Safety belt hand tracking method and system based on image recognition

Abstract

The invention relates to a safety belt hand tracking method and system based on image recognition, and relates to the technical field of automobile intelligence. Analyzing the driver video information to determine the number of hands of the current driver; if the number of the hands of the current driver falls into the interference hand number range, acquiring hand features of the current driver; determining an arm contour starting point of the driver; determining the hand position of a driver; obtaining a driver door direction; determining an expected door opening hand position of the driver; when the driver predicts that the door opening hand position falls into the safety belt sliding position area, a safety belt buckle sliding signal is output; in response to the safety belt buckle slippage signal, determining safety belt buckle slippage parameters through the position of the door opening hand predicted by the driver; and safety belt buckle slippage operation is executed based on the safety belt buckle slippage parameters. The method has the effects that the hand position of a driver is accurately tracked, the safety belt buckle is intelligently driven to move in a matched mode, and the convenience of getting-off unbinding is improved.

Description

Technical Field

[0001] This invention relates to the field of automotive intelligent technology, and in particular to a method and system for tracking seat belt hands based on image recognition. Background Technology

[0002] Seat belts are one of the most important passive safety devices in modern automobiles. Their core function is to quickly tighten the webbing through the locking function of the retractor in the event of a collision or emergency braking, effectively restraining the occupants and preventing them from secondary collisions with hard objects inside the vehicle due to inertia, thereby significantly reducing the risk of injury or death. Since its invention, seat belt technology has continuously evolved, from the initial two-point seat belts to the commonly used three-point seat belts, as well as pre-tensioned seat belts and force-limiting seat belts, greatly improving both safety and comfort.

[0003] Regarding the aforementioned technologies, in scenarios where passengers exit the vehicle, traditional seat belts require manual pressing of the buckle release button at a fixed position to unfasten them. When users are carrying items with both hands or are in a narrow parking space where turning around is inconvenient, the convenience of manually pressing the buckle at a fixed position is greatly reduced. Summary of the Invention

[0004] To address the difficulties and inconveniences of manually unlocking traditional seat belts in certain scenarios, this invention provides a seat belt hand tracking method and system based on image recognition.

[0005] In a first aspect, the present invention provides a seatbelt hand tracking method based on image recognition, employing the following technical solution: A seatbelt hand tracking method based on image recognition, comprising: Obtain driver video information; Analyze the driver's video information to determine the current number of the driver's hands; If the current number of driver's hands falls within the preset range of interfering hand numbers, the current driver's hand features are obtained; Determine the starting point of the driver's arm contour based on the current characteristics of the driver's hand; The driver's hand position is determined based on the starting point of the driver's arm contour; Get the driver's door direction; The driver's expected hand position for opening the door is determined based on the driver's door direction and the current position of the driver's hands; When the driver anticipates that their hand will fall into the preset seatbelt sliding position area when opening the door, a seatbelt buckle sliding signal is output. In response to the seatbelt buckle slip signal, the seatbelt buckle slip parameters are determined based on the driver's expected hand position when opening the door; The seatbelt buckle sliding operation is performed based on the seatbelt buckle sliding parameters.

[0006] By adopting the above technical solution, image recognition technology is used to detect the number of driver's hands, eliminate interfering hands, accurately locate the position of the driver's hands, and predict the position of the hand opening the door in combination with the direction of the door. This drives the seat belt buckle to slide to the corresponding position, shortens the operation distance for the driver to unfasten the seat belt, reduces manual adjustment steps, and improves the convenience and efficiency of unfastening the seat belt.

[0007] Optionally, methods for determining the driver's expected hand position for opening the door based on the driver's door direction and the current driver's hand position include: If the current driver's hand position falls within the preset steering wheel area, the driver's expected door opening hand position is determined by the driver's door direction and the preset driver's hand habit position; If the current driver's hand position is not within the steering wheel area, obtain the current driver's hand position off the steering wheel and the number of times the current driver's hand position is off the steering wheel; If the current number of times the driver's hands are off the steering wheel is the preset number of times both hands are off the steering wheel, the driver's expected hand position for opening the door is determined based on the direction of the driver's door. If the number of times the driver's hands are off the steering wheel is less than the number of times both hands are off the steering wheel, the driver's expected hand position for opening the door is determined based on the current position of the driver's hands off the steering wheel.

[0008] By adopting the above technical solutions, a door opening hand position prediction strategy is formulated for different scenarios based on whether the driver's hands are in the steering wheel area and the number of fingers off the steering wheel. This improves the accuracy of door opening hand position prediction under different driving postures and ensures that the seat belt buckle sliding operation is highly matched with the driver's actual door opening intention.

[0009] Optionally, it also includes a method for not outputting a seatbelt buckle slip signal when the driver anticipates that their hand position will fall into the seatbelt slip position area, the method comprising: Determine the seatbelt adjustment frequency when the driver anticipates that their hand will fall into the seatbelt slippage area when opening the door. When the seat belt adjustment frequency does not fall within the preset clothing adjustment frequency range, output a seat belt buckle slip signal; When the seat belt adjustment frequency falls within the range of clothing adjustment frequency, a driver clothing adjustment signal is output. In response to a signal indicating that the driver has adjusted their clothing, no signal indicating that the seatbelt buckle has slipped is output.

[0010] By adopting the above technical solution, the system can distinguish between the driver's clothing adjustment action and the normal door opening action by determining the seat belt adjustment frequency and comparing it with the clothing adjustment frequency range. This avoids the seat belt buckle slipping operation being mistakenly performed when the user is tidying up their clothes, improves the reliability of the system's action judgment, and prevents problems such as clothing tangling caused by accidental slipping.

[0011] Optionally, it also includes a method for performing seatbelt pressure parameter adjustment, the method comprising: When a signal indicating that the driver has adjusted their clothing is received, the distance the seatbelt has moved is obtained. When the seat belt moves beyond the preset range for seat belt movement due to vehicle bumps, a seat belt adjustment operation is initiated. If a seatbelt adjustment operation is performed, an abnormal seatbelt movement signal will be output. In response to a seatbelt abnormal movement signal, the seatbelt adjustment frequency is obtained; If the seat belt adjustment frequency falls within the preset abnormal seat belt adjustment frequency range, the seat belt pressure parameter adjustment operation will be performed according to the preset single pressure reduction parameter.

[0012] By adopting the above technical solution, after detecting the clothing adjustment signal, the active adjustment action of the seat belt is further identified. When the adjustment frequency reaches the abnormal threshold, the seat belt pressure relief operation is automatically executed, which effectively alleviates the restraint and discomfort caused by the seat belt being too tight and improves driving and riding comfort.

[0013] Optionally, it also includes a method for not outputting a seatbelt abnormal movement signal when a seatbelt adjustment operation is performed, the method comprising: Record the time range of seat belt adjustment based on the seat belt adjustment operation; By analyzing the driver's video information within the seatbelt adjustment time range, video information of the driver's hand movements can be obtained. Determine the driver's hand movement trend based on video information of the driver's hand movements; If the driver's hand movement tends to fall into the seat belt slippage area, it indicates that the driver is adjusting the seat belt; If the driver adjusts the seat belt, an abnormal seat belt movement signal will be output; If there is no driver adjusting the seat belt, no abnormal seat belt movement signal will be output.

[0014] By adopting the above technical solution and combining the hand movement video captured within the seat belt adjustment time range, the system analyzes the driver's hand movement trend, accurately distinguishes between the driver's active adjustment of the seat belt and passive movement caused by vehicle bumps, avoids misjudging non-active adjustment behavior and triggering pressure adjustment, and improves the accuracy and rigor of the system's judgment.

[0015] Optionally, it also includes a method for updating the driver's habitual hand position, which includes: After performing the seatbelt buckle sliding operation, determine the target position of the seatbelt sliding; A current seatbelt sliding target position queue is formed based on the seatbelt sliding target position; The maximum number of consistent target positions for seat belts is determined by the seat belt slip target position queue and seat belt slip target positions; Get the length of the queue for the seatbelt sliding target position; The proportion of the number of seat belt consistent target positions is determined based on the maximum number of seat belt consistent target positions and the length of the seat belt slippage target position queue. If the percentage of the number of consistent target positions for the maximum seat belt exceeds the preset threshold for the percentage of reliable seat belt target positions, the driver's hand position will be updated.

[0016] By adopting the above technical solution, a target position queue for seat belt sliding is constructed, and the driver's long-term operating habits are statistically analyzed. When the proportion of the target position reaches a reliable threshold, the habitual hand position is updated, and sliding is then performed directly according to the habitual position. This reduces the workload of real-time image recognition and calculation, improves the response speed of seat belt sliding, and achieves personalized adaptation.

[0017] Optional, also includes: Collect the current back pressure value of the driver; A queue of driver back pressure values ​​is generated based on the current driver back pressure value. The current driver baseline pressure value is generated based on the driver back pressure value queue; If the current driver's back pressure value is lower than the current driver's reference pressure value, determine the area of ​​reduced back pressure. If the current driver's back pressure value is greater than the current driver's baseline pressure value, determine the area of ​​increased back pressure. The force distribution relationship in the seat area is determined by the areas of increased back pressure and decreased back pressure. If the force distribution in the seat area is vertical, determine the vertical distance between the two positions; The current seatbelt pressure adjustment parameters are determined based on the vertical relative position distance. If the force distribution in the seat area is left-right, determine the relative distance between the left and right positions; The current seatbelt pressure adjustment parameters are determined based on the relative distance between the left and right sides. Perform the seatbelt pressure adjustment operation according to the current seatbelt pressure adjustment parameters.

[0018] By adopting the above technical solution, the driver's back pressure value is collected, the positional relationship of the pressure rise and fall area is analyzed to determine the driver's posture changes (forward leaning, backward leaning, and side leaning), and the seat belt pressure parameters are dynamically adjusted based on the relative position distance, so that the seat belt pressure is adapted to the driver's real-time posture, further improving driving comfort and fit.

[0019] Optional, also includes: Get the current vehicle speed in real time; If the current vehicle speed is not 0, the seatbelt pressure parameter adjustment operation will not be performed.

[0020] By adopting the above technical solution, seat belt pressure adjustment can be performed without changing the seat belt pressure during vehicle operation, ensuring that the seat belt maintains appropriate pressure at all times and protecting the safety of the driver and passengers. Because changes in vehicle speed and road conditions can cause frequent shifts in the body position of the driver and passengers during vehicle operation, adjusting the seat belt pressure under these circumstances could lead to unstable seat belt pressure, affecting its restraining effect on the occupants.

[0021] Optionally, it also includes a method for adjusting seatbelt pressure parameters during the seatbelt buckle sliding operation, the method comprising: When performing the seatbelt buckle sliding operation, obtain the buckle movement speed; When the movement speed of the buckle falls within the preset stationary speed range, a clothing friction signal is output; In response to clothing friction signals, the seatbelt pressure parameter is adjusted according to the single pressure reduction parameter; After the seatbelt pressure parameter reduction operation is performed, the seatbelt buckle sliding operation is performed based on the seatbelt buckle sliding parameter.

[0022] By employing the above technical solution, during the seatbelt buckle sliding operation, the presence of clothing friction is determined by detecting the buckle's movement speed. When the buckle's movement speed falls within the stationary speed range, it indicates that clothing friction may be causing obstruction to the movement. In this case, a clothing friction signal is output, and a pressure relief operation is performed. This effectively reduces discomfort caused by clothing friction and prevents clothing friction from affecting the smoothness of seatbelt sliding. Performing the sliding operation based on the seatbelt buckle sliding parameters after completing the pressure relief operation makes the seatbelt sliding process smoother, further improving the user experience for drivers and passengers.

[0023] Secondly, the present invention provides a seatbelt hand tracking system based on image recognition, which adopts the following technical solution: A seatbelt hand tracking system based on image recognition, comprising: The acquisition module is used to acquire driver video information and the driver's door direction; The memory is used to store the program of the image recognition-based seat belt hand tracking method described above; The processor loads and executes programs from memory.

[0024] By adopting the above technical solution, and by setting up an acquisition module, memory and processor, a hardware carrier is provided for the above seat belt hand tracking method, ensuring the stable implementation and efficient execution of the method, and realizing a closed loop of the entire process from data acquisition, logical operation to action execution.

[0025] In summary, the present invention has at least one of the following beneficial technical effects: By using image recognition technology to accurately detect the driver's hand position and eliminate interfering hands, and predicting the hand position when opening the door according to different scenarios, the seat belt buckle is driven to actively slide to the corresponding position, shortening the operation distance for the driver to contact and unfasten the seat belt, reducing manual adjustment steps, and greatly improving the convenience and efficiency of unfastening the seat belt when getting out of the car; A multi-layered identification mechanism is constructed, which can distinguish between clothing adjustment and normal door opening actions by adjusting the frequency, thus preventing seat belt slippage; and can also combine hand movement video to distinguish between driver's active adjustment of seat belt and passive movement caused by vehicle bumps, preventing false triggering of pressure adjustment, and ensuring the stability and accuracy of system operation. On the one hand, by learning drivers' long-term operating habits and updating their hand positions, the real-time computation load is reduced, and the seat belt slip response speed is improved. On the other hand, by combining back pressure sensing with changes in driver posture, the seat belt pressure parameters are dynamically adjusted to adapt the seat belt to different posture needs, significantly improving driving comfort and personalized experience. Attached Figure Description

[0026] Figure 1 This is a flowchart of a seatbelt hand tracking method based on image recognition, as described in an embodiment of this application. Detailed Implementation

[0027] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0028] This invention discloses a seatbelt hand tracking method based on image recognition. (Refer to...) Figure 1 A seatbelt hand tracking method based on image recognition includes: Step S1: Obtain driver video information.

[0029] Driver video information refers to the image or video stream data of the driver's area collected in real time by the in-vehicle camera, including visual information such as the driver's upper body posture, hand movements, and body position.

[0030] Step S2: Analyze the driver's video information to determine the current number of the driver's hands.

[0031] The current driver hand count refers to the total number of hands in the video frame after the driver's video information is analyzed by the YOLO object detection algorithm. The driver's hands may be present in the driver's area, as well as the hands of the front passenger and rear passengers. The current driver hand count is used to determine whether there is any external hand interference.

[0032] Step S3: If the current number of driver's hands falls within the preset range of interfering hand numbers, obtain the current driver's hand features.

[0033] The range of interfering hands refers to the preset range of the number of non-driver hand interferences that the system determines to exist. Specifically, it is set to more than 3 hands. That is, when 3 or more hands are detected, it is determined that there is external hand interference.

[0034] The current driver hand features refer to the visual characteristics of each detected hand, including the outline shape of the hand and the connection relationship with the surrounding limbs, which is the core basis for distinguishing the driver's hand from the interfering hand.

[0035] Step S4: Determine the starting point of the driver's arm contour based on the current driver's hand features.

[0036] The starting point of the driver's arm contour refers to the pixel position where a continuous and complete arm contour is first identified, extending from each detected hand toward the torso.

[0037] Step S5: Determine the position of the driver's hand based on the starting point of the driver's arm contour.

[0038] The driver's hand position refers to the real-time pixel coordinates or spatial region of the remaining driver's hand in the video frame after removing those whose continuous arm contours cannot be tracked (i.e., interfering hands) from all detected hands.

[0039] Step S6: Obtain the driver's door direction.

[0040] The driver's door direction refers to the lateral position of the driver's exit door as preset according to the vehicle's factory configuration, specifically the left side of a left-hand drive vehicle and the right side of a right-hand drive vehicle.

[0041] Step S7: Determine the driver's expected hand position for opening the door based on the driver's door direction and the current driver's hand position.

[0042] The driver's predicted hand position for opening the door refers to the spatial area in which the driver's hand, which is used to open the door, is predicted by the system based on the driver's door orientation and real-time hand position.

[0043] Step S8: When the driver expects the hand position to fall into the preset seat belt sliding position area, output a seat belt buckle sliding signal.

[0044] The seatbelt slippage area refers to the pixel area covered by the seatbelt webbing in the captured image, which is pre-marked using driver video information. Specifically, it is a rectangular area extending 5 to 10 cm outward from the edge of the webbing.

[0045] The seatbelt buckle slip signal refers to the electrical command signal output by the vehicle's ECU (Electronic Control Unit) to the seatbelt control module, which triggers the seatbelt buckle to slip.

[0046] When the driver anticipates that their hand will fall into the seatbelt sliding position area when opening the door, it indicates that the driver has a clear intention to open the door and get out of the vehicle. At this time, by moving the seatbelt buckle to the position mapped to the driver's hand, the operating distance for the driver to touch and unlock the seatbelt can be shortened, reducing the steps of manually adjusting the seatbelt position and helping the driver to complete the seatbelt unfastening operation more quickly.

[0047] Step S9: In response to the seat belt buckle slip signal, determine the seat belt buckle slip parameters based on the driver's expected hand position when opening the door.

[0048] The seatbelt buckle sliding parameters refer to the specific instruction parameters used to control the seatbelt buckle to move along the C-shaped track, calculated based on the expected hand position when opening the door. These parameters include the buckle's target position, direction of movement (up or down), speed of movement, and distance of movement. The C-shaped track here is not an independent physical guide rail, but rather a C-shaped sliding path formed by the seatbelt webbing along a preset path on the vehicle body. The webbing is fixed along the edge of the driver's seat back and the side pillar of the door, and its overall trajectory conforms to the spatial contour of the driver's upper body. The seatbelt buckle is directly fitted onto this webbing and can slide along the C-shaped track of the webbing. This webbing path is the buckle's sliding track.

[0049] Step S90: Perform seat belt buckle sliding operation based on seat belt buckle sliding parameters.

[0050] The seat belt buckle sliding operation refers to the process by which the seat belt control system, based on calculated sliding parameters, drives the motor and transmission mechanism to move the seat belt buckle along a C-shaped track to the target position.

[0051] The methods for determining the driver's expected hand position for opening the door based on the driver's door direction and the current position of the driver's hand include: Step S70: If the current driver's hand position falls into the preset steering wheel area, determine the driver's expected door opening hand position by the driver's door direction and the preset driver's habitual hand position.

[0052] The steering wheel area refers to a rectangular pixel area encompassing the entire steering wheel, pre-marked using driver video information, used to determine whether the driver's hands are still in a driving grip position.

[0053] The driver's habitual hand position refers to the spatial position of the driver's hand when opening the door, which is learned by the system through historical sliding data and is used stably by the driver over a long period of time. It is the optimal seat belt buckle positioning reference point adapted to the driver's operating habits.

[0054] If the driver's hands are currently positioned within the steering wheel area, it means that the driver's hands are still in a steering wheel gripping posture and have not yet made a door opening action. Therefore, it is impossible to determine the intention to open the door based on the real-time hand position. The positioning direction of the seat belt buckle needs to be planned in advance based on historical habit positions.

[0055] Step S71: If the current driver's hand position does not fall into the steering wheel area, obtain the current driver's hand position off the steering wheel and the number of times the current driver's hand position off the steering wheel is obtained.

[0056] The current position of the driver's hands off the steering wheel refers to the real-time pixel coordinates or spatial area of ​​all hands that are not in the steering wheel area in the driver's video information.

[0057] The current number of driver's hands off the steering wheel refers to the total number of the driver's hands that are not in the steering wheel area, and the value is 1 or 2.

[0058] If the driver's hands are not currently in the steering wheel area, it means that at least one of the driver's hands has left the steering wheel, indicating that the driver has entered the stage of preparing to get out of the car and has a preliminary tendency to open the door. The intention to open the door can be further judged by the position of the hand that is off the steering wheel.

[0059] Step S72: If the current number of the driver's hands off the steering wheel is the preset number of hands off the steering wheel, determine the driver's expected hand position for opening the door based on the direction of the driver's door.

[0060] The number of times the driver's hands are removed refers to the system's preset threshold for determining the number of times the driver's hands are completely off the steering wheel, specifically set to 2.

[0061] If the number of hands the driver has taken off the steering wheel is the preset number, it means that both hands of the driver have left the steering wheel. The driver may be preparing to get out of the car by organizing items, adjusting their seating position, or performing other preparatory actions. It is not possible to determine the driver's intention to open the door based on the position of a single hand. The driver's expected hand position for opening the door must be determined according to the direction of the door.

[0062] Step S73: If the number of times the driver's hands are off the steering wheel is less than the number of times both hands are off the steering wheel, determine the driver's expected hand position for opening the door based on the current position of the driver's hands off the steering wheel.

[0063] If the number of times the driver's hands are off the steering wheel is less than the number of times both hands are off the steering wheel, it means that the driver has only taken one hand off the steering wheel while the other hand is still holding the steering wheel to maintain body balance. The hand that is off the steering wheel is most likely the hand that will be used to open the door. The intention is clear and can be directly used as the positioning basis for the seat belt buckle, shortening the distance of subsequent unfastening operations.

[0064] This also includes a method for not outputting a seatbelt buckle slip signal when the driver anticipates that their hand will fall into the seatbelt slip position area when opening the door. This method includes: Step S80: When the driver anticipates that the hand position when opening the door will fall into the seat belt slippage area, determine the seat belt adjustment frequency.

[0065] Seatbelt adjustment frequency refers to the number of times a driver's hands complete the actions of entering, adjusting, and leaving the seatbelt within the seatbelt sliding position area per unit of time. It is used to characterize the frequency of hand activity in the seatbelt area.

[0066] Step S81: When the seat belt adjustment frequency does not fall within the preset clothing adjustment frequency range, output a seat belt buckle sliding signal.

[0067] The clothing adjustment frequency range refers to the frequency range that is pre-set through multiple experiments to determine whether the driver's actions constitute clothing adjustments.

[0068] When the seatbelt adjustment frequency does not fall within the range of clothing adjustment frequency, it indicates that the driver's hand activity frequency in the seatbelt area is low, which is a single or few touches before opening the door, rather than repeated clothing adjustment behavior. In this case, the seatbelt buckle sliding signal can be output normally to assist in opening the door.

[0069] Step S82: When the seat belt adjustment frequency falls within the clothing adjustment frequency range, output the driver's clothing adjustment signal.

[0070] The driver's clothing adjustment signal refers to the electrical command signal output by the vehicle ECU to the seat belt control module, which is used to pause the seat belt buckle sliding operation. It is the control signal after the system determines the clothing adjustment action.

[0071] When the frequency of seatbelt adjustment falls within the range of clothing adjustment frequency, it indicates that the driver's hand movements in the seatbelt area are high-frequency, repetitive, small-amplitude movements. This is most likely the adjustment action after the corner of the coat, the shoulder strap of the backpack, etc., are caught by the seatbelt, rather than the preparation action before opening the door.

[0072] Step S83: In response to the driver's clothing adjustment signal, do not output a seat belt buckle slip signal.

[0073] When a signal indicating that the driver is adjusting their clothing is present, it means that the system has clearly determined that the driver's current action priority is adjusting their clothing and that they have no intention of getting out of the vehicle. If the seat belt buckle is moved at this time, the clothing may become tangled further. Therefore, the system pauses the output of the movement signal and waits for the driver to finish adjusting their clothing before re-detecting the intention to open the door.

[0074] This also includes a method for performing seatbelt pressure parameter adjustment, the method comprising: Step S84: When there is a signal indicating that the driver has adjusted their clothing, obtain the seat belt movement distance.

[0075] The seat belt movement distance refers to the actual displacement length of the seat belt webbing along the C-shaped track during the duration of the driver's clothing adjustment signal, and is measured based on the starting mark of the C-shaped track.

[0076] Step S840: When the seat belt moves beyond the preset range for seat belt movement during vehicle bumps, determine the seat belt adjustment operation.

[0077] The range of seatbelt movement distance due to vehicle bumps refers to the passive displacement range determined by vehicle bumps, established through multiple experiments, used to distinguish passive displacement from active user adjustment. Seatbelt adjustment refers to actions determined by the driver to be actively touching or moving the seatbelt webbing or buckle, after eliminating vehicle bumps as a contributing factor.

[0078] Step S85: If a seatbelt adjustment operation is performed, output a seatbelt abnormal movement signal.

[0079] The seatbelt abnormal movement signal refers to the electrical command signal output by the vehicle ECU to the seatbelt pressure control module, which is used to prompt the system to detect abnormal behavior of the driver actively adjusting the seatbelt.

[0080] If seatbelt adjustment is performed, it indicates that the movement of the seatbelt is not caused by external factors such as vehicle bumps, but rather by the driver actively touching or moving the seatbelt while adjusting clothing or due to discomfort. This suggests that the seatbelt may be too tight, affecting comfort.

[0081] Step S86: In response to the abnormal movement signal of the seat belt, obtain the seat belt adjustment frequency.

[0082] Step S87: If the seat belt adjustment frequency falls within the preset abnormal seat belt adjustment frequency range, perform the seat belt pressure parameter adjustment operation according to the preset single pressure reduction parameter.

[0083] The abnormal seatbelt adjustment frequency range refers to a pre-set, high-frequency adjustment interval determined by the driver to indicate that the seatbelt is too tight. It distinguishes between occasional adjustments and frequent, uncomfortable adjustments. The frequency of abnormal seatbelt adjustments is higher than that of clothing adjustments, distinguishing between high-frequency adjustments made by the driver due to discomfort from the seatbelt and ordinary clothing adjustments. It reflects the high-frequency adjustments made by the driver due to discomfort from the seatbelt.

[0084] The single pressure reduction parameter refers to the length of webbing released by the seat belt retractor or the pressure reduction value when each pressure adjustment is performed, which is preset through multiple experiments. For example, releasing 2cm of webbing.

[0085] The seatbelt pressure parameter adjustment operation refers to the process by which the seatbelt control system, upon receiving a command, drives the retractor to release a preset length of webbing, thereby reducing the restraining pressure of the seatbelt on the driver's body.

[0086] If the seatbelt adjustment frequency falls within the preset abnormal seatbelt adjustment frequency range, it indicates that the driver is frequently moving the seatbelt to relieve discomfort while adjusting clothing because the seatbelt is too tight. In this case, it is necessary to actively reduce the seatbelt pressure to improve driving and riding comfort.

[0087] This also includes a method for not outputting a seatbelt abnormal movement signal when a seatbelt adjustment operation is performed, the method comprising: Step S850: Based on the seat belt adjustment operation, record the seat belt adjustment time range.

[0088] The seatbelt adjustment time range refers to the continuous time interval from when the driver first touches the seatbelt until the driver stops moving the seatbelt. It is used to lock the period of driver hand activity that needs to be analyzed.

[0089] Step S851: Obtain driver's hand activity video information by analyzing the driver's video information through the seat belt adjustment time range.

[0090] Driver hand movement video information refers to video segments extracted from driver video information that perfectly match the seat belt adjustment time range.

[0091] Step S852: Determine the driver's hand movement trend based on the video information of the driver's hand movements.

[0092] The driver's hand movement trend refers to the characteristics of the direction, trajectory and amplitude of hand movement obtained after analyzing the video information of the driver's hand activities through image frame difference algorithm and contour fitting algorithm, such as stroking up and down along the seat belt webbing.

[0093] Step S853: If the driver's hand movement tends to fall into the seat belt slippage area, confirm that the driver is adjusting the seat belt.

[0094] If the driver's hand movement tends to fall into the seat belt slip position area, it means that the driver's hand movement trajectory is completely concentrated in the seat belt slip position area. The purpose of the action is to actively adjust the seat belt position to relieve the feeling of restraint, rather than accidental touch or other irrelevant actions.

[0095] Step S854: If the driver is adjusting the seat belt, output an abnormal seat belt movement signal.

[0096] If the driver adjusts the seatbelt, it indicates that the driver has actively adjusted the seatbelt, possibly because the seatbelt is too tight and affects the driver's comfort. In this case, an abnormal seatbelt movement signal is output, allowing the system to further process the information and improve the driver's experience.

[0097] Step S855: If there is no driver adjusting the seat belt, no abnormal seat belt movement signal will be output.

[0098] If there is no driver adjusting the seat belt, it means that the driver's hand movement is not for adjusting the seat belt, but may be an accidental touch or other action unrelated to seat belt adjustment. In this case, no abnormal seat belt movement signal will be output to avoid unnecessary processing by the system.

[0099] This also includes a method for updating the driver's habitual hand position, which includes: Step S856: After performing the seat belt buckle sliding operation, determine the target position of the seat belt sliding.

[0100] The target position for seat belt sliding refers to the physical position where the seat belt buckle will finally stop after the seat belt buckle sliding operation is completed. This position corresponds precisely to the position of the driver's hand when opening the door.

[0101] Step S857: Form a current seat belt sliding target position queue based on the seat belt sliding target position.

[0102] The current seatbelt slip target position queue refers to a fixed-length dataset constructed using the first-in-first-out (FIFO) rule. It stores the seatbelt slip target positions of the most recent N times (N is a preset value, such as 20 times) in chronological order. This is used to accumulate long-term driver operation data and filter out accidental position deviations.

[0103] Step S858: Determine the maximum number of seat belt consistent target positions by using the seat belt sliding target position queue and seat belt sliding target position.

[0104] The maximum number of seat belt consistent target positions refers to the total number of positions in the current seat belt sliding target position queue whose deviation from the current seat belt sliding target position falls within the preset error range, where the error range is ±2cm.

[0105] Step S859: Obtain the length of the seatbelt sliding target position queue.

[0106] The seatbelt slip target position queue length refers to the maximum number of samples (i.e., N value) preset in the queue. It is a fixed denominator used to calculate the proportion. For example, setting it to 20 means that the driver's habits are judged only based on the data from the most recent 20 operations.

[0107] Step S860: Determine the percentage of the number of consistent target positions for seat belts based on the maximum number of consistent target positions for seat belts and the length of the seat belt sliding target position queue.

[0108] The percentage of seatbelt consistent target positions refers to the ratio of the maximum number of seatbelt consistent target positions to the length of the seatbelt sliding target position queue. The calculation formula is: the percentage of seatbelt consistent target positions equals the maximum number of seatbelt consistent target positions divided by the length of the seatbelt sliding target position queue × 100%.

[0109] Step S861: If the proportion of the number of consistent target positions of the maximum seat belt exceeds the preset threshold for the proportion of reliable seat belt target positions, update the driver's hand habit position.

[0110] The reliable seat belt target position percentage threshold refers to the minimum percentage value that is pre-set by the driver to determine that the driver's operating habits have stabilized.

[0111] If the percentage of the number of consistent target positions for the maximum seat belt exceeds the threshold for the percentage of reliable seat belt target positions, it indicates that the seat belt buckle positioning position corresponding to the driver's hand opening the door has formed a stable pattern in the driver's recent exit operations. At this time, the target position for seat belt sliding can be updated to the driver's habitual hand position, and the subsequent system can directly perform sliding operations based on this position.

[0112] This also includes: Step S870: Collect the current driver's back pressure value.

[0113] The current driver back pressure value refers to the pressure value of each contact point on the driver's back that is collected in real time by the distributed thin-film pressure sensor array integrated into the seat back. The unit is Newton (N), and each sensor unit corresponds to an independent pressure collection point.

[0114] Step S871: Generate a driver back pressure value queue based on the current driver back pressure value.

[0115] The driver back pressure value queue refers to a dataset of driver back pressure values ​​stored in chronological order within the most recent preset time period (e.g., 30 seconds).

[0116] Step S872: Form the current driver baseline pressure value based on the driver back pressure value queue.

[0117] The current driver baseline pressure value is the pressure value obtained by taking the median of all data in the driver's back pressure value queue. It is a reference benchmark for determining whether back pressure has increased or decreased.

[0118] Step S873: If the current driver's back pressure value is less than the current driver's reference pressure value, determine the area of ​​reduced back pressure.

[0119] The back pressure drop area refers to a sheet-like area on the seat back where the pressure values ​​collected by multiple consecutive sensor units are all lower than the current driver's reference pressure value, and it is composed of adjacent sensor units.

[0120] If the current driver's back pressure value is less than the current driver's reference pressure value, it means that the degree of fit between the driver's back and the seat back is reduced, or even slightly detached from the backrest. This is the physical basis for judging changes in body posture (such as turning or getting up).

[0121] Step S874: If the current driver's back pressure value is greater than the current driver's reference pressure value, determine the area of ​​increased back pressure.

[0122] The area of ​​increased back pressure refers to a sheet-like area on the seat back where multiple sensor units collect pressure values ​​that are higher than the current driver's baseline pressure value. It is often symmetrically distributed with the area of ​​decreased pressure.

[0123] If the current driver's back pressure value is greater than the current driver's baseline pressure value, it means that the pressure between the driver's back and the seat back in that area has increased, which is the physical basis for determining the shift of the body's center of gravity towards that area.

[0124] Step S875: Determine the force distribution relationship in the seat area by identifying the areas of increased back pressure and decreased back pressure.

[0125] The force position relationship of the seat area refers to the spatial distribution relationship between the pressure rise area and the pressure fall area in the plane of the seat back. It includes only two core types: left-right symmetry relationship and up-down distribution relationship, which are used to determine the specific posture changes of the driver.

[0126] Step S876: If the force position relationship of the seat area is vertical, determine the vertical relative position distance.

[0127] The vertical relative position distance refers to the straight-line distance between the geometric center of the pressure rise area and the geometric center of the pressure fall area in the vertical direction of the seat back.

[0128] If the force distribution in the seat area is vertical, it indicates that the change in pressure on the driver's back is concentrated in the vertical direction, most likely due to a posture adjustment of leaning forward or backward.

[0129] Step S877: Determine the current seatbelt pressure adjustment parameters based on the vertical relative position distance.

[0130] The current seat belt pressure adjustment parameters refer to the seat belt retractor action parameters, including webbing release length and pressure reduction range. The greater the distance, the greater the release length or pressure reduction range.

[0131] Step S878: If the force position relationship of the seat area is left-right, determine the relative distance between the left and right positions.

[0132] The left-right relative position distance refers to the straight-line distance between the geometric center of the pressure rise area and the geometric center of the pressure fall area in the horizontal direction of the seat back.

[0133] If the force distribution in the seat area is left-right, it indicates that the driver's back pressure changes are concentrated in the horizontal direction, clearly indicating a body turning action to the left or right, which closely matches the sideways posture when getting out of the car.

[0134] Step S879: Determine the current seatbelt pressure adjustment parameters based on the left and right relative position distances.

[0135] The current seat belt pressure adjustment parameters refer to the seat belt retractor action parameters calculated based on the distance between the left and right relative positions. The greater the steering distance, the greater the seat belt release length or decompression range, in order to adapt to the seat belt restraint needs when the driver is in a side position.

[0136] Step S880: Perform seat belt pressure parameter adjustment operation according to the current seat belt pressure adjustment parameters.

[0137] Optional, also includes: Step S881: Obtain the current vehicle speed in real time.

[0138] Current vehicle speed refers to the speed value recorded in real time during vehicle operation, which can be accurately collected through the vehicle's speed sensor.

[0139] Step S882: If the current vehicle speed is not 0, do not perform the seat belt pressure parameter adjustment operation.

[0140] If the current vehicle speed is not 0, it indicates that the vehicle is in motion. While in motion, the system should maintain normal seatbelt pressure to ensure driver safety and avoid increasing the risk of accidents due to arbitrary adjustments to seatbelt pressure. Therefore, seatbelt pressure parameter adjustment operations will not be performed.

[0141] This also includes a method for adjusting seatbelt pressure parameters during the seatbelt buckle sliding operation, the method comprising: Step S883: When performing the seat belt buckle sliding operation, obtain the buckle movement speed.

[0142] The buckle movement speed refers to the real-time linear velocity of the seat belt buckle as it slides along a preset C-shaped track. This velocity is collected and calculated in real time by the buckle's built-in position sensor, reflecting the smoothness of the buckle's movement.

[0143] Step S884: When the moving speed of the buckle falls within the preset stationary speed range, output the clothing friction signal.

[0144] The stationary speed range refers to a low-speed range that is pre-set by the user to indicate that the movement of the buckle is obstructed and close to a standstill. It is used to distinguish between normal buckle movement, mechanical malfunction jamming, and clothing friction jamming.

[0145] The clothing friction signal refers to the electrical command signal output by the vehicle ECU to the seat belt pressure control module, which is used to indicate that the buckle is obstructed due to clothing friction, triggering the pressure relief auxiliary action.

[0146] When the buckle's moving speed falls within the range of stationary speed, it indicates that the seat belt buckle is experiencing significant resistance during the sliding process. This resistance is not caused by a mechanical malfunction, but most likely by the driver's clothing (such as a coat or sweater) being caught in the buckle or webbing, preventing the buckle from moving smoothly.

[0147] Step S885: In response to the clothing friction signal, perform seat belt pressure parameter adjustment operation according to the single pressure reduction parameter.

[0148] When a clothing friction signal is detected, it indicates that the system has identified the cause of the buckle slippage as clothing friction. At this time, by releasing part of the seat belt webbing and reducing the restraint pressure, the friction between the clothing and the webbing and buckle can be reduced, clearing the obstruction for the buckle to slide smoothly, while avoiding forced sliding that could cause the clothing to be pulled or damaged.

[0149] Step S886: After performing the seat belt pressure parameter reduction operation, perform the seat belt buckle sliding operation based on the seat belt buckle sliding parameter.

[0150] After performing the seatbelt pressure reduction operation, it means that the restraining pressure of the seatbelt on the driver has been reduced to a reasonable range, and the friction between the clothing and the webbing and buckle has been significantly reduced. At this time, restarting the sliding operation can ensure that the buckle moves smoothly to the target position along the preset trajectory and successfully completes the assisted unfastening action.

[0151] Based on the same inventive concept, embodiments of the present invention provide a seatbelt hand tracking system based on image recognition.

[0152] A seatbelt hand tracking system based on image recognition, comprising: The acquisition module is used to acquire driver video information and the driver's door direction; Memory for storing a program for a seatbelt hand tracking method based on image recognition; The processor loads and executes programs from memory.

[0153] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A seatbelt hand tracking method based on image recognition, characterized in that, include: Obtain driver video information; Analyze the driver's video information to determine the current number of the driver's hands; If the current number of driver's hands falls within the preset range of interfering hand numbers, the current driver's hand features are obtained; Determine the starting point of the driver's arm contour based on the current characteristics of the driver's hand; The driver's hand position is determined based on the starting point of the driver's arm contour; Get the driver's door direction; The driver's expected hand position for opening the door is determined based on the driver's door direction and the current position of the driver's hands; When the driver anticipates that their hand will fall into the preset seatbelt sliding position area when opening the door, a seatbelt buckle sliding signal is output. In response to the seatbelt buckle slip signal, the seatbelt buckle slip parameters are determined based on the driver's expected hand position when opening the door; The seatbelt buckle sliding operation is performed based on the seatbelt buckle sliding parameters.

2. The seatbelt hand tracking method based on image recognition according to claim 1, characterized in that, Methods for determining the driver's expected hand position for opening the door based on the driver's door direction and the current position of the driver's hand include: If the current driver's hand position falls within the preset steering wheel area, the driver's expected door opening hand position is determined by the driver's door direction and the preset driver's hand habit position; If the current driver's hand position is not within the steering wheel area, obtain the current driver's hand position off the steering wheel and the number of times the current driver's hand position is off the steering wheel; If the current number of times the driver's hands are off the steering wheel is the preset number of times both hands are off the steering wheel, the driver's expected hand position for opening the door is determined based on the direction of the driver's door. If the number of times the driver's hands are off the steering wheel is less than the number of times both hands are off the steering wheel, the driver's expected hand position for opening the door is determined based on the current position of the driver's hands off the steering wheel.

3. The seatbelt hand tracking method based on image recognition according to claim 1, characterized in that, It also includes a method for not outputting a seatbelt buckle slip signal when the driver anticipates that their hand will fall into the seatbelt slip position area when opening the door, the method comprising: Determine the seatbelt adjustment frequency when the driver anticipates that their hand will fall into the seatbelt slippage area when opening the door. When the seat belt adjustment frequency does not fall within the preset clothing adjustment frequency range, output a seat belt buckle slip signal; When the seat belt adjustment frequency falls within the range of clothing adjustment frequency, a driver clothing adjustment signal is output. In response to a signal indicating that the driver has adjusted their clothing, no signal indicating that the seatbelt buckle has slipped is output.

4. The seatbelt hand tracking method based on image recognition according to claim 3, characterized in that, It also includes a method for performing seatbelt pressure parameter adjustment operations, the method comprising: When a signal indicating that the driver has adjusted their clothing is received, the distance the seatbelt has moved is obtained. When the seat belt moves beyond the preset range for seat belt movement due to vehicle bumps, a seat belt adjustment operation is initiated. If a seatbelt adjustment operation is performed, an abnormal seatbelt movement signal will be output. In response to a seatbelt abnormal movement signal, the seatbelt adjustment frequency is obtained; If the seat belt adjustment frequency falls within the preset abnormal seat belt adjustment frequency range, the seat belt pressure parameter adjustment operation will be performed according to the preset single pressure reduction parameter.

5. The seatbelt hand tracking method based on image recognition according to claim 4, characterized in that, It also includes a method for not outputting a seatbelt abnormal movement signal when a seatbelt adjustment operation is performed, the method comprising: Record the time range of seat belt adjustment based on the seat belt adjustment operation; By analyzing the driver's video information within the seatbelt adjustment time range, video information of the driver's hand movements can be obtained. Determine the driver's hand movement trend based on video information of the driver's hand movements; If the driver's hand movement tends to fall into the seat belt slippage area, it indicates that the driver is adjusting the seat belt; If the driver adjusts the seat belt, an abnormal seat belt movement signal will be output; If there is no driver adjusting the seat belt, no abnormal seat belt movement signal will be output.

6. The seatbelt hand tracking method based on image recognition according to claim 5, characterized in that, It also includes a method for updating the driver's habitual hand position, which includes: After performing the seatbelt buckle sliding operation, determine the target position of the seatbelt sliding; A current seatbelt sliding target position queue is formed based on the seatbelt sliding target position; The maximum number of consistent target positions for seat belts is determined by the seat belt slip target position queue and seat belt slip target positions; Get the length of the queue for the seatbelt sliding target position; The proportion of the number of seat belt consistent target positions is determined based on the maximum number of seat belt consistent target positions and the length of the seat belt slippage target position queue. If the percentage of the number of consistent target positions for the maximum seat belt exceeds the preset threshold for the percentage of reliable seat belt target positions, the driver's hand position will be updated.

7. The seatbelt hand tracking method based on image recognition according to claim 4, characterized in that, Also includes: Collect the current back pressure value of the driver; A queue of driver back pressure values ​​is generated based on the current driver back pressure value. The current driver baseline pressure value is generated based on the driver back pressure value queue; If the current driver's back pressure value is lower than the current driver's reference pressure value, determine the area of ​​reduced back pressure. If the current driver's back pressure value is greater than the current driver's baseline pressure value, determine the area of ​​increased back pressure. The force distribution relationship in the seat area is determined by the areas of increased back pressure and decreased back pressure. If the force distribution in the seat area is vertical, determine the vertical distance between the two positions; The current seatbelt pressure adjustment parameters are determined based on the vertical relative position distance. If the force distribution in the seat area is left-right, determine the relative distance between the left and right positions; The current seatbelt pressure adjustment parameters are determined based on the relative distance between the left and right sides. Perform the seatbelt pressure adjustment operation according to the current seatbelt pressure adjustment parameters.

8. A seatbelt hand tracking method based on image recognition according to claim 7, characterized in that, Also includes: Get the current vehicle speed in real time; If the current vehicle speed is not 0, the seatbelt pressure parameter adjustment operation will not be performed.

9. A seatbelt hand tracking method based on image recognition according to claim 8, characterized in that, It also includes a method for adjusting seatbelt pressure parameters during the seatbelt buckle sliding operation, the method comprising: When performing the seatbelt buckle sliding operation, obtain the buckle movement speed; When the movement speed of the buckle falls within the preset stationary speed range, a clothing friction signal is output; In response to clothing friction signals, the seatbelt pressure parameter is adjusted according to the single pressure reduction parameter; After the seatbelt pressure parameter reduction operation is performed, the seatbelt buckle sliding operation is performed based on the seatbelt buckle sliding parameter.

10. A seatbelt hand tracking system based on image recognition, characterized in that, include: The acquisition module is used to acquire driver video information and the driver's door direction; A memory for storing a program of a seatbelt hand tracking method based on image recognition as described in any one of claims 1 to 9; The processor loads and executes programs from memory.

Images