An Adaptive Streaming Media Rearview Mirror Control Method and System

Abstract

This invention provides an adaptive streaming media rearview mirror control method and system. First, it detects the driver's head rotation angle in real time. Second, it sets an angle threshold and intelligently selects a field-of-view adjustment path based on the comparison result of || and . When the head rotation angle is large, a digital image cropping path is used. By calculating the image cropping offset and adjusting the cropping window position, the field of view is quickly expanded. When the head rotation angle is small, a physical gimbal following path is used. By calculating the gimbal target rotation angle and driving the camera to rotate, the true field of view is captured with lossless image quality. This achieves real-time tracking of the driver's head rotation by the rearview mirror, eliminating blind spots caused by line-of-view deviation. Simultaneously, it avoids information redundancy caused by a fixed large field-of-view image, reduces the driver's cognitive load, and significantly improves the convenience and safety of driving observation.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving assistance system technology, and in particular to an adaptive streaming media rearview mirror control method and system. Background Technology

[0002] Streaming rearview mirrors use cameras to capture images of the area behind the vehicle and display them in real time on an in-vehicle screen, replacing traditional optical reflective lenses. This technology effectively eliminates blind spots caused by vehicle structure and passenger obstruction, and significantly improves visibility in adverse conditions such as nighttime, rain, and snow, making it an important technological direction for enhancing driving safety.

[0003] Currently, the mainstream technical approach to improving the field of view of streaming rearview mirrors mainly focuses on expanding the static physical field of view of the camera. Specifically, this manifests in two main solutions: one is to use a single wide-angle or fisheye camera with a large field of view (e.g., a field of view exceeding 150°), collect images, perform distortion correction processing, and then display the panoramic view fixedly on the screen; the other is to deploy multiple cameras on both sides and the rear of the vehicle, and generate an electronic panoramic surround view through image stitching algorithms.

[0004] Both of the above solutions display pre-set static images based on hardware configurations, completely disconnected from the driver's real-time observation intentions and head posture. Regardless of the width of the static image's field of view, when observing different directions (such as vehicles approaching from the side and rear, or pedestrians on the roadside), the driver still needs to actively search for targets within the fixed display, making the interaction logic essentially no different from traditional fixed-lens displays. This not only fails to fully utilize the technological advantage of electronic display systems that can dynamically reconstruct images, but the fixed, ultra-wide field of view, containing a large amount of irrelevant information, may actually increase the driver's cognitive load and distract their attention. Furthermore, to achieve the expansion of the static field of view, the above solutions generally rely on stacking high-cost hardware, resulting in high overall system costs and hindering widespread application in low- to mid-range vehicles.

[0005] In view of this, the inventor has specifically designed an adaptive streaming media rearview mirror control method and system, which leads to this invention. Summary of the Invention

[0006] To solve the above problems, the technical solution of the present invention is as follows: An adaptive streaming media rearview mirror control method includes the following steps: Real-time detection of the driver's head tilt angle relative to the vehicle's forward direction. ; Get the preset angle threshold Comparison | |and Size; Select the field of view adjustment path based on the comparison results: If | |> The field of view is adjusted by using digital image cropping paths, based on the head rotation angle. Calculate the image cropping offset, adjust the position of the video stream cropping window, and generate a new field of view; If | |≤ The physical gimbal is then used to adjust the field of view by following the path and adjusting according to the head's tilt angle. Calculate the target rotation angle of the gimbal Control the gimbal to rotate the camera to the target angle and capture a new field of view; The adjusted field of view is displayed in real time on the streaming rearview mirror.

[0007] Preferably, the image cropping offset is as follows: The physical offset for image cropping of the left and right rearview mirrors is calculated using a predefined piecewise linear mapping function; the piecewise linear mapping function consists of multiple calibrated discrete calibration points ( , The definition is given for any input angle. : if ≤ ,but = ; if ≥ ,but = ; If there exists k such that < < Then, it is calculated using linear interpolation: in, For the head tilt angle, This corresponds to the image cropping offset. Based on the pixel density of the display area, the physical offset is converted into a pixel offset using the following formula: in, To display the horizontal number of pixels in the area, To display the physical width of the area, This is the pixel offset; According to the pixel offset Determine the new position of the cropping window.

[0008] Preferably, it also includes security and optimization, which includes at least one of the following: Dead zone detection, setting an angle dead zone threshold. When | |< At this time, any field of view adjustment is prohibited, among which < ; Smoothing filtering is applied to the calculated field of view adjustment amount using first-order hysteresis filtering: current offset = α × newly calculated offset + (1-α) × previous frame offset, where α is the smoothing coefficient, and the value of α ranges from 0.1 to 0.3.

[0009] Preferably, the calculation of the gimbal target rotation angle Specifically, a mapping function with nonlinear saturation characteristics is used: in, This is the sensitivity coefficient. This is the scaling factor.

[0010] Preferably, controlling the rotation of the physical gimbal includes: Get the current actual angle of the gimbal ; Calculate the error between the target angle and the actual angle. ; The control quantity u is calculated using a proportional, integral, and derivative control algorithm: in , , These are the pre-tuned controller parameters; The control quantity u is converted into a pulse signal to drive the gimbal to move accurately to the target angle. .

[0011] Preferably, the detected deflection angle include: Acquire driver's facial image and locate the two-dimensional pixel coordinates of key facial feature points; A three-dimensional rigid head model is established, the model containing three-dimensional spatial coordinates corresponding to two-dimensional feature points; The rotation matrix of the head coordinate system relative to the camera coordinate system is solved using the perspective n-point algorithm; The rotation matrix is ​​converted into Euler angles to obtain the yaw angle of the head, which is then used as the yaw angle. .

[0012] This solution also provides an adaptive streaming media rearview mirror system, including: The head posture detection module is used to obtain the angle of the driver's head relative to the forward direction of the vehicle in real time. ; The mapping calculation module is connected to the head pose detection module and has a preset angle threshold. It also includes an embedded mapping relationship between head angle and field of vision adjustment. An execution module is connected to the mapping calculation module, and the execution module includes a digital image cropping unit and a physical gimbal following unit; The mapping calculation module is based on the current head deflection angle. With angle threshold Based on the comparison results, the physical gimbal following unit or the digital image cropping unit may be selectively activated: When | |≤ When the physical gimbal tracking unit is activated, the new field of view is directly captured by driving the camera to rotate. When | |> When the digital image cropping unit is activated, the display field of view is changed by adjusting the position of the video stream cropping window.

[0013] Preferably, the digital image cropping unit includes: The unit conversion subunit is used to convert the physical offset Δ into a pixel offset P according to the pixel density of the display area; The cropping subunit is used to dynamically adjust the starting coordinates of the video stream cropping window according to the pixel offset P; The display subunit is used to scale and display the cropped image in real time.

[0014] Preferably, the physical gimbal following unit includes: The gimbal mechanism provides horizontal rotational freedom; A drive motor is used to drive the gimbal mechanism to rotate; Angle sensor is used to provide real-time feedback on the actual rotation angle of the gimbal. ; The control subunit is used to determine the target angle. The actual rotation angle fed back by the angle sensor The error e is used to generate a drive signal to control the movement of the drive motor.

[0015] Preferably, it also includes a security and optimization module, which includes: The dead-zone detection submodule is used when | | Less than the preset dead zone threshold At this time, any field of view adjustment is prohibited; The smoothing filter submodule is used to perform first-order hysteresis filtering on the field of view adjustment. The boundary constraint submodule is used to ensure that the cropping window does not exceed the valid area of ​​the original image, and that the gimbal target angle does not exceed the mechanical limit range.

[0016] The technical solution provided by this invention has the following beneficial effects: This invention first detects the driver's head tilt angle in real time. Secondly, set an angle threshold. According to | |and The system intelligently selects the appropriate field-of-view adjustment path based on the comparison results. When the head rotation angle is large, a digital image cropping path is used. By calculating the image cropping offset and adjusting the cropping window position, the field of view is quickly expanded. When the head rotation angle is small, a physical gimbal following path is used. By calculating the gimbal target angle and driving the camera to rotate, the system captures the real field of view with lossless image quality. This achieves real-time tracking of the driver's head rotation by the rearview mirror, eliminating blind spots caused by line-of-sight deviation. The dual-path design balances response speed and image fidelity. At small angles, the gimbal provides a realistic image, while at large angles, cropping enables rapid expansion, with both approaches complementing each other. Simultaneously, it avoids information redundancy caused by a fixed large field of view, reducing the driver's cognitive load and significantly improving the convenience and safety of driving observation. Attached Figure Description

[0017] The accompanying drawings, which are provided to further illustrate the invention and constitute a part of this invention, are illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention.

[0018] in: Figure 1 This is a flowchart illustrating the implementation of this invention; Figure 2 This is a flowchart illustrating the implementation process of digital image cropping in this invention; Figure 3 This is a flowchart illustrating the implementation of physical gimbal following in this invention; Figure 4 This is a structural module diagram of the rearview mirror system in this invention. Detailed Implementation

[0019] To make the technical problems, solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of the invention.

[0020] Please refer to the details. Figures 1-4The adaptive streaming media rearview mirror control method provided by the present invention includes the following steps: Real-time detection of the driver's head tilt angle relative to the vehicle's forward direction. The system captures facial images of the driver using an in-vehicle camera and employs a pre-trained neural network model to calculate the yaw angle of the head relative to the vehicle's forward direction in real time. The raw angle data is low-pass filtered to eliminate high-frequency jitter;

[0021] Determine the current If the absolute value is greater than the preset dead zone threshold, skip the adjustment and keep the field of view stable. Get the preset angle threshold Comparison | |and Size; where The value range is 15° to 20°. Select the field of view adjustment path based on the comparison results: If | |> The field of view is adjusted by using digital image cropping paths, based on the head rotation angle. Calculate the image cropping offset, adjust the position of the video stream cropping window, and generate a new field of view; If | |≤ The physical gimbal is then used to adjust the field of view by following the path and adjusting according to the head's tilt angle. Calculate the target rotation angle of the gimbal Control the gimbal to rotate the camera to the target angle and capture a new field of view; The adjusted field of view is displayed in real time on the streaming rearview mirror. For the cropping path, ensure that the cropping window does not exceed the boundary of the effective image area; for the gimbal path, ensure that the command angle does not exceed the hardware mechanical limit. At the same time, a first-order hysteresis smoothing filter is applied to the final output adjustment amount (P or φ) to make the field of view change continuous without jumps.

[0022] In this invention, the rearview mirror's field of view can follow the driver's head movement in real time, fundamentally eliminating blind spots caused by line-of-sight deviation. The dual-path design fully leverages the advantages of each path: the gimbal provides optically accurate images within a small angle range, ensuring lossless image quality, while digital cropping enables rapid field of view expansion within a large angle range, avoiding mechanical limitations. The synergistic effect of both paths makes field of view adjustment both precise and natural. At the same time, it avoids information redundancy caused by a fixed large field of view, significantly reducing the driver's cognitive load and improving the convenience and safety of driving observation.

[0023] For details, please refer to Figures 1-3 , The specific image cropping offset is as follows: The physical offset for image cropping of the left and right rearview mirrors is calculated using a predefined piecewise linear mapping function; the piecewise linear mapping function consists of multiple calibrated discrete calibration points ( , The definition is given for any input angle. : if ≤ ,but = ; if ≥ ,but = ; If there exists k such that < < Then, it is calculated using linear interpolation: in, For the head tilt angle, This corresponds to the image cropping offset. Based on the pixel density of the display area, the physical offset is converted into a pixel offset using the following formula: in, To display the horizontal number of pixels in the area, To display the physical width of the area, This is the pixel offset; Based on pixel offset Determine the new position of the cropping window. First, determine the reference position of the cropping window. The default initial position displays the center area of ​​the original image. Then, calculate the new position of the cropping window. The calculation methods for the left and right rearview mirrors are as follows:

[0024] The starting horizontal coordinate for cutting = reference position + P Finally, based on the calculated cropping window position, the corresponding region is extracted from the original video stream, the extracted image is scaled to the display area size, and the display content is updated in real time.

[0025] Mapping function of the left rearview mirror For example, its calibration point set is {( )|i=1,2,...,n}, where This is the angle at which the head turns to the left. The corresponding physical offset for image cropping is calculated. For any input angle value x, the function value is calculated according to the following rules:

[0026] If x≤ ,but ( )= ; If x≥ ,but ( )= ; If there exists k such that <x< Then, it is calculated using linear interpolation: Mapping function of the right rearview mirror It is implemented using the exact same piecewise linear interpolation mechanism, but its calibration point set is independent to adapt to the different optical characteristics and field of view requirements of the right rearview mirror.

[0027] For details, please refer to Figures 1-4 Safety and optimization steps are implemented before calculations are performed, and an angle dead zone threshold is set to reduce false triggering. When | |< At this time, any field of view adjustment is prohibited, among which < ; typical Take 5°-10°; Secondly, to avoid screen jumps, a smoothing filter is used: the calculated field adjustment amount is subjected to a first-order hysteresis filter: current offset = α × newly calculated offset + (1-α) × previous frame offset, where α is the smoothing coefficient (0<α<1), and the coefficient α directly controls the trade-off between system response speed and smoothness.

[0028] When α approaches 1, the filtering effect is weak and the system response is fast, but it may not be able to effectively suppress the image jitter caused by sudden changes.

[0029] When α approaches 0, the filtering effect is strong and the output is very smooth, but it introduces a significant visual tracking delay, resulting in a "sluggish" operating experience.

[0030] Therefore, the preferred range for α is typically between 0.1 and 0.3. This range is determined through systematic testing: within this range, the system can significantly smooth the visuals (eliminating abrupt transitions) while keeping the visual feedback latency within 100-300 milliseconds. This latency falls within the comfortable range of human visual-motion perception and will not cause discomfort. For example, at (α = 0.2), the system's response setup time to a step input (reaching 95% of the target value) is approximately several frames, and at a typical 60Hz refresh rate, the latency is not noticeable.

[0031] For details, please refer to Figures 1-4 When the physical gimbal adjusts the field of view along the following path, the driver's head yaw angle θ is acquired in real time. After filtering, it is denoted as... .

[0032] Establish head tilt angle Angle of rotation of the gimbal Nonlinear mapping relationship: Nonlinear saturation characteristic: gimbal rotation angle The absolute value is always less than or equal to the head deflection angle. The absolute value, i.e., | |≤| |

[0033] High gain at small angles, low gain at large angles: When the head deflects at a small angle (e.g., | (|<20°), the mapping function has a high scaling factor, making the gimbal responsive; when the angle of deflection is large, the scaling factor decreases, making the rotation smooth and gradually approaching the preset mechanical limit value.

[0034] Defined using arctangent-class saturation functions For example, it can be defined as:

[0035] in This is the sensitivity coefficient. This is the scaling factor used to match the output range to the actual physical rotation range of the gimbal.

[0036] The control module calculates the target angle. The actual angle fed back by the angle sensor This generates drive signals. Specifically, it includes:

[0037] Error calculation: .

[0038] Control law calculation: The control quantity u is calculated using the proportional-integral-derivative control algorithm. in , , These are the pre-tuned controller parameters.

[0039] The control quantity u is converted into a pulse or voltage signal to drive the motor, thereby driving the gimbal to move accurately to the target angle. .

[0040] The footage captured by the rotating camera is displayed in real time on the streaming rearview mirror; The rotation range of the gimbal used in this solution should be limited by physical limiters, software limiters, and an emergency stop mechanism. These respectively implement three functions: preventing mechanical damage, ensuring a reasonable field of view, and stopping immediately when an anomaly is detected.

[0041] This invention systematically improves the static field-of-view mechanism of traditional streaming rearview mirrors, providing two parallel technical paths to achieve adaptive dynamic adjustment of the field of view: a digital image cropping scheme and a physical gimbal following scheme. Both establish a high-precision mapping relationship between the driver's head deflection angle and the adjustment amount of the rearview mirror's field of view, achieving real-time coordination between the rearview mirror's field of view and the driver's visual intent, thereby fundamentally solving the blind spots and observation burden problems caused by a fixed field of view.

[0042] Specifically, the digital image cropping solution is implemented through software algorithms, offering advantages such as low cost, fast response, and no mechanical wear. The system uses a predefined piecewise linear mapping function to accurately convert the head angle into pixel offsets for image cropping, and performs independent mapping and interpolation calculations for the left and right rearview mirrors, ensuring that the field of view adjustment conforms to human visual habits and the optical characteristics of different mirror surfaces. Simultaneously, the introduced dead-zone mechanism and smoothing filtering mechanism guarantee the continuity and safety of the field of view output.

[0043] The physical gimbal tracking solution, implemented through hardware mechanisms, provides a more intuitive and realistic viewing experience. This solution drives the gimbal to physically rotate the rear-view camera, directly capturing the real scene in the target direction, thus avoiding resolution loss that may result from image cropping.

[0044] When switching between the physical pan-tilt-zoom (PTZ) follow path and the digital image cropping path, a smooth transition processing step is also included, which employs at least one of the following methods: When the angle threshold is reached Previously, the cropping window was moved towards the target direction by a preset offset; When the angle threshold is reached Previously, the gimbal was rotated to a preset angle in the target direction; The amount of field of view adjustment during the switching process is gradually weighted and blended.

[0045] In this invention, when performing head pose detection, the first step is to locate and model facial feature points. Feature point localization involves acquiring the coordinates of several key two-dimensional pixels in the face region of the image. These key points typically include points with clear anatomical significance, such as the tip of the nose, the inner and outer corners of the left and right eyes, the corners of the mouth, and the center of the chin. Next, a three-dimensional head model is established. A general or calibrated three-dimensional rigid head model is predefined. This model contains three-dimensional spatial coordinates corresponding one-to-one with the aforementioned two-dimensional feature points. Its origin is usually set at the center of the head or near the root of the nose, forming a coordinate system centered on the head. Then, spatial pose parameters are calculated and mapping relationships are established.

[0046] Since solving head pose is essentially solving a 3D to 2D perspective projection problem, given a set of 3D spatial points (head model points), a set of corresponding 2D image projection points (detected feature points), and camera intrinsic parameters (such as focal length and principal point, obtained through camera calibration), the goal is to find the head rotation and translation parameters that best match the detected 2D points after the 3D points are projected onto the image plane. The core algorithm for pose determination is as follows:

[0047] This problem is solved using the perspective n-point algorithm. This algorithm directly calculates the rotation matrix (R) and translation vector (T) of the head coordinate system relative to the camera coordinate system by minimizing the reprojection error. This is a classic computer vision geometry problem with various mature numerical solutions (such as the direct linear transformation method and iterative optimization methods).

[0048] Finally, parameter transformation is performed to convert the obtained rotation matrix (R) into a more intuitive Euler angle representation, that is, the yaw angle (Yaw) is represented by the rotation of the head about the vertical axis (such as an axis passing through the top of the head). This invention also provides an adaptive streaming media rearview mirror system. The head posture detection module is used to obtain the angle of the driver's head relative to the forward direction of the vehicle in real time. ; The mapping calculation module is connected to the head pose detection module and has preset angle thresholds. It also includes an embedded mapping relationship between head angle and field of vision adjustment. The execution module is connected to the mapping calculation module. The execution module includes a digital image cropping unit and a physical gimbal following unit. The mapping calculation module is based on the current head deflection angle. With angle threshold Based on the comparison results, either the physical gimbal tracking unit or the digital image cropping unit can be selectively enabled: When | |≤ At this time, the physical gimbal tracking unit is activated, and the new field of view is directly captured by driving the camera to rotate; When | |> When this is done, the digital image cropping unit is activated, and the display field of view is changed by adjusting the position of the video stream cropping window.

[0049] The gimbal tracking unit includes: The gimbal mechanism provides horizontal rotational freedom; The drive motor is used to drive the gimbal mechanism to rotate; Angle sensor is used to provide real-time feedback on the actual rotation angle of the gimbal. ; The control subunit is used to determine the target angle. The actual rotation angle fed back by the angle sensor The error e is used to generate a drive signal to control the movement of the drive motor.

[0050] The digital image cropping unit includes: The unit conversion subunit is used to convert the physical offset Δ into a pixel offset P according to the pixel density of the display area; The cropping subunit is used to dynamically adjust the starting coordinates of the video stream cropping window based on the pixel offset P; The display subunit is used to scale and display the cropped image in real time.

[0051] The vehicle is set to cruise at high speed, with the driver's head tilted smoothly to the right by 30° to observe the road conditions to the right rear.

[0052] Example 1 (Digital Cropping Path): The system activates the cropping mapping of the right rearview mirror. The mapping table is consulted, assuming 30∘ is located between the calibration points (23.07∘, +1cm) and (32.50∘, −6cm). Through interpolation, the physical offset Δ≈−4.2cm is obtained. Assuming the screen parameter is 1cm = 100 pixels, the pixel offset instruction P = −420 pixels is generated. The image processing unit instantly moves the cropping center of the right rearview mirror image 420 pixels to the left. With almost no perceptible delay, the driver sees a wider field of view closer to the right side of the vehicle in the right rearview mirror, improving the lane change assist field of view coverage by approximately 40% compared to the fixed field of view.

[0053] Example 2 (Physical Gimbal Path): The system activates gimbal following. In a preferred embodiment, the head deflection angle... Angle of rotation of the gimbal mapping relationship for:

[0054] in This is the sensitivity coefficient. This is the scaling factor.

[0055] According to the nonlinear mapping model The calculation shows that the gimbal needs to rotate to the right. = 18°. The gimbal smoothly rotates to its position within 200ms, and the right rear-view camera directly captures the actual scene at an 18° angle to the right rear of the vehicle. The displayed image is an optically true image without any digital magnification loss, maintaining optimal image resolution and dynamic range. The measured display latency (from turning the head to the image stabilizing) is less than 250ms, meeting the persistence of vision requirements of the human eye.

[0056] In summary, this invention first detects the driver's head tilt angle in real time. Secondly, set an angle threshold. According to | |and The system intelligently selects the appropriate field-of-view adjustment path based on the comparison results. When the head rotation angle is large, a digital image cropping path is used. By calculating the image cropping offset and adjusting the cropping window position, the field of view is quickly expanded. When the head rotation angle is small, a physical gimbal following path is used. By calculating the gimbal target angle and driving the camera to rotate, the system captures the real field of view with lossless image quality. This achieves real-time tracking of the driver's head rotation by the rearview mirror, eliminating blind spots caused by line-of-sight deviation. The dual-path design balances response speed and image fidelity. At small angles, the gimbal provides a realistic image, while at large angles, cropping enables rapid expansion, with both approaches complementing each other. Simultaneously, it avoids information redundancy caused by a fixed large field of view, reducing the driver's cognitive load and significantly improving the convenience and safety of driving observation.

[0057] The present invention has been described above by way of example with reference to the accompanying drawings. Obviously, the specific implementation of the present invention is not limited to the above-described manner. Any non-substantial improvements made using the inventive concept and technical solution of the present invention, or the direct application of the inventive concept and technical solution of the present invention to other occasions without modification, are all within the protection scope of the present invention.

Claims

1. An adaptive streaming media rearview mirror control method, characterized in that, Includes the following steps: Real-time detection of the driver's head tilt angle relative to the vehicle's forward direction. ; Get the preset angle threshold Comparison | |and Size; Select the field of view adjustment path based on the comparison results: If | |> The field of view is adjusted by using digital image cropping paths, based on the head rotation angle. Calculate the image cropping offset, adjust the position of the video stream cropping window, and generate a new field of view; If | |≤ The view is then adjusted using a physical gimbal to follow the path and adjust according to the head's tilt angle. Calculate the target rotation angle of the gimbal Control the gimbal to rotate the camera to the target angle and capture a new field of view; The adjusted field of view is displayed in real time on the streaming rearview mirror.

2. The adaptive streaming media rearview mirror control method according to claim 1, characterized in that, The specific image cropping offset is as follows: The physical offset for image cropping of the left and right rearview mirrors is calculated using a predefined piecewise linear mapping function; the piecewise linear mapping function consists of multiple calibrated discrete calibration points ( , Defined as follows: for any input angle : if ≤ ,but = ; if ≥ ,but = ; If there exists k such that < < Then, it is calculated using linear interpolation: in, For the head tilt angle, This corresponds to the image cropping offset. Based on the pixel density of the display area, the physical offset is converted into a pixel offset using the following formula: in, To display the horizontal number of pixels in the area, To display the physical width of the area, This is the pixel offset; According to the pixel offset Determine the new position of the cropping window.

3. The adaptive streaming media rearview mirror control method according to claim 1, characterized in that, It also includes security and optimization, which includes at least one of the following: Dead zone detection, setting an angle dead zone threshold. When | | < At this time, any field of view adjustment is prohibited, among which < ; Smoothing filtering is applied to the calculated field of view adjustment amount using first-order hysteresis filtering: current offset = α × newly calculated offset + (1-α) × previous frame offset, where α is the smoothing coefficient, and the value of α ranges from 0.1 to 0.

3.

4. The adaptive streaming media rearview mirror control method according to claim 1, characterized in that, The calculation of the gimbal target rotation angle Specifically, a mapping function with nonlinear saturation characteristics is used: in, This is the sensitivity coefficient. This is the scaling factor.

5. The adaptive streaming media rearview mirror control method according to claim 1, characterized in that, The control of the physical gimbal rotation includes: Get the current actual angle of the gimbal ; Calculate the error between the target angle and the actual angle. ; The control quantity u is calculated using a proportional, integral, and derivative control algorithm: in , , These are the pre-tuned controller parameters; The control quantity u is converted into a pulse signal to drive the gimbal to move accurately to the target angle. .

6. The adaptive streaming media rearview mirror control method according to claim 1, characterized in that, The detected deflection angle include: Acquire driver's facial image and locate the two-dimensional pixel coordinates of key facial feature points; A three-dimensional rigid head model is established, the model containing three-dimensional spatial coordinates corresponding to two-dimensional feature points; The rotation matrix of the head coordinate system relative to the camera coordinate system is solved using the perspective n-point algorithm; The rotation matrix is ​​converted into Euler angles to obtain the yaw angle of the head, which is then used as the yaw angle. .

7. An adaptive streaming media rearview mirror system, based on the adaptive streaming media rearview mirror control method according to any one of claims 1 to 6, characterized in that, include: The head posture detection module is used to obtain the angle of the driver's head relative to the forward direction of the vehicle in real time. ; The mapping calculation module is connected to the head pose detection module and has a preset angle threshold. It also includes an embedded mapping relationship between head angle and field of vision adjustment. An execution module is connected to the mapping calculation module, and the execution module includes a digital image cropping unit and a physical gimbal following unit; The mapping calculation module is based on the current head deflection angle. With angle threshold Based on the comparison results, the physical gimbal following unit or the digital image cropping unit may be selectively activated: When | |≤ When the physical gimbal tracking unit is activated, the new field of view is directly captured by driving the camera to rotate; When | |> When the digital image cropping unit is activated, the display field of view is changed by adjusting the position of the video stream cropping window.

8. The adaptive streaming media rearview mirror system according to claim 7, characterized in that, The digital image cropping unit includes: The unit conversion subunit is used to convert the physical offset Δ into a pixel offset P according to the pixel density of the display area; The cropping subunit is used to dynamically adjust the starting coordinates of the video stream cropping window according to the pixel offset P; The display subunit is used to scale and display the cropped image in real time.

9. An adaptive streaming media rearview mirror system according to claim 7, characterized in that, The physical gimbal following unit includes: The gimbal mechanism provides horizontal rotational freedom; A drive motor is used to drive the gimbal mechanism to rotate; Angle sensor is used to provide real-time feedback on the actual rotation angle of the gimbal. ; The control subunit is used to determine the target angle. The actual rotation angle fed back by the angle sensor The error e is used to generate a drive signal to control the movement of the drive motor.

10. An adaptive streaming media rearview mirror system according to claim 7, characterized in that, It also includes a security and optimization module, which includes: The dead-zone detection submodule is used when | | Less than the preset dead zone threshold At this time, any field of view adjustment is prohibited; The smoothing filter submodule is used to perform first-order hysteresis filtering on the field of view adjustment. The boundary constraint submodule is used to ensure that the cropping window does not exceed the valid area of ​​the original image, and that the gimbal target angle does not exceed the mechanical limit range.

Images