Overhead working truck linkage emergency braking system based on multi-directional vision ranging

Abstract

The application discloses a kind of high-altitude working vehicle linkage emergency brake system based on multi-direction vision ranging, including multi-direction vision image acquisition module, vision color reconstruction module, contour fidelity vision enhancement module, binocular three-dimensional positioning module, vision ranging calibration module, high-altitude working vehicle motion envelope prediction module and linkage emergency brake module.The application belongs to the field of intelligent control, based on the vision color reconstruction of direction self-adaptive effective brightness range, improve the consistency of multidirectional binocular matching, provide stable input for subsequent obstacle distance calculation;Based on multiscale illumination estimation and contour fidelity calibration, improve the edge retention ability of obstacle, reduce disparity matching error;Based on the vision color reconstruction of direction self-adaptive effective brightness range, improve the consistency of multidirectional binocular matching;Based on multiscale illumination estimation and contour fidelity calibration, improve the edge retention ability of obstacle, reduce disparity matching error;Further reduce the reliability of final linkage emergency brake.

Description

Technical Field

[0001] This invention relates to the field of intelligent control, specifically to an emergency braking system for aerial work platforms based on multi-directional visual ranging. Background Technology

[0002] The aerial work platform's linked emergency braking system is a safety control system that senses the vehicle's surrounding environment and its own movement. When a potential collision or dangerous situation is detected, it automatically coordinates and controls the vehicle's movement, as well as the deceleration, stopping, or restriction of the movement of the work platform and boom. However, typical aerial work platform linked emergency braking systems suffer from problems such as fluctuating obstacle distances, inability to adapt to dynamically changing lighting environments, leading to false or missed emergency braking triggers; and inaccurate obstacle spatial positioning, making them unsuitable for complex working environments, resulting in low emergency braking safety. Summary of the Invention

[0003] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides a multi-directional visual ranging-based aerial work platform emergency braking system. Addressing the problems of conventional aerial work platform emergency braking systems, such as fluctuations in obstacle distance and inability to adapt to dynamically changing lighting environments in multiple directions, leading to false or missed emergency braking triggers, this solution improves the consistency of multi-directional binocular matching through visual color reconstruction based on directional adaptive effective brightness range, providing stable input for subsequent obstacle distance calculations. Furthermore, it enhances obstacle edge preservation capabilities and reduces disparity matching errors based on multi-scale illumination estimation and contour fidelity calibration, thereby reducing the final braking cost. This solution addresses the reliability of emergency braking in general aerial work platform vehicles, which suffer from inaccurate obstacle spatial positioning, inability to adapt to complex working environments, and consequently low emergency braking safety. It improves obstacle spatial positioning accuracy and reduces false braking and missed braking caused by single-camera errors through multi-directional 3D coordinate unification and distance measurement fusion. Visual feature learning is used to compensate for distance measurement deviations, improving the reliability of obstacle distance estimation. Motion envelope prediction enables early identification of dynamic collision zones. Finally, multi-directional risk calculation enables targeted emergency braking, enhancing the control safety of aerial work platforms during complex spatial operations.

[0004] The technical solution adopted by the present invention is as follows: The present invention provides a multi-directional visual ranging-based aerial work vehicle linkage emergency braking system, including a multi-directional visual image acquisition module, a visual color reconstruction module, a contour-fidelity visual enhancement module, a binocular three-dimensional positioning module, a visual ranging calibration module, an aerial work vehicle motion envelope prediction module, and a linkage emergency braking module.

[0005] The multi-directional visual image acquisition module acquires binocular images of the aerial work vehicle from different directions and simultaneously obtains motion status data.

[0006] The visual color reconstruction module takes multi-directional binocular images as input, determines the effective brightness range in each direction, dynamically adjusts the color channels, and outputs the color-reconstructed binocular image.

[0007] The contour-preserving visual enhancement module takes the reconstructed binocular image as input, and through multi-scale illumination estimation and structural information enhancement, outputs an enhanced binocular image that preserves the target edge and contour information.

[0008] The binocular 3D positioning module takes the enhanced binocular image as input, calculates the obstacle depth through binocular parallax, and performs coordinate transformation and distance fusion to obtain the fused obstacle distance.

[0009] The visual ranging calibration module takes enhanced binocular images as input, extracts target features and predicts distance errors through a lightweight visual model, calibrates the fused obstacle distances and outputs the results.

[0010] The aerial work platform motion envelope prediction module takes vehicle motion state data as input, predicts future motion trajectory based on vehicle, platform and boom motion parameters, and outputs the aerial work platform spatial motion envelope.

[0011] The linked emergency braking module takes the obstacle distance, spatial position, and motion envelope as input, calculates the collision risk, and generates linked braking control commands to realize linked emergency braking.

[0012] Furthermore, the visual color reconstruction module includes:

[0013] Gray-scale distribution statistics: For the input image, statistical histogram of channel pixel values;

[0014] Determine the effective visual range and calculate the cumulative distribution;

[0015] Set low-end and high-end response coefficients to determine the low and high effective grayscale boundaries respectively;

[0016] Dynamic channel adjustment is performed based on the effective pixel range, mapping the original grayscale values ​​to a unified effective dynamic range, and using effective low grayscale boundaries and effective high grayscale boundaries for adaptive channel mapping.

[0017] Furthermore, to avoid mixing brightness information from both sides of the building edge and causing edge blurring, the contour-fidelity visual enhancement module performs multi-scale illumination estimation and contour-fidelity calibration on the reconstructed images from different visual directions: it uses filtering at different spatial scales to obtain smooth illumination changes, and then uses the local structure of the brightness map itself to perform contour-fidelity calibration on the coarse illumination; including:

[0018] Multi-scale coarse illumination estimation converts the visually reconstructed images of each visual direction output by the visual color reconstruction module into luminance maps, sets the spatial scale, and calculates coarse illumination components for the spatial scale.

[0019] Filtered contour fidelity calibration establishes a coarse local linear relationship between illumination and brightness within a local window centered on a pixel and with a radius of r.

[0020] Structural information reconstruction involves logarithmic difference between the brightness map and the illumination components to extract local structural information that does not change synchronously with illumination changes, and to calculate structural information at each scale.

[0021] Brightness enhancement involves overlaying the reconstructed structural information onto the original brightness map to obtain enhanced brightness; the enhanced brightness map is then fused with the color information of the visual color reconstruction output image to obtain an enhanced binocular image.

[0022] Furthermore, the binocular three-dimensional positioning module includes:

[0023] Calculate the disparity and depth of the target points in the enhanced binocular image;

[0024] Calculate the target's three-dimensional coordinates and transform them to the vehicle coordinate system;

[0025] For the distance measurement results of the same obstacle in different directions, calculate the distance fluctuation in each direction; fuse the distance measurement variance to obtain the fused obstacle distance.

[0026] Furthermore, the visual ranging calibration module includes:

[0027] Set up a lightweight visual ranging calibration model, input the enhanced left and right images; use 4-level convolutional coding.

[0028] Set up two output branches:

[0029] The obstacle classification branch outputs the probability of the obstacle category.

[0030] The distance calibration branch uses supervised residual regression training with the actual distance error as the supervision target; the network output is combined with the binocular geometric ranging results to obtain the calibrated obstacle distance.

[0031] Furthermore, the linked emergency braking module includes:

[0032] The collision distance calculation for the moving area is performed by constructing the future occupied areas corresponding to the vehicle's walking direction, platform lifting direction, and boom movement direction based on the spatial relationship between the obstacle's spatial location and the future motion envelope of the aerial work platform. The effective collision distance in each direction is calculated based on the shortest distance between the obstacle's spatial location and the corresponding moving area. At the same time, the effective collision distance in each direction is constrained using the calibrated obstacle distance to obtain the constrained effective distance in each direction.

[0033] Collision time calculation involves calculating the relative motion of the obstacle and the vehicle in the future; and calculating the relative approach speed in the corresponding direction.

[0034] Directional risk calculation: Calculate the time required for the vehicle to complete braking; construct a directional risk response.

[0035] The system features coordinated control, defining directional risk conditions and generating braking commands for the vehicle, platform, and boom based on these risk conditions.

[0036] The beneficial effects achieved by the present invention using the above solution are as follows:

[0037] (1) In view of the problem that the general aerial work vehicle linkage emergency braking system is prone to obstacle distance fluctuations and cannot adapt to the dynamic changes in lighting environment in multiple directions, which leads to false triggering or missed triggering of emergency braking, this solution improves the consistency of multi-directional binocular matching by visual color reconstruction based on the effective brightness range of the direction, and provides a stable input for subsequent obstacle distance calculation; improves the obstacle edge preservation capability and reduces parallax matching error based on multi-scale lighting estimation and contour fidelity calibration; and thus reduces the reliability of the final linkage emergency braking.

[0038] (2) In view of the problem that the general aerial work vehicle linkage emergency braking system has inaccurate obstacle spatial positioning and cannot adapt to complex working environments, resulting in low emergency braking safety, this solution improves the accuracy of obstacle spatial positioning by unifying multi-directional three-dimensional coordinates and fusion of distance measurement, and reduces the false braking and missed braking caused by single camera error; it uses visual feature learning to compensate for distance measurement deviation and improves the reliability of obstacle distance estimation; it realizes the early judgment of dynamic collision area through motion envelope prediction; and finally realizes targeted emergency braking through multi-directional risk calculation; thus improving the control safety of aerial work vehicles in complex space operations. Attached Figure Description

[0039] Figure 1 This invention provides a flowchart of an aerial work platform linkage emergency braking system based on multi-directional visual ranging.

[0040] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0042] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0043] Example 1, see Figure 1 The present invention provides a multi-directional visual ranging-based aerial work vehicle linkage emergency braking system, including a multi-directional visual image acquisition module, a visual color reconstruction module, a contour-fidelity visual enhancement module, a binocular three-dimensional positioning module, a visual ranging calibration module, an aerial work vehicle motion envelope prediction module, and a linkage emergency braking module.

[0044] The multi-directional visual image acquisition module acquires binocular images of the aerial work vehicle from different directions and simultaneously obtains motion status data.

[0045] The visual color reconstruction module takes multi-directional binocular images as input, determines the effective brightness range in each direction, dynamically adjusts the color channels, and outputs the color-reconstructed binocular image.

[0046] The contour-preserving visual enhancement module takes the reconstructed binocular image as input, and through multi-scale illumination estimation and structural information enhancement, outputs an enhanced binocular image that preserves the target edge and contour information.

[0047] The binocular 3D positioning module takes the enhanced binocular image as input, calculates the obstacle depth through binocular parallax, and performs coordinate transformation and distance fusion to obtain the fused obstacle distance.

[0048] The visual ranging calibration module takes enhanced binocular images as input, extracts target features and predicts distance errors through a lightweight visual model, calibrates the fused obstacle distances and outputs the results.

[0049] The aerial work platform motion envelope prediction module takes vehicle motion state data as input, predicts future motion trajectory based on vehicle, platform and boom motion parameters, and outputs the aerial work platform spatial motion envelope.

[0050] The linked emergency braking module takes the obstacle distance, spatial position, and motion envelope as input, calculates the collision risk, and generates linked braking control commands to realize linked emergency braking.

[0051] Example 2, see Figure 1 This embodiment is based on the above embodiment. The multi-directional visual image acquisition module sets up visual acquisition units at the front, rear, left and right sides of the vehicle body, the working platform and the boom, respectively, according to the spatial structure and movement range of the aerial work vehicle body, the working platform and the boom. The field of view of each visual acquisition unit covers the vehicle's walking area, the platform's lifting area and the boom's swinging area. Each visual acquisition unit synchronously acquires binocular images in the corresponding direction and obtains the vehicle's current driving speed, longitudinal acceleration, lifting speed, boom attitude angle and angular velocity, forming vehicle motion state data that is synchronized with the visual images in time.

[0052] Example 3, see Figure 1 This embodiment, based on the above embodiment, addresses the differences in brightness distribution caused by direct sunlight, building reflections, metal structure reflections, and local shadows in different visual directions during outdoor operations of aerial work platforms. It performs pixel distribution statistics on the input images for each visual direction and determines the effective brightness range for that direction based on the proportion of pixels representing the main image element; including:

[0053] Gray-level distribution statistics for an input image with direction d and color channel c. Histogram of pixel values ​​for statistical channels: ; v is the number of pixels, and v is the grayscale value; It is a grayscale value; These are pixel coordinates; It is an indicator function that takes the value 1 if the condition is true, and 0 otherwise;

[0054] Determine the effective visual range and calculate the cumulative distribution: ; The grayscale value does not exceed The percentage of pixels; u is the grayscale index; It is the total number of pixels;

[0055] Set low-end response coefficient and high-end response coefficient Determine the low and high effective grayscale boundaries respectively:

[0056] ; ; and These are the effective low-grayscale boundary and effective high-grayscale boundary of the d-th direction and the c-th color channel, respectively; preferred settings , ∈[0.005,0.03], that is, discarding approximately 0.5% to 3% of extremely low grayscale and extremely high grayscale pixels respectively; this range is used to avoid extreme areas such as a small amount of sunlight reflection and local shadows determining the dynamic range of the entire image;

[0057] Dynamic channel adjustment is performed based on the effective pixel range, mapping the original grayscale values ​​to a unified effective dynamic range, and using effective low grayscale boundaries and effective high grayscale boundaries for adaptive channel mapping: ; These are the reconstructed grayscale values ​​of the color channels; This is the color channel gain coefficient, preferably between 0.9 and 1.1; It is the color channel offset coefficient, preferably -5 to 5; It is a smoothing term, preferably 10. −6 ~10 −3 Constraints on the reconstruction results: ; It is the grayscale value of the color channel after visual color reconstruction; It is a truncation function.

[0058] Example 4, see Figure 1 This embodiment is based on the above embodiment. To avoid mixing brightness information on both sides of the building edge, thus blurring the edge, the contour-fidelity visual enhancement module performs multi-scale illumination estimation and contour-fidelity calibration on the reconstructed images from different visual directions: smooth illumination changes are obtained using filtering at different spatial scales, and then contour-fidelity calibration is performed on the coarse illumination using the local structure of the brightness map itself; including:

[0059] Multi-scale coarse illumination estimation converts the reconstructed images of each visual direction output by the visual color reconstruction module into brightness maps. Set K spatial scales, and for the k-th spatial scale, calculate the coarse illumination component: ; It is the coarse illumination component of direction d at the k-th scale; It is a spatial scale; The standard deviation is A two-dimensional Gaussian kernel; small scales are used to describe local lighting changes, while large scales are used to describe overall lighting changes around the vehicle.

[0060] Filtered contour fidelity calibration establishes a coarse local linear relationship between illumination and brightness within a local window centered at pixel (x,y) with radius r. ; It is the illumination component after contour fidelity calibration; and These are local coefficients; ; ; It is the illumination component after contour fidelity calibration; It is covariance; It is variance; This is the regularization parameter, preferably 10. −4 ~10 −2 ; It is a localized, rough average of illumination; Therefore The average brightness of a local window centered on the center, with a radius of 4~8;

[0061] Structural information reconstruction involves logarithmic difference between the brightness map and the illumination components to extract local structural information that does not change synchronously with illumination variations, and calculation of structural information at various scales: ; It is the structural information at the k-th scale;

[0062] Brightness enhancement involves overlaying the reconstructed structural information onto the original brightness map, ensuring that building edges, boom outlines, and obstacle boundaries remain discernible under different lighting conditions, while avoiding over-enhancement that creates false edges, resulting in enhanced brightness. ; ; ; This is the enhanced brightness map; It is the structural reinforcement factor, preferably ranging from 0.5 to 2.0; It is the overall structural enhancement amount after fusing the d-th visual direction; It is the fusion weight of the k-th spatial scale; the enhanced brightness map is fused with the color information of the visual color reconstruction output image to obtain the enhanced binocular image.

[0063] By performing the above operations, this solution addresses the problem that general aerial work platform emergency braking systems are prone to obstacle distance fluctuations and cannot adapt to dynamically changing lighting environments in multiple directions, leading to false or missed emergency braking triggers. This solution improves the consistency of multi-directional binocular matching by using visual color reconstruction based on the effective brightness range of the direction-adaptive system, providing stable input for subsequent obstacle distance calculations. Furthermore, it enhances obstacle edge preservation capabilities and reduces parallax matching errors based on multi-scale lighting estimation and contour fidelity calibration, thereby reducing the reliability of the final emergency braking system.

[0064] Example 5, see Figure 1This embodiment is based on the above embodiment. The binocular 3D positioning module uses the enhanced binocular image output by the contour-fidelity visual enhancement module as input, performs intrinsic and extrinsic parameter calibration on the binocular visual acquisition units in each direction, and extracts the disparity of the same obstacle target point based on the enhanced left and right binocular images to calculate the obstacle depth and reconstruct the 3D coordinates. Subsequently, the 3D coordinates of the obstacle obtained from different directions are uniformly transformed to the aerial work platform vehicle coordinate system. When the same obstacle is detected simultaneously from multiple visual directions, weighted fusion is performed based on the fluctuation of the continuous ranging results in each direction to obtain a unified obstacle spatial position and fusion distance; including:

[0065] For direction d, the disparity of the target point in the enhanced binocular image is: The target depth is: ; It's parallax; and The x-coordinate of the target point in the enhanced binocular image; It refers to the camera's pixel focal length; It is the binocular baseline, in meters; It is the target depth;

[0066] The target's three-dimensional coordinates are: ; ; ; These are the coordinates of the target image; These are the coordinates of the camera's principal point; It refers to the position of the target in each coordinate system of the visual unit;

[0067] Transform to vehicle coordinate system: ; It is the rotation matrix from the visual unit to the vehicle coordinate system; It is the translation vector of the d-th visual unit relative to the vehicle coordinate system, in meters; It refers to the spatial location of the obstacle;

[0068] For distance measurement results from multiple directions for the same obstacle, calculate the distance fluctuation in the d-th direction: ; This is the ranging value of the i-th frame, in meters. It is the average distance across M consecutive frames, in meters; This is the standard deviation of the distance measurement, in meters; it is fused based on the variance of the distance measurement. ; This is the distance to the merged obstacles, in meters. It is the set of visual directions from which obstacles are observed; It is the smoothing distance, set to 10. −6 ~10 − 4 m2 .

[0069] Example 6, see Figure 1 This embodiment is based on the above embodiment. The visual ranging calibration module takes the enhanced binocular image as input, sets a lightweight visual model, extracts features from the visually enhanced binocular image, and uses this feature to identify obstacle targets and calibrate the initial obstacle distance to correct the depth measurement deviation caused by binocular parallax error; including:

[0070] Lightweight vision ranging calibration model, input ; These are the enhanced left and right images, respectively.

[0071] The network structure, which employs 4-level convolutional coding, is shown in Table 1.

[0072]

[0073] Feature extraction at each level is as follows: ; This is the i-th level feature, with the initial feature being X; Conv1 has 2 input channels and performs joint encoding on the stereo image;

[0074] Set up two output branches:

[0075] The obstacle classification branch outputs the probability of the obstacle category. ; It is the category probability. and These are classification branch parameters; It is a level 4 feature; It is global average pooling;

[0076] The distance calibration branch is trained using supervised residual regression, with the true distance error as the supervision target. Combine the network output with the binocular geometric ranging results: ; It is the distance residual compensation amount predicted by the network, in meters; and These are distance calibration branch parameters; This is the calibrated obstacle distance, in meters (m).

[0077] The network training function uses classification loss plus distance residual loss: ; These are the loss weights, with values ​​ranging from 0 to 1; L is the network training function.

[0078] Classification loss ;

[0079] Distance residual loss: N is the number of training samples; C is the number of obstacle categories; These are real category labels; It predicts the class probability; This represents the true depth; the distance residual is standardized to convert the distance loss into a dimensionless form for use in network training loss; N is the total number of samples, and i is the sample index; It represents the obstacle distance for the i-th sample; a sample is a set of enhanced stereo images;

[0080] To avoid the numerical scale difference between distance loss and classification loss, the distance residual is standardized before training and then reconstructed after the network training is completed.

[0081] Example 7, see Figure 1 This embodiment is based on the above embodiment. The aerial work platform vehicle motion envelope prediction module uses the vehicle motion state data obtained by the multi-directional visual image acquisition module as input to predict the future motion envelope of the aerial work platform vehicle, and the vehicle's longitudinal displacement: Vertical displacement of the platform: Boom angle change: ; This is the longitudinal displacement of the vehicle, measured in meters (m). It is the predicted time, in seconds; This is the current longitudinal speed of the vehicle, in m / s; This is the current longitudinal acceleration of the vehicle, measured in m / s². 2 ; This is the vertical displacement of the platform, measured in meters (m). This is the current vertical velocity of the platform, in m / s; This is the current vertical acceleration of the platform, in m / s². 2 ; It is the boom angle at the predicted moment, in rad; This is the current boom angle, in rad. This is the current boom angular velocity, in rad / s; This is the current boom angular acceleration, in rad / s². 2 ;

[0082] Based on vehicle dimensions and boom geometry, the vehicle body, platform, and boom are transformed into a time-varying spatial envelope: ; This is the future motion envelope of the aerial work platform vehicle, measured in meters (m). 3 ; It is a geometric mapping function that generates a spatial envelope based on the motion state and vehicle geometry.

[0083] Example 8, see Figure 1This embodiment is based on the above embodiment. The linkage emergency braking module calculates the collision time in the vehicle's walking, platform lifting, and boom movement directions based on the calibrated obstacle distance, obstacle spatial position, and the future motion envelope of the aerial work platform. It then compares the collision time with the time required for the vehicle to complete braking. When the collision time is less than the braking time, it generates the corresponding braking control quantity for the motion mechanism; including:

[0084] Collision distance calculation within the movement area is based on the spatial position of the obstacle. Future motion envelope of aerial work platforms The spatial relationships are constructed by establishing the vehicle's traveling direction v, the platform's lifting direction h, and the boom's movement direction. Corresponding future occupied areas: ;Calculate the effective collision distance in each direction based on the shortest distance between the spatial location of the obstacle and the corresponding movement area: Simultaneously, the calibrated obstacle distances are used to constrain the effective collision distances in each direction. ; It is the future occupied area in the d direction; It is the effective collision distance in the d-th direction of motion; It is the shortest distance from a point to a spatial region; It is the effective distance in the constrained direction; This is the distance calibration weighting coefficient, ranging from 0 to 1;

[0085] Collision time calculation involves calculating the relative motion of the obstacle and the vehicle in the future; and calculating the relative approach velocity in the corresponding directions. ; This is the speed of the aerial work platform vehicle, measured in m / s. It is a relative approximate speed, measured in m / s; It is the speed of the obstacle, measured in m / s; if Then the collision time is predicted. For platforms and booms, Speed ​​of movement using equipment;

[0086] Directional risk calculation, time required for the vehicle to complete braking ; It is the braking response time, measured in seconds, representing the time from the detection of the braking command to the actual start of braking force generated by the vehicle. For platform lifting and boom movement direction, the mechanism stop response time is used. This is the current longitudinal speed of the vehicle, in m / s; It is the braking response time, measured in seconds, representing the time from the detection of a braking command to the actual generation of braking force by the vehicle; It is the magnitude of the vehicle's braking deceleration, measured in m / s². 2 Take positive values; for each potential collision direction corresponding to vehicle movement, platform lifting, and boom movement, construct directional risk responses: ; It is the risk response value in the d-th direction; It is the collision time in the d-th direction; This is the time smoothing term, measured in seconds (s). This indicates that the predicted collision time is less than the time required for the vehicle to complete braking, necessitating the triggering of appropriate emergency control measures.

[0087] Interlocking control defines risk states in three directions. ; It represents the vehicle's driving risk status, a binary variable that takes the value 1 when there is a collision risk and 0 otherwise. This indicates a risk status related to platform upgrades or downgrades. This represents the boom's movement risk state; T represents the transposition operation; braking commands are generated for the vehicle, platform, and boom respectively based on the risk state. ;in, This is the current normal motion control command; It is the actual control command after linkage control; These are three directional control commands;

[0088] when This indicates a collision risk in the direction the vehicle is traveling, and the vehicle should be stopped and braking should be applied.

[0089] when This indicates that there is a risk of collision in the platform's lifting area, and platform lifting should be stopped.

[0090] when This indicates that there is a risk of collision within the boom's motion envelope, and the corresponding boom movement should be stopped.

[0091] It is the final control instruction after risk assessment.

[0092] By performing the above operations, this solution addresses the problem of inaccurate obstacle spatial positioning in general aerial work platform emergency braking systems, which are unable to adapt to complex working environments and thus lead to low emergency braking safety. It improves obstacle spatial positioning accuracy and reduces false braking and missed braking caused by single-camera errors through multi-directional three-dimensional coordinate unification and distance measurement fusion. It compensates for distance measurement deviations using visual feature learning, improving the reliability of obstacle distance estimation. It achieves early judgment of dynamic collision zones through motion envelope prediction. Finally, it enables targeted emergency braking through multi-directional risk calculation, thereby improving the control safety of aerial work platforms during complex spatial operations.

[0093] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0094] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A multi-directional visual ranging-based aerial work platform vehicle linkage emergency braking system, characterized in that: The system includes a multi-directional visual image acquisition module, a visual color reconstruction module, a contour-fidelity visual enhancement module, a binocular 3D positioning module, a visual ranging calibration module, a high-altitude work vehicle motion envelope prediction module, and a linkage emergency braking module. The multi-directional visual image acquisition module acquires binocular images of the aerial work vehicle from different directions and simultaneously obtains motion status data. The visual color reconstruction module takes multi-directional binocular images as input, determines the effective brightness range in each direction, dynamically adjusts the color channels, and outputs the color-reconstructed binocular image. The contour-preserving visual enhancement module takes the reconstructed binocular image as input, and through multi-scale illumination estimation and structural information enhancement, outputs an enhanced binocular image that preserves the target edge and contour information. The binocular 3D positioning module takes the enhanced binocular image as input, calculates the obstacle depth through binocular parallax, and performs coordinate transformation and distance fusion to obtain the fused obstacle distance. The visual ranging calibration module takes enhanced binocular images as input, extracts target features and predicts distance errors through a lightweight visual model, calibrates the fused obstacle distances and outputs the results. The aerial work platform motion envelope prediction module takes vehicle motion state data as input, predicts future motion trajectory based on vehicle, platform and boom motion parameters, and outputs the aerial work platform spatial motion envelope. The linked emergency braking module takes the obstacle distance, spatial position, and motion envelope as input, calculates the collision risk, and generates linked braking control commands to realize linked emergency braking.

2. The aerial work platform vehicle linkage emergency braking system based on multi-directional visual ranging according to claim 1, characterized in that: The visual color reconstruction module includes: Gray-level distribution statistics for an input image with direction d and color channel c. Histogram of pixel values ​​for statistical channels: ; v is the number of pixels, and v is the grayscale value; It is a grayscale value; These are pixel coordinates; It is an indicator function; Determine the effective visual range and calculate the cumulative distribution: ; The grayscale value does not exceed The percentage of pixels; u is the grayscale index; It is the total number of pixels; Set low-end response coefficient and high-end response coefficient Determine the low and high effective grayscale boundaries respectively: ; ; and These are the effective low-grayscale boundary and effective high-grayscale boundary of the d-th direction and the c-th color channel, respectively; Dynamic channel adjustment is performed based on the effective pixel range, mapping the original grayscale values ​​to a unified effective dynamic range, and using effective low grayscale boundaries and effective high grayscale boundaries for adaptive channel mapping: ; These are the reconstructed grayscale values ​​of the color channels; It is the color channel gain coefficient; It is the color channel offset coefficient; It is a smoothing term; it restricts the reconstruction result: ; It is the grayscale value of the color channel after visual color reconstruction; It is a truncation function.

3. The aerial work platform vehicle linkage emergency braking system based on multi-directional visual ranging according to claim 2, characterized in that: The contour-fidelity visual enhancement module includes: Multi-scale coarse illumination estimation converts the reconstructed images of each visual direction output by the visual color reconstruction module into brightness maps. Set K spatial scales, and for the k-th spatial scale, calculate the coarse illumination component: ; It is the coarse illumination component of direction d at the k-th scale; It is a spatial scale; The standard deviation is Two-dimensional Gaussian kernel; Filtered contour fidelity calibration establishes a coarse local linear relationship between illumination and brightness within a local window centered at pixel (x,y) with radius r. ; It is the illumination component after contour fidelity calibration; and These are local coefficients; ; ; It is the illumination component after contour fidelity calibration; It is covariance; It is variance; It is a regularization parameter; It is a localized, rough average of illumination; Therefore The average brightness of the local window centered on the target; Structural information reconstruction involves logarithmic difference between the brightness map and the illumination components to extract local structural information that does not change synchronously with illumination variations, and calculation of structural information at various scales: ; It is the structural information at the k-th scale; Brightness enhancement involves overlaying the reconstructed structural information onto the original brightness map to obtain enhanced brightness. ; ; ; This is the enhanced brightness map; It is the structural reinforcement factor; It is the overall structural enhancement amount after fusing the d-th visual direction; It is the fusion weight of the k-th spatial scale; the enhanced brightness map is fused with the color information of the visual color reconstruction output image to obtain the enhanced binocular image.

4. The aerial work platform vehicle linkage emergency braking system based on multi-directional visual ranging according to claim 3, characterized in that: The binocular three-dimensional positioning module includes: For direction d, the disparity of the target point in the enhanced binocular image is: The target depth is: ; It's parallax; and The x-coordinate of the target point in the enhanced binocular image; It refers to the camera's pixel focal length; It is a binocular baseline; It is the target depth; The target's three-dimensional coordinates are: ; ; ; These are the coordinates of the target image; These are the coordinates of the camera's principal point; It refers to the position of the target in each coordinate system of the visual unit; Transform to vehicle coordinate system: ; It is the rotation matrix from the visual unit to the vehicle coordinate system; It is the translation vector of the d-th visual unit relative to the vehicle coordinate system; It refers to the spatial location of the obstacle; For distance measurement results from multiple directions for the same obstacle, calculate the distance fluctuation in the d-th direction: ; It is the ranging value of the i-th frame; It is the average distance across M consecutive frames; It is the standard deviation of the distance measurement; fusion is performed based on the variance of the distance measurement: ; It is the distance to the merged obstacle; It is the set of visual directions from which obstacles are observed; It is a smooth distance.

5. The aerial work platform vehicle linkage emergency braking system based on multi-directional visual ranging according to claim 4, characterized in that: The visual ranging calibration module includes: Set up a lightweight visual ranging calibration model, input the enhanced left and right images; use 4-level convolutional coding. Set up two output branches: The obstacle classification branch outputs the probability of the obstacle category. The distance calibration branch uses supervised residual regression training with the actual distance error as the supervision target; the network output is combined with the binocular geometric ranging results to obtain the calibrated obstacle distance.

6. The aerial work platform vehicle linkage emergency braking system based on multi-directional visual ranging according to claim 5, characterized in that: The linked emergency braking module includes: The collision distance calculation for the moving area is performed by constructing the future occupied areas corresponding to the vehicle's walking direction, platform lifting direction, and boom movement direction based on the spatial relationship between the obstacle's spatial location and the future motion envelope of the aerial work platform. The effective collision distance in each direction is calculated based on the shortest distance between the obstacle's spatial location and the corresponding moving area. At the same time, the effective collision distance in each direction is constrained using the calibrated obstacle distance to obtain the constrained effective distance in each direction. Collision time calculation involves calculating the relative motion of the obstacle and the vehicle in the future; and calculating the relative approach speed in the corresponding direction. Directional risk calculation: Calculate the time required for the vehicle to complete braking; construct a directional risk response. The system features coordinated control, defining directional risk conditions and generating braking commands for the vehicle, platform, and boom based on these risk conditions.

Images