A vehicle non-contact speed measurement method, device, computer equipment and storage medium

Abstract

The application is suitable for the field of speed measurement, and provides a vehicle non-contact speed measurement method, device, computer equipment and storage medium. The method comprises the following steps: acquiring continuous road surface images, inertial measurement data and environmental illumination parameters in the driving process of a vehicle; obtaining image displacement speed, road surface feature displacement speed and texture richness parameters based on the continuous road surface images; obtaining inertial speed and vibration amplitude parameters according to the inertial measurement data; determining the fusion weights of the image displacement speed, the road surface feature displacement speed and the inertial speed respectively according to the environmental illumination parameters, the texture richness parameters and the vibration amplitude parameters; and performing fusion processing on the image displacement speed, the road surface feature displacement speed and the inertial speed based on the fusion weights, and outputting the driving speed of the vehicle. The scheme can improve the stability and environmental adaptability of speed measurement accuracy.

Description

Technical Field

[0001] This application belongs to the field of speed measurement, and in particular relates to a non-contact speed measurement method, device, computer equipment and storage medium for vehicles. Background Technology

[0002] Existing vehicle speed measurement methods typically include contact-based speed measurement methods that are based on wheel rotation and non-contact speed measurement methods that are based on image or inertial measurement data. Contact-based speed measurement methods require the installation of detection components at locations such as wheels and axles, which are susceptible to installation errors, loose components, and wear, and also have high installation and maintenance costs.

[0003] Visual speed measurement methods based on road images can achieve high accuracy in scenarios with clear road markings and stable lighting. However, real-world road environments are complex and variable. Ambient light intensity changes in real time with factors such as weather, time of day, and shading. Furthermore, the paving materials and wear levels vary significantly across different road sections, all of which can affect the stability of image feature extraction and matching. Speed ​​measurement methods based on inertial measurement data obtain speed information by integrating acceleration or angular velocity. This method has the advantage of being unaffected by external environmental interference. However, inertial sensors have inherent bias and noise, causing their integration results to drift over time, making it difficult to maintain long-term accuracy when used alone.

[0004] To address this, some solutions attempt to combine or switch between visual and inertial speed measurement to compensate for each other's shortcomings. However, these combinations typically employ fixed fusion strategies or hard-switching logic, making it difficult to adapt to the continuous dynamic changes in road conditions, lighting environments, and vehicle motion during driving, resulting in significant fluctuations in speed measurement accuracy. Summary of the Invention

[0005] This application provides a contactless vehicle speed measurement method, device, computer equipment, and storage medium to solve the above-mentioned problems.

[0006] The first aspect of this application provides a contactless speed measurement method for vehicles, including: Acquire continuous road surface images, inertial measurement data, and environmental illumination parameters during vehicle operation; Based on the continuous road surface images, image displacement velocity, road surface feature displacement velocity, and texture richness parameters are obtained. The inertial velocity and vibration amplitude parameters are obtained based on the inertial measurement data. Based on the ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter, determine the respective fusion weights of the image displacement velocity, the road surface feature displacement velocity, and the inertial velocity; The image displacement velocity, the road feature displacement velocity, and the inertial velocity are fused based on the fusion weights to output the vehicle's driving speed.

[0007] A second aspect of this application provides a contactless vehicle speed measurement device, comprising: The image acquisition unit is used to acquire continuous road surface images during vehicle movement; An inertial measurement unit is used to collect inertial measurement data of the vehicle. The illumination information acquisition unit is used to acquire ambient illumination parameters; and The processing unit is configured to obtain image displacement velocity, road feature displacement velocity, and texture richness parameters based on the continuous road surface images; obtain inertial velocity and vibration amplitude parameters based on the inertial measurement data; determine the fusion weights of the image displacement velocity, road feature displacement velocity, and inertial velocity based on the ambient illumination parameters, texture richness parameters, and vibration amplitude parameters; perform fusion processing on the image displacement velocity, road feature displacement velocity, and inertial velocity based on the fusion weights, and output the vehicle's driving speed.

[0008] A third aspect of this application provides a computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in the first aspect.

[0009] A fourth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0010] In this embodiment, the above-described scheme dynamically determines the fusion weights of image displacement velocity, road feature displacement velocity, and inertial velocity based on real-time acquired ambient illumination parameters, texture richness parameters, and vibration amplitude parameters. The vehicle speed is then output through weighted fusion based on these determined weights. This dynamic weight determination mechanism adaptively adjusts the participation of image speed measurement information in the fusion process according to the current actual lighting conditions and road texture. It adjusts the weight ratios accordingly in scenarios with significant changes in illumination or road texture differences, effectively addressing the interference of ambient illumination fluctuations and road feature differences on visual speed measurement. Simultaneously, by dynamically adjusting the fusion weights using vibration amplitude parameters, the fusion ratio of each speed source can be adjusted based on the vehicle's current vibration state. Compared to existing fixed fusion strategies or hard-switching logic, this method allows the fusion weights to change in real-time with driving conditions, maintaining a reasonable speed information fusion ratio under different lighting conditions, road textures, and vibration states, thereby improving the stability and environmental adaptability of speed measurement accuracy. Attached Figure Description

[0011] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiments below. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 This is a flowchart of a vehicle contactless speed measurement method according to some embodiments of this application. Figure 1 ; Figure 2 This is a flowchart of a vehicle contactless speed measurement method according to some embodiments of this application. Figure 2 ; Figure 3 This is a structural diagram of a vehicle contactless speed measuring device according to some embodiments of this application; Figure 4 This is a structural diagram of the computer device provided in the embodiments of this application. Detailed Implementation

[0012] The embodiments of the technical solution of this application will now be described in detail with reference to the accompanying drawings. These embodiments are only used to more clearly illustrate the technical solution of this application and are therefore merely examples, and should not be used to limit the scope of protection of this application.

[0013] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0014] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0015] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0016] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0017] To illustrate the technical solution described in this application, specific embodiments are provided below.

[0018] like Figure 1 As shown, the first embodiment of the present invention provides a contactless speed measurement method for vehicles, including: S10. Acquire continuous road surface images, inertial measurement data, and environmental illumination parameters during vehicle operation; In a specific example, each frame of road surface image can be associated with an image timestamp and image size; inertial measurement data can include triaxial acceleration, triaxial angular velocity, and inertial timestamp; ambient illuminance parameters can include illuminance measurements and illuminance timestamps.

[0019] Following the example above, in this embodiment, the vehicle contactless speed measurement method is implemented based on a vehicle contactless speed measurement device, which may include an image acquisition unit, an inertial measurement unit, a light information acquisition unit, and a processing unit. The image acquisition unit acquires road surface images at a frame rate of 20 frames per second to 60 frames per second, and the inertial measurement unit acquires inertial measurement data at a sampling rate of 100 Hz to 800 Hz. The processing unit can establish a time correspondence between multiple sets of inertial measurement data between two adjacent road surface images and the corresponding image frames.

[0020] S20. Based on the continuous road surface image, obtain the image displacement velocity, road surface feature displacement velocity, and texture richness parameters; In a specific example, the process of obtaining image displacement velocity is as follows: The processing unit obtains the image displacement velocity based on the image content displacement between adjacent road surface images; the image content displacement can be obtained through optical flow, region matching, or phase correlation, etc.; the processing unit converts the displacement in the image coordinates into the displacement in the road surface coordinates according to the calibration relationship of the image acquisition unit.

[0021] Let the image coordinates of the i-th valid image position in the previous frame and the current frame be p, respectively. i and p i_prime G(·) represents the mapping relationship from image coordinates to road surface coordinates, then the corresponding road surface displacement ΔP i It can be represented as: ΔP i =G(p i_prime )-G(p i ) The processing unit obtains multiple ΔP i The component along the vehicle's direction of travel is obtained, and the image displacement velocity is obtained based on the time interval between adjacent frames. Multiple displacement components can be summarized using median, truncated average, or robust model fitting to reduce the impact of a few outlier matches.

[0022] Following the example above, the process of obtaining the road surface feature displacement velocity and texture richness parameters is as follows: The processing unit extracts natural texture feature points of the road surface from continuous road surface images, generates descriptors for the feature points, and establishes feature point matching relationships between adjacent road surface images; The processing unit filters valid feature point pairs based on descriptor distance, orientation consistency, and geometric model consistency.

[0023] Let the image coordinates of the j-th valid feature point pair in the previous frame and the current frame be q, respectively. j and q j_prime The corresponding road surface displacement ΔQ j It can be represented as: ΔQ j =G(q j_prime )-G(q j ) The processing unit is based on multiple ΔQ j The displacement velocity of road features is obtained by the component along the vehicle's direction of travel and the time interval between adjacent frames. Although the image displacement velocity and the road feature displacement velocity are obtained based on the same continuous road image, the former reflects the overall image content displacement, while the latter reflects the matching displacement of discrete natural texture feature points.

[0024] Texture richness parameters can be determined based on the number of effective feature points, the density of feature points per unit image area, image gradient energy, or image information entropy.

[0025] For example, let N valid A represents the number of valid feature points within the region of interest of the road surface. roi If the pixel area of ​​the region is represented by T, then the feature point density T N It can be represented as: T N =N valid / A roi S30. Obtain the inertial velocity and vibration amplitude parameters based on the inertial measurement data; In a specific example, the processing unit estimates the attitude of the inertial measurement unit based on the angular velocity data, transforms the acceleration in the vehicle coordinate system to the reference coordinate system, and removes the gravity component and acceleration zero bias to obtain the inertial velocity and vibration amplitude parameters.

[0026] Let the acceleration of the vehicle in the direction of travel be a. n The acceleration with zero bias is b n The inertial measurement sampling time interval is Δt. i Then the inertial velocity V imu (n) can be derived recursively as follows: V imu (n)=V imu (n-1)+(a n -b n )×Δt i Where n is the sampling number of the inertial measurement data; the zero-bias acceleration can be obtained by averaging multiple sets of acceleration data while the vehicle is stationary. When the vehicle is determined to be stationary, the processing unit can perform zero-velocity updates to reduce the integral drift of the inertial velocity.

[0027] The vibration amplitude parameter can be obtained from the root mean square of the high-frequency components of angular velocity or acceleration within the sliding time window; the processing unit can also normalize and combine the angular velocity vibration index and the acceleration vibration index to simultaneously characterize image acquisition posture jitter and vehicle impact.

[0028] S40. Determine the fusion weights of the image displacement velocity, the road surface feature displacement velocity, and the inertial velocity based on the ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter. S50. Based on the fusion weights, the image displacement velocity, the road feature displacement velocity, and the inertial velocity are fused to output the vehicle's driving speed.

[0029] In a specific example, the processing unit determines the fusion weights of image displacement velocity, road feature displacement velocity, and inertial velocity based on ambient illumination parameters, texture richness parameters, and vibration amplitude parameters; and the fusion weights can be determined by lookup tables, piecewise functions, or continuous mapping functions.

[0030] Specifically, let the image displacement velocity V img , Road surface characteristic displacement velocity V feat and inertial velocity V imu The fusion weights are w img w feat and w imu And w img w feat and w imu If the sum of the three is 1, then the fusion speed V can be expressed as: V=w img ×V img +w feat ×V feat +w imu ×V imu The processing unit can also achieve fusion through an adaptive weighted Kalman filter. When the data of a certain speed source is missing or the quality is abnormal, the processing unit reduces the corresponding fusion weight or stops using the speed source. When all three speeds are invalid, the processing unit outputs a speed invalid status.

[0031] By having three speeds participate in vehicle speed estimation together, and adjusting the degree of influence of each speed source according to the current ambient illumination, road texture, and vehicle vibration, the impact of a single speed measurement path failing under specific operating conditions on the final output can be reduced.

[0032] As can be seen from the above processing, image displacement velocity is used to characterize the overall content displacement of adjacent road surface images, road surface feature displacement velocity is used to characterize the matching displacement of natural texture feature points on the road surface, and inertial velocity is obtained based on the vehicle's inertial measurement data. These three velocities have different data sources and formation methods, and can respectively reflect different aspects of vehicle motion.

[0033] Based on this, the processing unit adjusts the fusion weights of the three speeds by combining ambient illumination parameters, texture richness parameters, and vibration amplitude parameters. This allows speed sources less affected by the current operating state to have a greater impact on the fusion result, while speed sources affected by low illumination, insufficient texture, vehicle vibration, or data anomalies have a correspondingly reduced impact on the fusion result. Therefore, when the measurement quality of one speed measurement path deteriorates, other speed measurement paths can still participate in vehicle speed estimation, thereby reducing the impact of anomalies in a single speed measurement path on the output result and improving the continuity and stability of vehicle speed output.

[0034] In addition, both image displacement velocity and road feature displacement velocity are obtained based on continuous road images, but they respectively utilize the overall image content displacement and the matching displacement of discrete natural texture feature points; the two visual speed measurement results can complement each other and participate in the fusion with inertial velocity, thereby improving the adaptability of vehicle non-contact speed measurement to different road textures and driving conditions.

[0035] In one possible implementation, such as Figure 2 As shown, obtaining image displacement velocity, road feature displacement velocity, and texture richness parameters based on the continuous road surface image includes: S21. Perform optical flow calculation on the continuous road surface images to obtain the displacement of two adjacent road surface images, and calculate the image displacement velocity based on the displacement and the time interval between the two adjacent road surface images.

[0036] S22. Extract the natural texture feature points of the road surface in the continuous road surface images, match the natural texture feature points of the road surface in two adjacent frames of road surface images to obtain matching feature point pairs, and calculate the road surface feature displacement velocity based on the position change of each matching feature point pair in the two adjacent frames of road surface images and the time interval between the two adjacent frames of road surface images.

[0037] S23. Determine the texture richness parameter based on at least one of the following: the number of effective feature points, feature point density, image gradient energy, and image information entropy in the continuous road surface image.

[0038] In a specific example, the image displacement velocity described above can be obtained through optical flow calculation. The processing unit preprocesses two adjacent frames of road surface images. Preprocessing may include grayscale conversion, lens distortion correction, brightness normalization, noise suppression, and cropping of the road surface region of interest.

[0039] In one implementation, the processing unit employs sparse optical flow; the processing unit selects pixels that meet the corner quality conditions in the previous frame of the road surface image and searches for the corresponding positions in the current frame of the road surface image.

[0040] The processing unit calculates the two-dimensional motion vector of each pixel based on the condition that the brightness of the local image remains stable within a short period of time.

[0041] In another implementation, the processing unit uses dense optical flow; the processing unit calculates motion vectors for multiple pixels in the region of interest of the road surface. Dense optical flow can obtain a more complete image motion field and is suitable for road surfaces with a small number of detectable discrete feature points but still with continuous gray-level changes.

[0042] The processing unit determines the main optical flow direction based on the directions of multiple optical flow vectors and removes optical flow vectors that deviate significantly from the main optical flow direction. Subsequently, the processing unit further removes optical flow vectors with low reliability based on the calculation error or motion consistency of the optical flow vectors. The processing unit transforms the retained optical flow vectors after filtering to the road coordinate system and obtains the image displacement velocity based on the transformed optical flow vectors.

[0043] The displacement velocity of road surface features can be obtained by extracting natural texture feature points of the road surface and matching feature points between adjacent road surface images.

[0044] In one implementation, the processing unit employs the Oriented FAST and Rotated BRIEF (ORB) algorithm to detect feature points from adjacent road surface images and generate binary descriptors for each feature point. The processing unit determines candidate matching feature point pairs based on the Hamming distance between the binary descriptors and performs a bidirectional matching check on these candidate matching feature point pairs, retaining only feature point pairs that are nearest neighbors in both matching directions.

[0045] In another implementation, the processing unit uses Scale-Invariant Feature Transform (SIFT) to extract feature points and floating-point descriptors, and filters the matching results based on the Euclidean distance and the nearest neighbor distance ratio; SIFT can be used when there are certain changes in the image scale or the pose of the image acquisition unit.

[0046] The processing unit can use the random sampling consensus algorithm to perform geometric consistency screening on candidate feature point pairs. The processing unit estimates the inter-frame geometric transformation model from the candidate feature point pairs and takes the feature point pairs that conform to the model as inliers. The processing unit obtains the road feature displacement velocity based on the road surface coordinate displacement corresponding to the inliers.

[0047] When obtaining image displacement velocity based on optical flow, the processing unit uses pixel motion information from multiple locations in the road surface image to determine the image displacement between adjacent frames. Therefore, even when the number of extractable discrete feature points is small, image displacement can still be estimated based on the pixel motion information. When obtaining road surface feature displacement velocity based on feature point matching, the processing unit uses feature descriptors to distinguish different natural road surface textures and determines the road surface feature displacement based on the positional changes of matched feature points between adjacent frames. Therefore, when local brightness changes or some optical flow vectors are unstable, an alternative visual displacement estimation result can be provided. Thus, image displacement velocity and road surface feature displacement velocity can reflect the vehicle's motion relative to the road surface from different perspectives and provide complementary velocity observations for subsequent fusion processing.

[0048] In one possible implementation, before performing optical flow calculations on the continuous road surface images to obtain the displacement between two adjacent road surface images, the method further includes: Based on the angular velocity data in the inertial measurement data, the image acquisition attitude corresponding to each frame of the road surface image and the attitude change between adjacent frames are obtained, and motion compensation is performed on the continuous road surface images based on the attitude change to obtain the motion-compensated continuous road surface images; and / or, The continuous road surface image is enhanced by estimating the illuminance component and enhancing the reflectance component using a Retinex network that has undergone channel pruning and quantization. The Retinex network includes an encoding structure and a decoding structure, both of which include depthwise separable convolutional layers.

[0049] In a specific example, before performing optical flow calculations on the continuous road surface images to obtain the displacement between two adjacent road surface images, the processing unit determines the image acquisition posture corresponding to each road surface image frame based on the angular velocity data in the inertial measurement data, and determines the posture change between adjacent frames based on the image acquisition postures corresponding to two adjacent road surface images. The image acquisition posture is used to represent the rotation state of the image acquisition unit relative to the vehicle coordinate system or the road surface coordinate system, and can be represented by Euler angles, rotation matrices, or quaternions.

[0050] Specifically, the processing unit integrates the angular velocity data between the acquisition time of the previous frame of the road surface image and the acquisition time of the current frame of the road surface image to obtain the attitude change of the current frame relative to the previous frame. When the sampling frequency of the angular velocity data is higher than the acquisition frequency of the road surface images, the processing unit integrates multiple sets of angular velocity data between two adjacent frames of the road surface images. The processing unit can also combine acceleration data to correct the image acquisition attitude, thereby reducing the cumulative error caused by the integration of angular velocity.

[0051] The processing unit performs coordinate transformation on the current frame road image based on the attitude change between adjacent frames, compensating for the image displacement caused by the attitude change of the image acquisition unit in the current frame road image. This motion compensation primarily eliminates additional image displacement caused by vehicle bumps or image acquisition unit jitter, while preserving the image content displacement caused by the vehicle's movement relative to the road surface.

[0052] The processing unit can perform motion compensation frame by frame on continuous road surface images, or it can perform motion compensation when the vibration amplitude parameter reaches the compensation start condition. When using the conditional triggering method, compensation start threshold and compensation exit threshold can be set separately to reduce frequent switching of motion compensation state.

[0053] After motion compensation, the processing unit obtains image displacement velocity, road feature displacement velocity, and texture richness parameters based on the motion-compensated continuous road surface images. This reduces the impact of image acquisition pose changes on image displacement calculation, feature point matching, and texture evaluation, allowing the obtained visual velocity to better reflect the vehicle's actual driving motion relative to the road surface.

[0054] If angular velocity data is missing or abnormal within the current image acquisition cycle, the processing unit may not perform motion compensation on the corresponding road surface image and may reduce the real-time confidence of the image displacement velocity and the road surface feature displacement velocity to reduce the impact of insufficiently compensated visual velocity on the fusion result.

[0055] In another example, when the ambient illuminance parameter is less than the preset illuminance threshold, the processing unit enhances the continuous road surface image through an image enhancement network based on the Retina-Cortex Theory (Retinex) after channel pruning and quantization, and obtains the enhanced continuous road surface image.

[0056] The Retinex network processes low-illuminance road images based on a model formed by the reflection and illumination components of the road image. For the input road image... , can be represented as:

[0057] in, Represents the reflection component, used to characterize road surface material and texture; The illuminance component represents the spatial distribution of light in an image; x and y represent the horizontal and vertical coordinates of a pixel in the road image, respectively. The Retinex network includes an encoding structure and a decoding structure. The encoding structure is used to extract texture features, edge features, and illumination features from the input road surface image. The decoding structure is used to estimate the illumination component and enhance the reflection component based on the encoding features to obtain an enhanced road surface image. Both the encoding and decoding structures include depthwise separable convolutional layers. The depthwise separable convolutional layers may include channel-wise convolutional layers and point-wise convolutional layers to reduce the number of network parameters and computational cost.

[0058] In a specific example, the Retinex network outputs an illuminance estimation map and a reflection enhancement map. The processing unit obtains the enhanced road surface image based on the reflection enhancement map. During network training, the network parameters can be adjusted based on the difference between the enhanced road surface image and the normal illuminance image, the smoothness of the illuminance component, and the degree of texture preservation of the reflection component.

[0059] Channel pruning removes less important convolutional channels from the Retinex network. After channel pruning, the network can be retrained to reduce the impact of channel pruning on image enhancement. Quantization reduces the representation bit width of network weights and intermediate feature data, further reducing model storage and inference computation.

[0060] The processing unit obtains image displacement velocity, road feature displacement velocity, and texture richness parameters based on the enhanced continuous road surface image. By estimating the illuminance component and enhancing the reflection component in the low-illuminance road surface image, the discernibility of road surface texture and edges can be improved, thereby enhancing image displacement calculation, feature point matching, and texture richness evaluation under low-illuminance conditions.

[0061] When the ambient illuminance parameter is not less than the preset illumination threshold, the processing unit can disable the Retinex network and directly process the acquired continuous road surface images. To avoid frequent activation and deactivation of the Retinex network due to fluctuations in ambient illuminance around the preset illumination threshold, enhanced activation and deactivation thresholds can be set separately.

[0062] In one possible implementation, motion compensation of the continuous road surface image based on the attitude change includes: Based on the angular velocity data between the acquisition time of the previous frame road image and the acquisition time of the current frame road image, determine the change in attitude angle of the current frame road image relative to the previous frame road image. Obtain the intrinsic parameter matrix of the image acquisition unit used to acquire the continuous road surface images and the attitude angle change amount, and construct the homography transformation matrix of the current frame road surface image relative to the previous frame road surface image based on the intrinsic parameter matrix and the attitude angle change amount; The current frame road surface image is subjected to coordinate transformation according to the homography transformation matrix to obtain the motion-compensated current frame road surface image, so as to compensate for the image displacement of the current frame road surface image relative to the previous frame road surface image caused by the change in image acquisition posture.

[0063] In a specific example, the processing unit determines the change in attitude angle of the current frame road surface image relative to the previous frame road surface image based on the angular velocity data between the acquisition time of the previous frame road surface image and the acquisition time of the current frame road surface image; let the angular velocity data be ω(t), the acquisition time of the previous frame be t(k-1), and the acquisition time of the current frame be t k Then the change in attitude angle Δθ k It can be represented as: Δθ k =∫[t(k-1),t k ]ω(t)dt Where, Δθk This represents the change in attitude angle corresponding to the kth group of adjacent frames.

[0064] The processing unit acquires the intrinsic parameter matrix K of the image acquisition unit. The intrinsic parameter matrix may include focal length, principal point position, and pixel ratio parameters in two image coordinate directions. The intrinsic parameter matrix can be obtained during the production calibration phase and stored in non-volatile memory.

[0065] The change in attitude angle can be converted into a rotation matrix R(Δθ) k In an implementation primarily based on rotation compensation, the homography transformation matrix H of the current frame relative to the previous frame can be expressed as: H=K×R(Δθ k )×K -1 For a pixel homogeneous coordinate p in the current frame road surface image, the compensated pixel homogeneous coordinate pc can be determined based on the homogeneous correspondence between pc and H multiplied by p.

[0066] The processing unit performs coordinate transformation on the current frame based on H, and determines the transformed pixel value through bilinear interpolation, bicubic interpolation, or nearest neighbor interpolation.

[0067] Based on the definition of the direction of the pose change, the processing unit can also perform inverse mapping on the current frame using the inverse matrix of H. The direction of the homography transformation matrix should be consistent with the pose relationship of the current frame relative to the previous frame. Using inverse mapping can reduce hole pixels after coordinate transformation.

[0068] After image coordinate transformation, the processing unit can crop the overlapping region that exists in both the previous frame and the compensated current frame; visual velocimetry can be performed within the overlapping region to avoid invalid pixels at the transformation edge affecting optical flow or feature matching.

[0069] The homography transformation described above is mainly used to compensate for the rotational changes of the image acquisition unit; the translational displacement caused by the vehicle traveling normally along the road surface is still retained in the compensated image; the subsequent visual speed measurement obtains the vehicle speed based on the remaining inter-frame road surface displacement.

[0070] When the height of the image acquisition unit relative to the road surface changes significantly or the road surface is not approximately flat, rotational homography compensation may have residual errors. The processing unit can reflect the residual errors in the basic confidence of visual velocity, or it can perform compensation and velocity estimation separately for multiple image regions.

[0071] By utilizing the angular velocity data between two adjacent frames, the attitude change can be mapped to a specific time interval. By converting the attitude change into an image coordinate transformation, the additional image displacement caused by the attitude change during image acquisition can be compensated while preserving the vehicle's displacement.

[0072] In one possible implementation, determining the fusion weights of the image displacement velocity, the road feature displacement velocity, and the inertial velocity based on the ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter includes: In each fusion cycle, based on the numerical relationships between the current ambient illumination parameter, texture richness parameter, and vibration amplitude parameter and their respective preset thresholds, the real-time confidence levels of the image displacement velocity, road feature displacement velocity, and inertial velocity under the corresponding numerical relationships are determined. Specifically, the magnitude of the ambient illumination parameter is positively correlated with the real-time confidence levels of the image displacement velocity and road feature displacement velocity; the magnitude of the vibration amplitude parameter is negatively correlated with the real-time confidence levels of the image displacement velocity and road feature displacement velocity; and the magnitude of the texture richness parameter is positively correlated with the real-time confidence level of the road feature displacement velocity. The real-time confidence level can be a value between 0 and 1. The closer the real-time confidence level is to 1, the more suitable the corresponding velocity source is as a vehicle speed observation in the current fusion cycle.

[0073] Based on the real-time confidence level, the fusion weights corresponding to the image displacement velocity, the road surface feature displacement velocity, and the inertial velocity are determined.

[0074] The step of determining the real-time confidence levels of the image displacement velocity, the road feature displacement velocity, and the inertial velocity under the corresponding numerical relationships between the current ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter and their respective preset thresholds includes: When the ambient illuminance parameter is less than a preset illuminance threshold, the real-time confidence level of the inertial velocity is determined to be greater than the real-time confidence levels of the image displacement velocity and the road feature displacement velocity, respectively. When the vibration amplitude parameter is greater than the preset vibration threshold, the real-time confidence level of the inertial velocity is determined to be greater than the real-time confidence levels of the image displacement velocity and the road surface feature displacement velocity, respectively. When the texture richness parameter is less than a preset texture threshold, the real-time confidence level of the road surface feature displacement velocity is determined to be less than the real-time confidence levels of the image displacement velocity and the inertial velocity, respectively.

[0075] In a specific example, the processing unit calculates the confidence C of the illuminator. L Texture subconfidence C T and oscillator confidence C AThe confidence level of the illumination sub-sub can increase with the increase of the ambient illuminance parameter; the confidence level of the texture sub-sub ...

[0076] Real-time confidence level C of image displacement velocity img It can be determined based on the illumination sub-confidence, vibrator sub-confidence, and optical flow fundamental quality index; the optical flow fundamental quality index can be determined by the effective optical flow quantity, forward and backward consistency error, or region matching peak; the real-time confidence level C of the road surface characteristic displacement velocity. feat The basic quality index of feature matching can be determined based on the confidence of illumination sub-sub ...

[0077] Real-time confidence level C of inertial velocity imu The confidence level of the inertial velocity can be determined based on the validity of the inertial measurement data, the integration duration, the sensor range status, and the zero-bias stability. The real-time confidence level of the inertial velocity does not need to be continuously maintained at a high value. When inertial measurement data exhibits saturation, timestamp interruptions, zero-bias anomalies, or excessively long integration durations, the processing unit can reduce the C-value. imu .

[0078] In a specific implementation that satisfies the preset corresponding rule described in the claim, when the ambient illuminance parameter is less than the preset illumination threshold, the processing unit will C imu Set to greater than C img and C feat The value of C. At this time, C imu It can remain at its original value or decrease due to a decline in the quality of inertial data, as long as it is still greater than the real-time confidence of each of the two visual velocities within the current fusion cycle.

[0079] When the vibration amplitude parameter exceeds the preset vibration threshold, the processing unit will... imu Set to greater than C img and C feat The numerical value; this relative relationship is used to reflect that the influence of image acquisition posture jitter on visual velocity is greater than its influence on short-term inertial velocity; if the inertial measurement data saturates at the same time, the processing unit can mark all three velocities as low confidence and output an abnormal state.

[0080] When the texture richness parameter is less than the preset texture threshold, the processing unit will C feat Set to less than C img and C imuThe numerical value. Image displacement velocity can still be obtained through region correlation or dense optical flow using continuous grayscale structure, without requiring a sufficient number of discrete feature points.

[0081] When multiple conditions, including low illumination, strong vibration, and low texture, are simultaneously met, the processing unit can determine the confidence level according to a rule-based combination. For example, under low illumination and low texture conditions, a C-value can be formed. imu Greater than C img And C img Greater than C feat The confidence level relationship. Under strong vibration and saturation of inertial data, the processing unit may not force the maintenance of a high confidence level for inertial velocity, but instead place all velocity sources in the anomaly processing branch.

[0082] The processing unit can determine the fusion weights based on three real-time confidence levels: w img =C img / (C img +C feat +C imu +ε) w feat =C feat / (C img +C feat +C imu +ε) w imu =C imu / (C img +C feat +C imu +ε) Here, ε is a positive number to prevent the denominator from being zero. ε can be a value much smaller than the sum of normal confidence levels. The processing unit can also query the fusion weight table based on the real-time confidence level. The fusion weight table can be divided into multiple state combinations according to illumination state, texture state, and vibration state, and each state combination corresponds to a set of three-way fusion weights.

[0083] The ambient illumination threshold, texture threshold, and vibration threshold can be determined through calibration data. During calibration, three speeds and a reference speed are collected simultaneously under different lighting conditions, different road surfaces, and different vibration conditions. The error of each speed relative to the reference speed is calculated, and the state parameter value that can distinguish the error change area is selected as the threshold.

[0084] By recalculating the real-time confidence level in each fusion cycle, it is possible to avoid the perception that increases or decreases are continuously accumulating on historical values; by characterizing the effects of illumination, texture, and vibration on different velocity sources separately, the fusion weights can be redistributed according to the current working state.

[0085] Following the example above, when any of the real-time confidence levels is lower than a preset confidence threshold, the fusion weight corresponding to that real-time confidence level will be adjusted to zero.

[0086] In one possible implementation, the fusion processing of the image displacement velocity, the road feature displacement velocity, and the inertial velocity based on each of the fusion weights includes: One or more algorithms, including adaptive weighted Kalman filtering, weighted least squares, and weighted averaging, are used to fuse the image displacement velocity, the road feature displacement velocity, and the inertial velocity based on the fusion weights.

[0087] Specifically, when using the weighted average method, the image displacement velocity, the road feature displacement velocity, and the inertial velocity can be multiplied by their respective fusion weights, and the results can be summed to obtain the vehicle's speed; wherein, the sum of each fusion weight can be 1.

[0088] When using the weighted least squares method, weighted residuals corresponding to the image displacement velocity, the road feature displacement velocity, and the inertial velocity can be constructed according to each of the fusion weights. The vehicle speed is obtained with the goal of minimizing the sum of the weighted residuals. The larger the fusion weight, the greater the influence of its corresponding residual on the solution result.

[0089] When using adaptive weighted Kalman filtering, the vehicle's speed can be used as a state variable, and the image displacement velocity, the road feature displacement velocity, and the inertial velocity can be used as observations. The observation noise parameter or observation gain of the corresponding observations can be adjusted according to the fusion weights to recursively estimate the vehicle's speed. The larger the fusion weight, the greater the influence of the corresponding observation in the state update process.

[0090] When using multiple algorithms, the speeds can be filtered first using an adaptive weighted Kalman filter, and then the filtered speeds can be fused using a weighted average method or a weighted least squares method to obtain the vehicle's speed.

[0091] In one possible implementation, when the ambient illuminance parameter is less than a preset illumination threshold, acquiring continuous road surface images during vehicle travel includes: In each of multiple consecutive image acquisition cycles, road surface images are acquired using at least two different sets of exposure parameters, wherein the exposure parameters include at least one of exposure time and acquisition gain. Road surface images acquired using different exposure parameters within the same image acquisition cycle are fused to obtain a fused road surface image corresponding to the image acquisition cycle. The continuous road surface image is composed of fused road surface images corresponding to multiple consecutive image acquisition cycles.

[0092] The second embodiment of this application provides a contactless speed measuring device 100 for vehicles, such as... Figure 3 As shown, the vehicle contactless speed measuring device 100 includes an image acquisition unit 101, an inertial measurement unit 102, a light information acquisition unit 103, and a processing unit 104. The image acquisition unit 101, the inertial measurement unit 102, and the light information acquisition unit 103 are respectively connected to the processing unit 104 for data transmission. The image acquisition unit 101 is used to acquire continuous road surface images during vehicle travel and transmit the continuous road surface images to the processing unit 104; The inertial measurement unit 102 is used to collect the inertial measurement data of the vehicle and transmit the inertial measurement data to the processing unit 104; The illumination information acquisition unit 103 is used to acquire ambient illumination parameters and transmit the ambient illumination parameters to the processing unit 104; The processing unit 104 is used to obtain image displacement velocity, road feature displacement velocity and texture richness parameters based on the continuous road surface images, obtain inertial velocity and vibration amplitude parameters based on the inertial measurement data, determine the fusion weights of the image displacement velocity, road feature displacement velocity and inertial velocity based on the ambient illumination parameters, texture richness parameters and vibration amplitude parameters, and perform fusion processing on the image displacement velocity, road feature displacement velocity and inertial velocity based on each fusion weight, and output the vehicle's driving speed.

[0093] The data connection can be formed through a data bus, interface circuit, or other connection methods that enable data transmission.

[0094] In a specific example, the vehicle contactless speed measuring device provided in this application embodiment includes: The system includes an image acquisition unit, an inertial measurement unit, a lighting information acquisition unit, and a processing unit. The image acquisition unit, inertial measurement unit, and lighting information acquisition unit are each connected to the processing unit, which receives the data output from each unit and outputs the vehicle's speed.

[0095] The image acquisition unit is used to acquire continuous road surface images facing the road surface where vehicles are traveling. The continuous road surface images are multiple frames of road surface images arranged according to acquisition time. The image acquisition unit can output color images or grayscale images.

[0096] An inertial measurement unit (IMU) is used to acquire inertial measurement data. An IMU may include an accelerometer and an angular velocity sensor to output triaxial acceleration and triaxial angular velocity, respectively. The accelerometer and angular velocity sensor can be integrated into the same IMU or configured separately.

[0097] The illumination information acquisition unit is used to acquire ambient illumination parameters. In some embodiments, the illumination information acquisition unit includes an ambient light sensor. In other embodiments, the illumination information acquisition unit includes an illumination estimation module executed by a processing unit, which estimates the ambient illumination parameters based on the average brightness of the road surface image, a brightness histogram, exposure time, or acquisition gain.

[0098] The processing unit can be composed of one or more of a general-purpose processor, digital signal processor, image processor, neural network processor, or field-programmable gate array. The processing unit performs visual velocimetry, inertial velocimetry, state parameter calculation, and multi-source fusion processing, respectively.

[0099] The data output by the image acquisition unit, inertial measurement unit, and illumination information acquisition unit can carry timestamps. The processing unit aligns continuous road surface images, inertial measurement data, and ambient illumination parameters according to the timestamps. When the sampling frequency of inertial measurement data is higher than the image acquisition frequency, the processing unit extracts the inertial measurement data between the acquisition times of two adjacent frames of road surface images for inertial velocity calculation and image motion compensation.

[0100] The vehicle can be a bicycle, electric bicycle, scooter, or balance bike, or any other vehicle capable of having an image acquisition unit facing the road surface. The image acquisition unit can be located at the front of the handlebars, the front of the frame, or the bottom of the frame, with its acquisition direction facing the road surface in front of or below the vehicle.

[0101] Image displacement velocity represents the vehicle speed obtained from the overall image content displacement in continuous road surface images. Road surface feature displacement velocity represents the vehicle speed obtained from the matching displacement of natural texture feature points of the road surface in adjacent road surface images. Inertial velocity represents the vehicle speed obtained from inertial measurement data.

[0102] The texture richness parameter characterizes the amount of texture information in the road image that can be used for image matching or feature matching. The vibration amplitude parameter characterizes the degree of vibration of the vehicle or image acquisition unit within the current time period. The fusion period represents the time period during which the processing unit updates the three speeds, state parameters, fusion weights, and vehicle speed.

[0103] In a specific example, the vehicle contactless speed measurement device includes an image acquisition unit, an inertial measurement unit, a light information acquisition unit, and a processing unit.

[0104] The image acquisition unit includes a camera and an image interface circuit. The camera is positioned facing the road surface where the vehicle is traveling and outputs continuous road surface images to the processing unit via the image interface circuit. The image acquisition unit can be installed at the front or bottom of the vehicle, and its installation position should ensure that the field of view covers the road surface area along the vehicle's travel path.

[0105] The inertial measurement unit includes an accelerometer and an angular velocity sensor, and outputs inertial measurement data to the processing unit. The inertial measurement unit can be arranged adjacent to the image acquisition unit, or it can be arranged on the circuit board where the processing unit is located.

[0106] The illumination information acquisition unit may include an ambient light sensor and output ambient illuminance parameters to the processing unit. The ambient light sensor may be positioned at a light-transmitting location on the device housing to reduce the impact of the housing obstructing the illuminance measurement.

[0107] The processing unit is connected to the image acquisition unit, the inertial measurement unit, and the illumination information acquisition unit, respectively. Based on continuous road surface images, the processing unit obtains image displacement velocity, road feature displacement velocity, and texture richness parameters. Based on inertial measurement data, it obtains inertial velocity and vibration amplitude parameters. Furthermore, it determines the fusion weights for each of the three velocities based on the ambient illumination parameters, texture richness parameters, and vibration amplitude parameters. The processing unit fuses the three velocities according to their fusion weights and outputs the vehicle's travel speed.

[0108] The processing unit can be configured with an image buffer and an inertial data buffer. The image buffer stores multiple frames of road surface images arranged according to acquisition time, while the inertial data buffer stores multiple sets of acceleration and angular velocity data acquired between adjacent image frames. The processing unit reads the image data and inertial measurement data corresponding to the same fusion period based on the timestamp.

[0109] After the device is powered on, the processing unit completes the initialization of each unit and obtains zero bias in inertial measurement. After entering continuous speed measurement, the processing unit cyclically acquires multi-source data, calculates three types of speed and state parameters, updates the fusion weights, and outputs the vehicle speed.

[0110] When the image acquisition unit is temporarily unable to provide a valid road surface image, the processing unit can reduce the fusion weight of the two visual velocities and maintain the output using the inertial velocity within a limited time. When the inertial measurement data is invalid, the processing unit can fuse the two visual velocities.

[0111] By integrating image acquisition, inertial measurement, illumination information acquisition, and multi-source processing into the same device, vehicle speed can be obtained without contacting the vehicle wheels.

[0112] In a specific example, the following describes an end-to-end processing procedure using a contactless speed measurement device mounted on the front of a bicycle frame. This process combines dual-vision speed measurement, real-time confidence, adaptive weighted Kalman filtering, image motion compensation, and low-light Retinex enhancement.

[0113] After the device is powered on, the processing unit initializes the image acquisition unit, inertial measurement unit, and illumination information acquisition unit, and reads the intrinsic parameter matrix of the image acquisition unit and the mapping parameters from image coordinates to road surface coordinates. When the vehicle remains stationary, the processing unit determines the zero-bias acceleration and zero-bias angular velocity based on multiple sets of inertial measurement data.

[0114] After the vehicle begins to move, the image acquisition unit continuously acquires images of the road surface, the inertial measurement unit acquires acceleration and angular velocity at a sampling frequency higher than the image frame rate, and the illumination information acquisition unit continuously obtains ambient illumination parameters. The processing unit aligns the above data according to the timestamp.

[0115] Between two adjacent road surface images, the processing unit determines the change in attitude angle of the current frame relative to the previous frame based on the angular velocity data, and constructs a homography transformation matrix based on the intrinsic parameter matrix of the image acquisition unit to perform motion compensation on the current frame road surface image.

[0116] The processing unit performs optical flow calculations on the motion-compensated continuous road surface images to obtain image displacement velocity; simultaneously, it extracts natural texture feature points of the road surface and performs inter-frame matching to obtain road surface feature displacement velocity and texture richness parameters. The processing unit also obtains inertial velocity and vibration amplitude parameters based on inertial measurement data.

[0117] Within each fusion cycle, the processing unit determines the real-time confidence level and fusion weight of the three velocities based on ambient illumination parameters, texture richness parameters, and vibration amplitude parameters. When a vehicle enters a low-texture road surface, the fusion weight of the road feature displacement velocity decreases; when a vehicle passes through a bumpy road section, the fusion weight of the inertial velocity relatively increases.

[0118] When a vehicle enters a low-light area, the processing unit enhances the continuous road surface image through a Retinex network that has undergone channel pruning and quantization. Then, based on the enhanced continuous road surface image, it calculates the image displacement velocity, road feature displacement velocity, and texture richness parameters.

[0119] The processing unit uses the three speeds as observation vectors for an adaptive weighted Kalman filter, adjusts the system noise covariance matrix and observation noise covariance matrix based on real-time confidence, and outputs the vehicle speed. The above feature combination is only one typical implementation and does not constitute a limitation on various extended feature combination methods.

[0120] In one possible implementation, the processing unit may functionally include a visual processing module, an inertial processing module, a weight determination module, and a fusion processing module. The visual processing module performs optical flow calculations on continuous road surface images to obtain image displacement velocity, and extracts natural texture feature points of the road surface and performs inter-frame matching on these feature points to obtain road surface feature displacement velocity and texture richness parameters. The inertial processing module obtains inertial velocity and vibration amplitude parameters based on inertial measurement data. The weight determination module determines the real-time confidence level and fusion weight for each of the three velocities based on the ambient illumination parameter, texture richness parameter, and vibration amplitude parameter. The fusion processing module adjusts the system noise covariance matrix and observation noise covariance matrix of the adaptive weighted Kalman filter based on the real-time confidence levels, and performs filtering and fusion on the three velocities.

[0121] The processing unit may further include a motion compensation module and a low-light enhancement module. The motion compensation module determines the attitude change between adjacent road surface images based on angular velocity data, and performs motion compensation on continuous road surface images based on the attitude change. The low-light enhancement module enhances continuous road surface images using a Retinex network that has undergone channel pruning and quantization when the ambient illumination parameter is less than a preset illumination threshold. Each module is a logical module divided according to its processing function, and can be implemented by the processing unit running corresponding programs.

[0122] In one possible implementation, the vehicle contactless speed measurement device may further include a supplementary lighting unit connected to the processing unit and positioned towards the road surface area captured by the image acquisition unit. The supplementary lighting unit may include an infrared supplementary light or other supplementary light source capable of improving the brightness of the road surface image.

[0123] When the ambient illuminance parameter is less than a preset illumination threshold, the processing unit controls the supplementary lighting unit to turn on, providing supplementary lighting to the road image acquisition area of ​​the image acquisition unit. The processing unit can also control the supplementary lighting unit to work synchronously with the image acquisition unit according to the image acquisition cycle, ensuring the supplementary lighting unit emits light during road image acquisition. Based on continuous road images acquired under supplementary lighting conditions, the processing unit obtains image displacement velocity, road feature displacement velocity, and texture richness parameters; when the ambient illuminance parameter is not less than the preset illumination threshold, the processing unit can control the supplementary lighting unit to turn off.

[0124] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0125] In the above embodiments of this application, the various embodiments and implementation methods can be combined with each other, and there is no obstacle to their combination due to the separate description of the embodiments and implementation methods. The implementation processes of the various embodiments and implementation methods can be referred to each other, and features can be integrated to form an overall solution that includes the technical features of the various embodiments or implementation methods.

[0126] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0127] In one embodiment, such as Figure 4 As shown, a computer device is provided. The computer device 5 of this embodiment includes: at least one processor 500 (… Figure 4 (Only one is shown in the diagram), memory 501, and computer program 502 stored in said memory 501 and executable on said at least one processor 500, wherein said processor 500 executes said computer program 502 to implement the steps in any of the above method embodiments.

[0128] The computer device 5 can be a desktop computer, laptop, handheld computer, or other computing device. The computer device 5 may include, but is not limited to, a processor 500 and a memory 501. Those skilled in the art will understand that... Figure 4 This is merely an example of computer device 5 and does not constitute a limitation on computer device 5. It may include more or fewer components than shown, or combine certain components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.

[0129] The processor 500 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0130] The memory 501 can be an internal storage unit of the computer device 5, such as a hard disk or memory of the computer device 5. The memory 501 can also be an external storage device of the computer device 5, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device 5. Furthermore, the memory 501 can include both internal and external storage units of the computer device 5. The memory 501 is used to store the computer program and other programs and data required by the computer device. The memory 501 can also be used to temporarily store data that has been output or will be output.

[0131] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0132] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0133] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0134] In the embodiments provided in this application, it should be understood that the disclosed apparatus / computer devices and methods can be implemented in other ways. For example, the apparatus / computer device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0135] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0136] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0137] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0138] The methods described in this application can be implemented in whole or in part by a computer program product. When the computer program product is run on a computer device, the computer device executes the steps in the various method embodiments described above.

[0139] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A non-contact speed measurement method for vehicles, characterized in that, include: Acquire continuous road surface images, inertial measurement data, and environmental illumination parameters during vehicle operation; Based on the continuous road surface images, image displacement velocity, road surface feature displacement velocity, and texture richness parameters are obtained. The inertial velocity and vibration amplitude parameters are obtained based on the inertial measurement data. Based on the ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter, determine the respective fusion weights of the image displacement velocity, the road surface feature displacement velocity, and the inertial velocity; The image displacement velocity, the road feature displacement velocity, and the inertial velocity are fused based on the fusion weights to output the vehicle's driving speed.

2. The non-contact vehicle speed measurement method according to claim 1, characterized in that, The process of obtaining image displacement velocity, road feature displacement velocity, and texture richness parameters based on the continuous road surface image includes: Optical flow calculation is performed on the continuous road surface images to obtain the displacement between two adjacent road surface images. The image displacement velocity is then calculated based on the displacement and the time interval between the two adjacent road surface images. Extract the natural texture feature points of the road surface in the continuous road surface images, match the natural texture feature points of the road surface in two adjacent frames of road surface images to obtain matching feature point pairs, and calculate the road surface feature displacement velocity based on the position change of each matching feature point pair in the two adjacent frames of road surface images and the time interval between the two adjacent frames of road surface images. The texture richness parameter is determined based on at least one of the following: the number of effective feature points, feature point density, image gradient energy, and image information entropy in the continuous road surface image.

3. The non-contact vehicle speed measurement method according to claim 2, characterized in that, Before performing optical flow calculation on the continuous road surface images to obtain the displacement between two adjacent road surface images, the method further includes: Based on the angular velocity data in the inertial measurement data, the image acquisition attitude corresponding to each frame of the road surface image and the attitude change between adjacent frames are obtained, and motion compensation is performed on the continuous road surface images based on the attitude change to obtain the motion-compensated continuous road surface images; and / or, The continuous road surface image is enhanced by estimating the illuminance component and enhancing the reflectance component using a Retinex network that has undergone channel pruning and quantization. The Retinex network includes an encoding structure and a decoding structure, both of which include depthwise separable convolutional layers.

4. The non-contact vehicle speed measurement method according to claim 3, characterized in that, The motion compensation of the continuous road surface image based on the attitude change includes: Based on the angular velocity data between the acquisition time of the previous frame road image and the acquisition time of the current frame road image, determine the change in attitude angle of the current frame road image relative to the previous frame road image. Obtain the intrinsic parameter matrix of the image acquisition unit used to acquire the continuous road surface images and the attitude angle change amount, and construct the homography transformation matrix of the current frame road surface image relative to the previous frame road surface image based on the intrinsic parameter matrix and the attitude angle change amount; The current frame road surface image is subjected to coordinate transformation according to the homography transformation matrix to obtain the motion-compensated current frame road surface image, so as to compensate for the image displacement of the current frame road surface image relative to the previous frame road surface image caused by the change in image acquisition posture.

5. The non-contact vehicle speed measurement method according to claim 1, characterized in that, The step of determining the fusion weights of the image displacement velocity, the road feature displacement velocity, and the inertial velocity based on the ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter includes: In each fusion cycle, based on the numerical relationships between the current ambient illumination parameter, texture richness parameter, and vibration amplitude parameter and their respective preset thresholds, the real-time confidence levels of the image displacement velocity, road feature displacement velocity, and inertial velocity under the corresponding numerical relationships are determined. Specifically, the magnitude of the ambient illumination parameter is positively correlated with the real-time confidence levels of the image displacement velocity and road feature displacement velocity; the magnitude of the vibration amplitude parameter is negatively correlated with the real-time confidence levels of the image displacement velocity and road feature displacement velocity; and the magnitude of the texture richness parameter is positively correlated with the real-time confidence level of the road feature displacement velocity. Based on the real-time confidence level, the fusion weights corresponding to the image displacement velocity, the road surface feature displacement velocity, and the inertial velocity are determined.

6. The non-contact vehicle speed measurement method according to claim 5, characterized in that, The step of determining the real-time confidence levels of the image displacement velocity, the road feature displacement velocity, and the inertial velocity under the corresponding numerical relationships between the current ambient illumination parameter, the texture richness parameter, and the vibration amplitude parameter and their respective preset thresholds includes: When the ambient illuminance parameter is less than a preset illuminance threshold, the real-time confidence level of the inertial velocity is determined to be greater than the real-time confidence levels of the image displacement velocity and the road feature displacement velocity, respectively. When the vibration amplitude parameter is greater than the preset vibration threshold, the real-time confidence level of the inertial velocity is determined to be greater than the real-time confidence levels of the image displacement velocity and the road feature displacement velocity, respectively. When the texture richness parameter is less than a preset texture threshold, the real-time confidence of the road surface feature displacement velocity is determined to be less than the real-time confidence of the image displacement velocity and the inertial velocity, respectively.

7. The non-contact vehicle speed measurement method according to claim 1, characterized in that, The process of fusing the image displacement velocity, the road feature displacement velocity, and the inertial velocity based on the fusion weights includes: One or more algorithms, including adaptive weighted Kalman filtering, weighted least squares, and weighted averaging, are used to fuse the image displacement velocity, the road feature displacement velocity, and the inertial velocity based on the fusion weights.

8. A contactless speed measuring device for vehicles, characterized in that, include: The image acquisition unit is used to acquire continuous road surface images during vehicle movement; An inertial measurement unit is used to collect inertial measurement data of the vehicle. The illumination information acquisition unit is used to acquire ambient illumination parameters; and The processing unit is configured to obtain image displacement velocity, road feature displacement velocity, and texture richness parameters based on the continuous road surface images; obtain inertial velocity and vibration amplitude parameters based on the inertial measurement data; determine the fusion weights of the image displacement velocity, road feature displacement velocity, and inertial velocity based on the ambient illumination parameters, texture richness parameters, and vibration amplitude parameters; perform fusion processing on the image displacement velocity, road feature displacement velocity, and inertial velocity based on the fusion weights, and output the vehicle's driving speed.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-7.

Images