensively judging false alarms using the above strategies, the probability of false alarms can be reduced.
[0121] Figure 6 is a schematic flowchart of a gesture recognition method according to an embodiment of the present disclosure. As shown in Figure 6, this gesture recognition method collects point cloud data based on a radar sensor. The radar sensor acquires point cloud data in the environment in real time, and each point cloud data point includes parameters such as velocity, distance, horizontal angle, and pitch angle. The algorithm processes the point cloud data to recognize specific gestures (including but not limited to grasping actions, waving up and down), and finally determines whether to execute the grasping action or ignore false positives. The overall process includes steps such as data acquisition, object detection, gesture feature extraction, and decision judgment.
[0122] Specifically, it includes the following steps.
[0123] Acquire radar point cloud data.
[0124] The processing unit reads the point cloud data of the current frame from the radar sensor. Each frame of point cloud data contains multiple point cloud data points, and each point cloud data point has fields such as velocity, distance, horizontal angle, and pitch angle.
[0125] Calculate the absolute value of the velocity of each point, and filter out the point cloud data point corresponding to the maximum velocity, which helps to focus on the hand object with the most significant motion.
[0126] Accumulate and cache the point cloud data with the maximum velocity in dozens of frames of data to form a historical data queue. The purpose of caching is for subsequent time series analysis and filtering to reduce the influence of noise. The cached dozens of frames of data correspond to approximately 1 second of data, which can be adjusted according to the gesture speed.
[0127] Perform mean filtering processing, that is, perform mean filtering on the data parameters (velocity, distance, horizontal angle, pitch angle) of the dozens of frames of point cloud data in the cache to smooth data fluctuations.
[0128] After performing the mean filtering operation, smoothed velocity, distance, horizontal angle and pitch angle data can be obtained.
[0129] Object existence judgment, that is, judging whether an object exists in a preset spatial region, where the preset spatial region is defined by a horizontal position and a vertical position (or distance), and the specific representation is as follows.
[0130] The horizontal position judgment range is -0.2d<dx<0.4d (d represents a specific unit);
[0131] The vertical position judgment range is 0.7d<dy<0.9d (d represents a specific unit).
[0132] Further judge whether there are peaks and troughs in the velocity curve, perform time-domain analysis on the cached historical velocity data, and detect its local extreme points.
[0133] On the basis of judging that peaks and troughs exist, further analyze the number and arrangement order of the peaks and troughs.
[0134] When the velocity curve satisfies the pattern of "two peaks and troughs, with the peaks and troughs intersecting and the peaks appearing first," a series of continuous and sequential parameter constraint checks are then initiated. These checks aim to verify the spatial stability and regularity of the gesture motion.
[0135] The specific steps of the judgment are as follows: First, determine whether the parameters related to the horizontal and pitch angles meet the preset conditions. Next, determine whether the distance change is within the allowable range. Finally, determine whether the speed change meets the stability requirements.
[0136] Figure 7 is a flowchart illustrating the determination of the validity of gesture recognition results according to an embodiment of this disclosure. As shown in Figure 7, the system starts, the radar sensor begins operation, and outputs point cloud data at a predetermined frame rate. The process begins and includes the following steps: Acquiring radar point cloud data: The processing unit reads the point cloud data of the current frame from the radar interface. The point cloud data of the current frame contains information on multiple detected target points. Judging spatial location and density: That is, judging whether there are enough point cloud data points located within a preset effective interaction space. This step ensures that targets are concentrated in the designated operation area. Angular feature filtering: For point clouds located within the effective space, their angular features are further statistically analyzed. When the number of data points with a pitch angle greater than a first threshold is greater than a first threshold, a first quantity condition is satisfied; when the number of data points with a horizontal angle greater than a second threshold is greater than a second threshold, a second quantity condition is satisfied. The first threshold is in the range of 8° to 14°, and the first quantity condition is 3 to 7 points. The second threshold is in the range of 12° to 18°, and the second quantity condition is no more than 5 points. This step is set to initially filter targets that conform to hand shape characteristics. The stability of angle changes is assessed by determining whether the number of data points in the point cloud data that simultaneously meet the following conditions is within five. Then, data points whose absolute pitch angle value is greater than a first threshold (within the range of 15° to 25°) are detected. Similarly, data points whose absolute horizontal angle value is greater than a first threshold (also within the range of 15° to 25°) are detected. This step is designed to ensure smooth angle changes in gesture movements.
[0137] The distance change is judged to determine whether the distance change of the target between consecutive frames is no greater than 0.4±0.1m. This judgment is verified in two steps to ensure the stability of the distance change.
[0138] Valid action determination: When the target meets all the above judgment conditions, the system determines it as a valid gesture action and triggers the corresponding response command.
[0139] The velocity change assessment serves as a supplementary assessment, verifying whether the target's velocity change is no greater than 0.4 ± 0.1 m / s, further ensuring motion stability. Finally, false alarm handling is performed.
[0140] In one alternative embodiment, a point cloud data queue is constructed using point cloud data, including the following execution steps.
[0141] For each frame of point cloud data, the velocities corresponding to multiple point cloud data points in each frame are obtained, and target point cloud data points are selected. The target point cloud data point is the point cloud data point with the highest corresponding velocity in the current frame. A point cloud data queue is then constructed based on these multiple target point cloud data points.
[0142] In this step, the velocity information of each point cloud data point is extracted from each frame of point cloud data, and then the point with the highest velocity is selected as the target point cloud data point. This is because at the same point in time, hand movements are usually reflected at the point with the highest velocity; therefore, this point is considered to be the most likely to represent the current hand gesture state.
[0143] The point cloud data mentioned above is a collection of three-dimensional position information of target objects in the environment collected by millimeter-wave radar. Each point contains information such as distance, horizontal angle, pitch angle and velocity.
[0144] The target point cloud data point has the highest velocity value in the current frame point cloud data. The target point cloud data point is the most likely to reflect the dynamic movement of the hand, which is crucial for gesture recognition.
[0145] The target point cloud data points selected from multiple consecutive frames of point cloud data are stored in a queue in chronological order. This queue can record the speed changes of hand movements over time, providing a continuous data stream for subsequent gesture recognition algorithms.
[0146] The aforementioned point cloud data queue is an ordered data set containing target point cloud data points from multiple consecutive frames, arranged in chronological order. This point cloud data queue serves as the input to the gesture recognition algorithm, designed to capture the dynamic changes in hand gestures.
[0147] 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 set for analysis, stored data, displayed data, etc.) involved in this disclosure 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 entry points are provided for users to choose to authorize or refuse.
[0148] According to another aspect of the present disclosure, a vehicle door control system based on gesture recognition is also provided. Figure 8 is a structural block diagram of a vehicle door control system based on gesture recognition according to an embodiment of the present disclosure. As shown in Figure 8, the vehicle door control system 800 based on gesture recognition includes: a gesture detection module 801 and a door control module 802; wherein, the gesture detection module 801 is communicatively connected to the door control module 802, and is configured to determine the gesture action type of the target object and the door opening method corresponding to the gesture action type according to the vehicle door control method based on gesture recognition described above; the door control module 802 includes a door controller, and is configured to control the opening of the door according to the door opening method.
[0149] According to another aspect of the present disclosure, a vehicle is also provided, including the above-described gesture recognition-based door control system.
[0150] Embodiments of this disclosure also provide a computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device on which the computer-readable storage medium resides to perform the methods of the various embodiments of this disclosure.
[0151] Optionally, in this embodiment, the storage medium may be configured to store a computer program configured to perform the following steps.
[0152] Step S11: In response to the door being closed and a target object existing in a preset space area, acquire point cloud data collected by the vehicle radar sensor within a preset time period, wherein the point cloud data is the motion feature data of the target object.
[0153] Step S12: Construct a point cloud data queue using the point cloud data.
[0154] Step S13: Perform temporal change analysis on the point cloud data in the point cloud data queue to obtain the temporal change characteristics of the target object's hand movements.
[0155] Step S14: Determine the type of gesture action of the target object based on the temporal change characteristics.
[0156] Step S15: Determine the door opening method corresponding to the gesture action type, and control the door opening according to the door opening method.
[0157] Embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements the methods of various embodiments of this disclosure.
[0158] Optionally, in this embodiment, the computer program performs the following steps when executed by the processor.
[0159] Step S11: In response to the door being closed and a target object existing in a preset space area, acquire point cloud data collected by the vehicle radar sensor within a preset time period, wherein the point cloud data is the motion feature data of the target object.
[0160] Step S12: Construct a point cloud data queue using the point cloud data.
[0161] Step S13: Perform temporal change analysis on the point cloud data in the point cloud data queue to obtain the temporal change characteristics of the target object's hand movements.
[0162] Step S14: Determine the type of gesture action of the target object based on the temporal change characteristics.
[0163] Step S15: Determine the door opening method corresponding to the gesture action type, and control the door opening according to the door opening method.
[0164] Embodiments of this disclosure also provide a computer program product including a non-volatile computer-readable storage medium configured to store a computer program that, when executed by a processor, implements the methods described in various embodiments of this disclosure.
[0165] Embodiments of this disclosure also provide a computer program that, when executed by a processor, implements the methods described in the various embodiments of this disclosure.
[0166] In the above embodiments of this disclosure, 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.
[0167] In the several embodiments provided in this disclosure, 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 couplings, direct couplings, or communication connections may be through some interfaces; indirect couplings or communication connections between units or modules may be electrical or other forms.
[0168] 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.
[0169] Furthermore, the functional units in the various embodiments of this disclosure 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.
[0170] 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 disclosure, 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 disclosure. 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.
[0171] The above description is only a preferred embodiment of this disclosure. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principles of this disclosure, and these improvements and modifications should also be considered within the scope of protection of this disclosure. Industrial applicability
[0172] In this disclosure, firstly, by responding to the closed state of the vehicle door and monitoring the target object within a preset spatial area, the gesture recognition process can be initiated promptly and accurately, avoiding invalid or meaningless data processing and ensuring the effective utilization of system resources. Secondly, a point cloud data queue is constructed using point cloud data collected by the vehicle-mounted radar sensor within a preset time period. This process not only collects the raw data of the target object's hand movements but also preserves the continuity and dynamism of the movements through time series analysis, laying the foundation for subsequent accurate analysis of hand movement characteristics. Furthermore, time-series change analysis is performed on the point cloud data queue to extract the time-series change features of multiple key parameters. The extraction and matching of these features can significantly distinguish different gestures, achieving high-precision gesture recognition even in complex environments or backgrounds. Then, the gesture type is determined based on the time-series change features, enabling intelligent recognition and classification of numerous gestures, including but not limited to grasping, waving, and drawing circles, thereby achieving a response to multi-mode gesture commands. Finally, based on the determined gesture type, the corresponding door opening method can be accurately matched, avoiding door control errors caused by gesture recognition mistakes, and directly improving the intelligence level of door control and user experience. In summary, this disclosure achieves the technical effect of improving gesture recognition accuracy and thus enabling precise door control, thereby solving the technical problem of low gesture recognition accuracy leading to poor door control accuracy in related technologies.
Claims
1. A vehicle door control method based on gesture recognition, characterized in that, include: In response to the vehicle door being closed and the presence of a target object in a preset space area, point cloud data collected by the vehicle radar sensor within a preset time period is acquired, wherein the point cloud data is the motion feature data of the target object; A point cloud data queue is constructed using the aforementioned point cloud data; Temporal variation analysis is performed on the point cloud data in the point cloud data queue to obtain the temporal variation characteristics of the hand movements of the target object; Based on the temporal change characteristics, the type of gesture action of the target object is determined; Determine the door opening method corresponding to the gesture action type, and control the door opening according to the door opening method.
2. The vehicle door control method based on gesture recognition according to claim 1, characterized in that, Temporal variation analysis is performed on the point cloud data in the point cloud data queue to obtain the temporal variation characteristics of the hand movements of the target object, including: Perform velocity temporal variation characteristic analysis on the point cloud data in the point cloud data queue to determine the velocity variation characteristics; Perform distance temporal variation feature analysis on the point cloud data in the point cloud data queue to determine the distance variation features; Angle temporal change feature analysis is performed on the point cloud data in the point cloud data queue to determine the angle change features; The temporal variation characteristics are determined based on the velocity variation characteristics, the distance variation characteristics, and the angle variation characteristics.
3. The vehicle door control method based on gesture recognition according to claim 2, characterized in that, The velocity temporal variation feature analysis is performed on the point cloud data in the point cloud data queue to determine the velocity variation features, including: Based on the point cloud data queue, a velocity temporal variation curve is plotted; Based on the velocity time-series variation curve, determine the number of velocity peaks, the number of velocity troughs, and their relative positions, wherein the relative position is the position of the velocity peak relative to the velocity trough. The velocity change characteristics are determined based on the number of velocity peaks, the number of velocity troughs, and their relative positions.
4. The vehicle door control method based on gesture recognition according to claim 2, characterized in that, The distance temporal variation feature analysis is performed on the point cloud data in the point cloud data queue to determine the distance variation features, including: Based on the point cloud data queue, a distance time series variation curve is plotted; Based on the distance time-series change curve, the distance change characteristics are determined.
5. The vehicle door control method based on gesture recognition according to claim 2, characterized in that, Perform the angle temporal change feature analysis on the point cloud data in the point cloud data queue to determine the angle change features, including: Based on the point cloud data queue, plot the pitch angle time-series variation curve and the horizontal angle time-series variation curve; The angle change characteristics are determined based on the pitch angle time-series change curve and the horizontal angle time-series change curve.
6. The vehicle door control method based on gesture recognition according to claim 5, characterized in that, The pitch angle time-series variation curve is set to determine the angle change of the hand in the horizontal direction, and the horizontal angle time-series variation curve is set to determine the angle change of the hand in the vertical direction.
7. The vehicle door control method based on gesture recognition according to claim 2, characterized in that, The gesture recognition-based door control method further includes: Determine the velocity change and distance change corresponding to two adjacent point cloud data points in the point cloud data queue; Determine the number of first point cloud data points and the number of second point cloud data points in the point cloud data queue, wherein the pitch angle corresponding to the first point cloud data point is greater than a preset pitch angle threshold, and the horizontal angle corresponding to the second point cloud data point is greater than a preset horizontal angle threshold. In response to the following conditions: the speed change corresponding to any two adjacent point cloud data points is less than or equal to a preset speed change threshold; the distance change corresponding to any two adjacent point cloud data points is less than or equal to a preset distance change threshold; the number of the first point cloud data points meets a first quantity condition; and the number of the second point cloud data points meets a second quantity condition, the determined gesture action type is identified as a valid action, and the door is opened according to the door opening method corresponding to the gesture action type.
8. The vehicle door control method based on gesture recognition according to claim 7, characterized in that, The gesture recognition-based door control method further includes: In response to any of the following: the speed change corresponding to any two adjacent point cloud data points is greater than the preset speed change threshold; or the distance change corresponding to any two adjacent point cloud data points is greater than the preset distance change threshold; or the number of the first point cloud data points does not meet the first quantity condition; or the number of the second point cloud data points does not meet the second quantity condition, the determined gesture action type is identified as a mis-touch action, and the door is controlled to remain in the closed state.
9. The vehicle door control method based on gesture recognition according to claim 1, characterized in that, Constructing the point cloud data queue using the point cloud data includes: For each frame of point cloud data included in the point cloud data, the velocity corresponding to multiple point cloud data points in each frame of point cloud data is obtained, and the target point cloud data point is selected, wherein the target point cloud data point is the point cloud data point with the maximum velocity in the current frame of point cloud data. The point cloud data queue is constructed based on multiple target point cloud data points.
10. The vehicle door control method based on gesture recognition according to claim 2, characterized in that, Based on the temporal change characteristics, the gesture action type of the target object is determined, including: In response to the timing change characteristics satisfying a first preset condition, the gesture action type is determined to be a pull action; In response to the timing change characteristics satisfying the second preset condition, the gesture action type is determined to be a push action; The first preset condition is that the velocity change feature is a single-peak feature, the distance change feature continues to decrease, the horizontal angle change feature is a first change feature, and the pitch angle change feature is a first change feature. The second preset condition is that the velocity change feature is the single-peak feature, the distance change feature continues to increase, the horizontal angle change feature is the first change feature, and the pitch angle change feature is the first change feature.
11. The vehicle door control method based on gesture recognition according to claim 10, characterized in that, Based on the temporal change characteristics, the gesture action type of the target object is determined, including: In response to the timing change characteristics satisfying a third preset condition, the gesture action type is determined to be a pressing action; The third preset condition is that the speed change feature is the single-peak feature, the distance change feature decreases to the first preset distance value, the horizontal angle change feature is the first change feature, and the pitch angle change feature is the first change feature.
12. The vehicle door control method based on gesture recognition according to claim 10, characterized in that, Based on the temporal change characteristics, the gesture action type of the target object is determined, including: In response to the temporal change feature satisfying the fourth preset condition, the gesture action type is determined to be a hand-raising action; The fourth preset condition is that the speed change feature is the single-peak feature, the distance change feature decreases to the second preset distance value, the horizontal angle change feature is the first change feature, and the pitch angle change feature increases to the preset pitch angle value.
13. The vehicle door control method based on gesture recognition according to claim 10, characterized in that, Based on the temporal change characteristics, the gesture action type of the target object is determined, including: In response to the timing change feature satisfying the fifth preset condition, the gesture type is determined to be a cross-waving gesture; The fifth preset condition is that the velocity change feature is a double-peak feature, the horizontal angle change feature first increases to a first preset horizontal angle value and then decreases to a second preset horizontal angle value, and the pitch angle change feature is the first change feature; or, the velocity change feature is the double-peak feature, the horizontal angle change feature first decreases to the second preset horizontal angle value and then increases to the first preset horizontal angle value, and the pitch angle change feature is the first change feature.
14. A vehicle door control system based on gesture recognition, characterized in that, include: Gesture detection module and door control module; The gesture detection module is communicatively connected to the door control module and is configured to determine the gesture action type of the target object and the door opening method corresponding to the gesture action type according to any one of claims 1 to 13. The door control module includes a door controller, which is configured to control the opening of the door according to the door opening method.
15. A vehicle, characterized in that, include: The gesture recognition-based door control system as described in claim 14.
