Braking performance tester based on optical flow sensor

Abstract

This invention discloses a braking performance tester based on an optical flow sensor, belonging to the field of braking performance testing technology. It includes an optical flow sensor component and a portable computing terminal. The optical flow sensor component collects and preprocesses road surface image data, acceleration data, vehicle pitch rate data, and vertical height data during vehicle movement. The portable computing terminal includes a feature calculation module, an optical flow correction module, an inertial estimation module, and a braking evaluation module. This invention solves the detection problems caused by missing road surface texture, insufficient lighting, braking pitch error, mechanical vibration interference, and mud / water obstruction under complex working conditions by integrating multiple units of the optical flow sensor component with multiple modules of the portable computing terminal, combined with active texture projection, geometric attitude compensation, and adaptive weight allocation strategies. This achieves all-weather, high-precision, and high-reliability testing of motor vehicle braking performance, and is portable and easy to operate.

Description

Technical Field

[0001] This invention belongs to the field of braking performance testing technology, specifically relating to a braking performance testing instrument based on an optical flow sensor. Background Technology

[0002] Brake performance testing equipment is used to quantitatively measure the kinematic parameters of a motor vehicle during braking to evaluate the performance of the vehicle's braking system. It mainly consists of sensor components, a signal acquisition circuit, and a microprocessor unit. The sensor components are responsible for sensing the vehicle's speed changes and displacement on the road surface in real time, converting physical motion into electrical signals. The signal acquisition circuit filters and amplifies the electrical signals before transmitting them to the microprocessor unit. The microprocessor unit calculates the data according to kinematic formulas and outputs key indicators such as initial braking speed, braking distance, braking time, and average deceleration. Brake performance testing equipment is a core device on motor vehicle safety inspection lines, widely used in vehicle manufacturing companies' off-line testing, annual inspections at motor vehicle inspection stations, and on-site enforcement by agricultural machinery safety supervision agencies. It objectively reflects the mechanical performance of the vehicle's braking system, ensuring vehicle driving safety.

[0003] Existing technologies, while employing laser ranging, contact wheel speed sensors, and real-time dynamic differential positioning via GPS to address the difficulty of obtaining vehicle braking performance parameters under complex working conditions, all suffer from significant technical limitations due to their respective measurement principles. These limitations make it difficult to meet the all-weather, high-precision requirements for agricultural machinery inspection. Laser ranging technology has stringent requirements on road surface smoothness and vehicle straight-line stability, and is prone to losing the measured target. Contact wheel speed sensors have a cumbersome installation process, and their measurement results depend on the friction between the detection wheel and the ground, making them susceptible to measurement distortion on wet, slippery, or soft surfaces. GPS is easily obstructed by trees, buildings, etc., leading to signal loss or drift, and its low data update frequency makes it unable to accurately capture high-speed dynamic braking processes.

[0004] These existing technological shortcomings ultimately manifest themselves as follows: they cannot adapt to testing conditions with missing road surface texture features and insufficient ambient light intensity; they cannot eliminate geometric measurement errors introduced by pitch motion during emergency braking of vehicles; they cannot suppress signal interference caused by high-frequency mechanical vibration of agricultural machinery; and they cannot avoid detection failures caused by sensor lenses being blocked by mud and water.

[0005] Therefore, it is possible to consider improving the braking performance tester based on optical flow sensors to solve the above-mentioned defects. Summary of the Invention

[0006] The purpose of this invention is to provide a braking performance tester based on an optical flow sensor, which can achieve all-weather, high-precision, and high-reliability testing of motor vehicle braking performance, and is portable and easy to operate.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A braking performance tester based on an optical flow sensor includes an optical flow sensing component and a portable computing terminal. The optical flow sensing component collects and preprocesses road surface image data, acceleration data, vehicle pitch rate data, and vertical height data during vehicle movement. The portable computing terminal has the following built-in features:

[0009] The feature calculation module, based on the preprocessed road surface image data, uses the Laplacian operator to generate a visual confidence factor and uses a corner detection algorithm to extract optical flow vectors. It identifies the vibration dominant frequency of the preprocessed acceleration data through fast Fourier transform to generate a vibration interference factor, and integrates the preprocessed vehicle pitch rate data to obtain the vehicle pitch angle.

[0010] The optical flow correction module uses a trigonometric transformation matrix to correct the optical flow vector projection error based on the preprocessed vertical height data and the vehicle body pitch angle, thus obtaining the geometric compensation speed.

[0011] The inertial estimation module integrates the preprocessed acceleration data to obtain the inertial estimation velocity.

[0012] The braking assessment module dynamically allocates visual and inertial weights based on visual confidence factors and vibration interference factors. It calculates the real-time vehicle speed by weighted summation of geometric compensation speed and inertial estimation speed, and performs integration to obtain the braking distance. The braking distance is then compared with a preset braking distance threshold to determine whether the braking performance is qualified.

[0013] Preferably, the optical flow sensing component consists of a high frame rate imaging unit, a structured light projection unit, a multispectral illumination unit, a laser ranging unit, and a microelectromechanical inertial measurement unit. It adopts an integrated packaging structure and is adsorbed onto the chassis of the vehicle under test, including a protective housing and a magnetic mounting base set on the top of the protective housing.

[0014] The high frame rate imaging unit is located at the center of the bottom of the protective housing and is used to acquire road surface image data containing artificial texture patterns;

[0015] The structured light projection unit and the multispectral illumination unit are symmetrically arranged on both sides of the high frame rate imaging unit, and are used to project artificial texture patterns onto the road surface and provide a constant lighting environment, respectively.

[0016] The laser ranging unit is positioned adjacent to the high frame rate imaging unit and is used to collect vertical height data relative to the road surface. The working windows of the high frame rate imaging unit, structured light projection unit, multispectral illumination unit, and laser ranging unit all face the ground.

[0017] The microelectromechanical inertial measurement unit is encapsulated inside a protective housing and closely attached to the back of the circuit board of the high frame rate imaging unit, and is used to collect acceleration data and vehicle pitch rate data;

[0018] The portable computing terminal includes a human-computer interaction touch screen and an alarm buzzer, and is connected to an optical flow sensing component via a wireless link.

[0019] Preferably, the process by which the optical flow sensing component acquires and preprocesses road surface image data, acceleration data, vehicle pitch rate data, and vertical height data during vehicle motion includes:

[0020] The optical flow sensor component is magnetically attached to a rigid plane near the centerline of the chassis of the vehicle under test. The installation posture is adjusted so that the working windows of the high frame rate imaging unit, structured light projection unit, multispectral illumination unit and laser ranging unit are perpendicular to the ground. The measurement axis of the microelectromechanical inertial measurement unit is aligned with the body coordinate system axis of the vehicle under test. The portable computing terminal is placed in the driver's cab of the vehicle under test to establish a data transmission link with the optical flow sensor component.

[0021] The multispectral illumination unit and structured light projection unit are turned on to emit an illumination beam with constant light power and project an artificial texture pattern with pseudo-random speckle characteristics into the field of view of the high frame rate imaging unit.

[0022] The high frame rate imaging unit continuously exposes at a preset image sampling frequency to capture road surface images containing artificial texture patterns, collects road surface image data, and uses a Gaussian smoothing filter algorithm to suppress shot noise from the image sensor and generate preprocessed road surface image data.

[0023] The microelectromechanical inertial measurement unit measures the linear acceleration signal along the driving direction and the angular rate signal around the lateral axis of the vehicle under test according to the preset inertial sampling frequency. These signals are used as acceleration data and vehicle pitch rate data, respectively. The infinite impulse response low-pass filtering algorithm is used to filter out high-frequency mechanical vibration noise and generate pre-processed acceleration data and pre-processed vehicle pitch rate data, respectively.

[0024] The laser ranging unit measures the instantaneous vertical distance between the optical flow sensing component and the road surface according to the preset ranging sampling frequency, which is used as the vertical height data. The sliding window midpoint filtering algorithm is used to remove sudden pulse interference caused by uneven road surface, and the preprocessed vertical height data is generated.

[0025] Preferably, in the feature calculation module, the process of generating a visual confidence factor using the Laplacian operator and extracting the optical flow vector using a corner detection algorithm based on the preprocessed road surface image data includes:

[0026] The Laplacian convolution kernel is used to perform sliding convolution on the preprocessed road surface image data to calculate the second derivative of the image pixel gray value and obtain the image edge gradient map.

[0027] Calculate the variance of pixel values ​​in the image edge gradient map to obtain the image edge energy value, and map the image edge energy value to a dimensionless interval of 0 to 1 as a visual confidence factor;

[0028] The Harris corner detection algorithm is used to traverse the preprocessed road surface image data, calculate the corner response function value of each pixel, and select pixels with corner response function values ​​greater than a preset threshold as strong feature points.

[0029] Using the Pyramid Lucas-Cannard optical flow algorithm, strong feature points are tracked between two consecutive preprocessed road surface image data, and optical flow constraint equations are constructed. The overdetermined equation system composed of the optical flow constraint equations of each pixel in the neighborhood of the strong feature point is solved by the least squares method, and the horizontal and vertical displacement components of the strong feature points in the image pixel coordinate system are obtained. These components are then combined to form the optical flow vector.

[0030] Preferably, in the feature calculation module, the process of identifying the dominant vibration frequency of the preprocessed acceleration data through fast Fourier transform to generate a vibration interference factor, and integrating the preprocessed vehicle pitch rate data to obtain the vehicle pitch angle includes:

[0031] Preprocessed acceleration data of a preset time window length is extracted to construct a time-domain acceleration sequence;

[0032] The time-domain acceleration sequence is converted into a frequency-domain power spectral density sequence using the discrete fast Fourier transform algorithm. The maximum amplitude point in the frequency-domain power spectral density sequence is searched, and the frequency corresponding to the maximum amplitude point is determined as the vibration dominant frequency.

[0033] The amplitude energy corresponding to the dominant vibration frequency is converted into a value in the closed interval between 0 and 1 using a normalized mapping function, which serves as the vibration disturbance factor.

[0034] The initial pitch angle at the moment braking begins is obtained. The trapezoidal numerical integration algorithm is used to perform a cumulative integral operation on the preprocessed vehicle pitch rate data with respect to time. The cumulative integral operation result is added to the initial pitch angle to obtain the vehicle pitch angle at the current moment.

[0035] Preferably, in the optical flow correction module, the process of correcting the optical flow vector projection error based on the preprocessed vertical height data and vehicle body pitch angle using a trigonometric geometric transformation matrix to obtain the geometric compensation speed includes:

[0036] Using the reciprocal of the cosine of the vehicle pitch angle as the longitudinal correction factor, a triangular geometric transformation matrix with the longitudinal correction factor as the diagonal element is constructed.

[0037] Obtain the pre-calibrated equivalent focal length parameters and optical flow vectors extracted by the feature calculation module in the high frame rate imaging unit;

[0038] The preprocessed vertical height data is divided by the equivalent focal length parameter of the high frame rate imaging unit to obtain the scaling factor from the image pixel coordinate system to the road surface physical coordinate system. The scaling factor, the triangular geometric transformation matrix and the optical flow vector are then multiplied together to convert the optical flow motion in the image plane into the geometric compensation velocity in the road surface plane.

[0039] Preferably, in the inertial estimation module, the process of integrating the preprocessed acceleration data to obtain the inertial estimated velocity includes:

[0040] Read the inertial velocity calculated at the previous sampling time. If the current time is the start time of the detection cycle, retrieve the geometric compensation velocity at the current time as the initial value of the inertial velocity.

[0041] Using a numerical integration algorithm, the product of the preprocessed acceleration data at the current sampling time and the inertial sampling time interval is calculated to obtain the velocity increment of the current sampling period. The velocity increment is then added to the inertial inference velocity at the previous sampling time to obtain the inertial inference velocity at the current sampling time.

[0042] Preferably, in the braking assessment module, the process of dynamically allocating visual weights and inertial weights based on visual confidence factors and vibration interference factors includes:

[0043] An adaptive weight generation model based on the Sigmoid function is constructed. The difference between the visual confidence factor and the vibration interference factor is used as the input variable of the adaptive weight generation model. The visual weight is calculated by nonlinear mapping and the value between 0 and 1 is used as the visual weight. The inertial weight is obtained by subtracting the visual weight from 1, ensuring that the sum of the visual weight and the inertial weight at the same time is equal to 1.

[0044] Preferably, in the braking assessment module, the process of calculating the real-time vehicle speed by weighted summation of the geometrically compensated speed and the inertial-calculated speed, and then integrating the results to obtain the braking distance, includes:

[0045] Using the calculated visual weight and inertial weight, the product of geometric compensation speed and visual weight, and the product of inertial calculated speed and inertial weight are calculated respectively. The two product results are added together to obtain the real-time speed of the vehicle.

[0046] By monitoring the real-time speed change characteristics of the vehicle to determine the braking start time and the vehicle stopping time, a numerical integration algorithm is used to perform definite integral calculation on the real-time speed of the vehicle over the time interval from the braking start time to the vehicle stopping time. The area under the speed-time curve is calculated to obtain the braking distance.

[0047] Preferably, in the braking assessment module, the process of comparing the braking performance with a preset braking distance threshold to determine whether the braking performance is qualified includes:

[0048] Read the vehicle type and initial braking speed of the vehicle under test, and retrieve the braking distance threshold corresponding to the vehicle type and initial braking speed from the built-in standard database;

[0049] Calculate the difference between the braking distance and the retrieved braking distance threshold. If the braking distance is less than or equal to the braking distance threshold, generate a braking performance qualified judgment signal. If the braking distance is greater than the braking distance threshold, generate a braking performance unqualified judgment signal.

[0050] The portable computing terminal receives the judgment signal. If it receives a qualified braking performance judgment signal, it displays a green pass mark on the human-machine interaction touch screen. If it receives a failed braking performance judgment signal, it drives the alarm buzzer to sound an alarm and displays a red warning mark and the value exceeding the standard on the human-machine interaction touch screen.

[0051] The beneficial effects of this invention are:

[0052] This invention integrates a structured light projection unit and a multispectral illumination unit into the optical flow sensing component. The structured light projection unit projects artificial texture patterns onto the road surface, while the multispectral illumination unit provides a constant illumination environment. Combined with a high frame rate imaging unit, it acquires road surface image data containing artificial texture patterns. This solves the problem that existing detection equipment cannot adapt to the lack of road surface texture features and insufficient ambient light intensity, ensuring the stability of optical flow feature extraction of the brake performance tester under complex outdoor lighting conditions and different road surface materials.

[0053] This invention acquires vertical height data through a laser ranging unit, acquires vehicle pitch rate data through a microelectromechanical inertial measurement unit and integrates it to obtain the vehicle pitch angle, and uses a trigonometric geometric transformation matrix to correct the optical flow vector projection error based on the preprocessed vertical height data and vehicle pitch angle, and calculates the geometrically compensated speed. This eliminates the geometric measurement error introduced by the vehicle's pitch motion during emergency braking, and improves the accuracy of speed calculation and distance measurement of the braking performance tester under dynamic attitude change conditions of the vehicle.

[0054] This invention utilizes a feature calculation module to generate a visual confidence factor using the Laplace operator to characterize the degree of lens occlusion. It also employs a fast Fourier transform to identify the dominant vibration frequency of preprocessed acceleration data to generate a vibration interference factor. The braking evaluation module dynamically allocates visual and inertial weights based on the visual confidence factor and the vibration interference factor, and performs weighted fusion of geometric compensation speed and inertial calculation speed. This suppresses signal interference caused by high-frequency mechanical vibrations in agricultural machinery, avoids detection failures caused by sensor lenses being obstructed by mud and water, and ensures the data reliability of the braking performance tester under harsh operating environments. Attached Figure Description

[0055] Figure 1 This is a schematic diagram illustrating the overall application scenario of the present invention;

[0056] Figure 2 This is a bottom-view structural diagram of the optical flow sensing component in this invention;

[0057] Figure 3 This is a schematic diagram of the module connections of the portable computing terminal in this invention;

[0058] Figure 4 This is a simulation diagram of environmental perception and feature calculation during the verification process of this invention. Figure 4 (a) in the figure is a simulation diagram of environmental perception and feature calculation under normal working conditions; Figure 4 (b) in the figure is a simulation diagram of environmental perception and feature calculation under the condition of mud splashing;

[0059] Figure 5 This is a simulation diagram of the adaptive weight allocation strategy during the verification process of this invention; Figure 5 (a) in the figure is a simulation diagram of the adaptive weight allocation strategy under visual failure state; Figure 5 (b) in the figure is a simulation diagram of the adaptive weight allocation strategy under strong vibration conditions; Figure 5 (c) in the figure is a simulation diagram of the adaptive weight allocation strategy under normal working conditions;

[0060] Figure 6 This is a simulation diagram showing the comparison of multi-source speed data and the calculation of braking distance during the verification process of this invention.

[0061] In the diagram: 1. Optical flow sensing component; 2. Portable computing terminal; 3. Wireless link; 4. Artificial texture pattern; 5. Field of view; 6. Magnetic mounting base; 7. Protective housing; 8. High frame rate imaging unit; 9. Structured light projection unit; 10. Multispectral illumination unit; 11. Laser ranging unit; 12. Microelectromechanical inertial measurement unit. Detailed Implementation

[0062] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0063] like Figure 1-3 As shown, the braking performance tester based on optical flow sensor includes an optical flow sensing component 1 and a portable computing terminal 2. The optical flow sensing component 1 collects and preprocesses road surface image data, acceleration data, vehicle pitch rate data and vertical height data when the vehicle is moving.

[0064] The optical flow sensing component 1 consists of a high frame rate imaging unit 8, a structured light projection unit 9, a multispectral illumination unit 10, a laser ranging unit 11, and a microelectromechanical inertial measurement unit 12. It adopts an integrated packaging structure and is attached to the chassis of the vehicle under test. It includes a protective housing 7 and a magnetic mounting base 6 set on the top of the protective housing 7.

[0065] The high frame rate imaging unit 8 is located at the center of the bottom of the protective housing 7 and is used to acquire road surface image data containing artificial texture patterns 4.

[0066] The structured light projection unit 9 and the multispectral illumination unit 10 are symmetrically arranged on both sides of the high frame rate imaging unit 8, and are used to project artificial texture patterns 4 onto the road surface and provide a constant lighting environment, respectively.

[0067] The laser ranging unit 11 is positioned adjacent to the high frame rate imaging unit 8 and is used to collect vertical height data relative to the road surface. The working windows of the high frame rate imaging unit 8, the structured light projection unit 9, the multispectral illumination unit 10, and the laser ranging unit 11 all face the ground.

[0068] The microelectromechanical inertial measurement unit 12 is encapsulated inside the protective housing 7 and closely attached to the back of the circuit board of the high frame rate imaging unit 8, and is used to collect acceleration data and vehicle pitch rate data.

[0069] The portable computing terminal 2 includes a human-computer interaction touch screen and an alarm buzzer, and is connected to the optical flow sensing component 1 via a wireless link 3.

[0070] The optical flow sensing component 1 is attached to a rigid plane near the centerline of the chassis of the vehicle under test using the magnetic mounting base 6. The installation posture is adjusted so that the working windows of the high frame rate imaging unit 8, the structured light projection unit 9, the multispectral illumination unit 10, and the laser ranging unit 11 are perpendicular to the ground. The measurement axis of the microelectromechanical inertial measurement unit 12 is aligned with the body coordinate system axis of the vehicle under test. The portable computing terminal 2 is placed in the driver's cab of the vehicle under test to establish a data transmission link with the optical flow sensing component 1.

[0071] The multispectral illumination unit 10 and the structured light projection unit 9 are turned on to emit a constant light power illumination beam to eliminate the shadow area under the chassis and project an artificial texture pattern 4 with pseudo-random speckle characteristics into the field of view 5 of the high frame rate imaging unit 8 to enhance the contrast of the road surface texture.

[0072] The high frame rate imaging unit 8 continuously exposes according to the preset image sampling frequency, captures the road surface containing the artificial texture pattern 4, collects road surface image data, and uses a Gaussian smoothing filter algorithm to suppress the shot noise of the image sensor and generate preprocessed road surface image data.

[0073] The microelectromechanical inertial measurement unit 12 measures the linear acceleration signal along the driving direction and the angular rate signal around the transverse axis of the vehicle under test according to the preset inertial sampling frequency. These signals are used as acceleration data and vehicle pitch rate data, respectively. The infinite impulse response low-pass filtering algorithm is used to filter out high-frequency mechanical vibration noise, and preprocessed acceleration data and preprocessed vehicle pitch rate data are generated, respectively.

[0074] The laser ranging unit 11 measures the instantaneous vertical distance between the optical flow sensing component 1 and the road surface according to the preset ranging sampling frequency, and uses it as vertical height data. The sliding window midpoint filtering algorithm is used to remove sudden pulse interference caused by uneven road surface, and preprocessed vertical height data is generated.

[0075] Portable computing terminal 2 has the following built-in features:

[0076] The feature calculation module, based on the preprocessed road surface image data, uses the Laplacian operator to generate a visual confidence factor and uses a corner detection algorithm to extract optical flow vectors. It identifies the vibration dominant frequency of the preprocessed acceleration data through fast Fourier transform to generate a vibration interference factor, and integrates the preprocessed vehicle pitch rate data to obtain the vehicle pitch angle.

[0077] The preprocessed road surface image data is subjected to sliding convolution operation using a Laplacian convolution kernel to calculate the second derivative of the image pixel grayscale value and obtain the image edge gradient map.

[0078] The variance of pixel values ​​in the image edge gradient map is calculated to obtain the image edge energy value. The image edge energy value is then mapped to a dimensionless interval of 0 to 1 and used as a visual confidence factor.

[0079] The Harris corner detection algorithm is used to traverse the preprocessed road surface image data, calculate the corner response function value of each pixel, and select pixels with corner response function values ​​greater than a preset threshold as strong feature points.

[0080] Using the Pyramid Lucas-Cannard optical flow algorithm, strong feature points are tracked between two consecutive preprocessed road surface image data, and optical flow constraint equations are constructed. The overdetermined equation system composed of the optical flow constraint equations of each pixel in the neighborhood of the strong feature point is solved by the least squares method, and the horizontal and vertical displacement components of the strong feature points in the image pixel coordinate system are obtained. These components are then combined to form the optical flow vector.

[0081] Preprocessed acceleration data of a preset time window length is extracted to construct a time-domain acceleration sequence.

[0082] The time-domain acceleration sequence is converted into a frequency-domain power spectral density sequence using the discrete fast Fourier transform algorithm. The maximum amplitude point in the frequency-domain power spectral density sequence is searched, and the frequency corresponding to the maximum amplitude point is determined as the dominant vibration frequency.

[0083] The amplitude energy corresponding to the dominant vibration frequency is converted into a value in the closed interval from 0 to 1 using a normalized mapping function, which serves as the vibration disturbance factor.

[0084] The initial pitch angle at the moment braking begins is obtained. The trapezoidal numerical integration algorithm is used to perform a cumulative integral operation on the preprocessed vehicle pitch rate data with respect to time. The cumulative integral operation result is added to the initial pitch angle to obtain the vehicle pitch angle at the current moment.

[0085] The optical flow correction module uses a trigonometric transformation matrix to correct the optical flow vector projection error based on the preprocessed vertical height data and the vehicle body pitch angle, thereby obtaining the geometric compensation speed.

[0086] Using the reciprocal of the cosine of the vehicle body pitch angle as a longitudinal correction factor, a triangular geometric transformation matrix with the longitudinal correction factor as the diagonal element is constructed to specifically compensate for the perspective shortening deformation caused by the vehicle body tilt in the driving direction.

[0087] Obtain the pre-calibrated equivalent focal length parameters and optical flow vectors extracted by the feature calculation module in the high frame rate imaging unit 8.

[0088] The preprocessed vertical height data is divided by the equivalent focal length parameter of the high frame rate imaging unit 8 to obtain the scaling factor from the image pixel coordinate system to the road surface physical coordinate system. The scaling factor, the triangular geometric transformation matrix and the optical flow vector are then multiplied together to convert the optical flow motion in the image plane into the geometric compensation velocity in the road surface plane.

[0089] The inertial estimation module integrates the preprocessed acceleration data to obtain the inertial estimation velocity.

[0090] Read the inertial velocity calculated at the previous sampling time. If the current time is the start time of the detection cycle, retrieve the geometric compensation velocity at the current time as the initial value of the inertial velocity.

[0091] Using a numerical integration algorithm, the product of the preprocessed acceleration data at the current sampling time and the inertial sampling time interval is calculated to obtain the velocity increment of the current sampling period. The velocity increment is then added to the inertial inference velocity at the previous sampling time to obtain the inertial inference velocity at the current sampling time.

[0092] The braking assessment module dynamically allocates visual and inertial weights based on visual confidence factors and vibration interference factors. It calculates the real-time vehicle speed by weighted summation of geometric compensation speed and inertial estimation speed, and performs integration to obtain the braking distance. The braking distance is then compared with a preset braking distance threshold to determine whether the braking performance is qualified.

[0093] An adaptive weight generation model based on the Sigmoid function is constructed. The difference between the visual confidence factor and the vibration interference factor is used as the input variable of the adaptive weight generation model. The visual weight is calculated by nonlinear mapping and the value between 0 and 1 is used as the visual weight. The inertial weight is obtained by subtracting the visual weight from 1, ensuring that the sum of the visual weight and the inertial weight at the same time is equal to 1.

[0094] Using the calculated visual weights and inertial weights, the product of the geometric compensation speed and the visual weights, and the product of the inertial calculated speed and the inertial weights are calculated respectively. The two product results are added together to obtain the real-time speed of the vehicle.

[0095] By monitoring the real-time speed change characteristics of the vehicle to determine the braking start time and the vehicle stopping time, a numerical integration algorithm is used to perform definite integral calculation on the real-time speed of the vehicle over the time interval from the braking start time to the vehicle stopping time. The area under the speed-time curve is calculated to obtain the braking distance.

[0096] The system reads the vehicle type and initial braking speed of the tested vehicle, and retrieves the braking distance threshold corresponding to the vehicle type and initial braking speed from the built-in standard database.

[0097] Calculate the difference between the braking distance and the retrieved braking distance threshold. If the braking distance is less than or equal to the braking distance threshold, generate a braking performance qualified judgment signal. If the braking distance is greater than the braking distance threshold, generate a braking performance unqualified judgment signal.

[0098] Portable computing terminal 2 receives the judgment signal. If it receives a braking performance qualified judgment signal, it displays a green pass mark on the human-machine interaction touch screen. If it receives a braking performance unqualified judgment signal, it drives the alarm buzzer to sound an alarm and displays a red warning mark and the value exceeding the standard on the human-machine interaction touch screen.

[0099] In summary, the complete testing process of the braking performance tester based on the optical flow sensor in this embodiment is as follows:

[0100] Before using the brake performance tester based on optical flow sensors, the optical flow sensor component 1 is first physically installed. Using the magnetic mounting base 6 set on the top of the protective housing 7, the optical flow sensor component 1 is adsorbed and fixed on a rigid plane near the centerline of the chassis of the vehicle under test. During the fixing process, the installation posture of the optical flow sensor component 1 is adjusted to ensure that the working windows of the high frame rate imaging unit 8, structured light projection unit 9, multispectral illumination unit 10 and laser ranging unit 11 inside the protective housing 7 are all perpendicular to the road surface, and the measurement axis of the microelectromechanical inertial measurement unit 12 encapsulated inside the protective housing 7 is aligned with the body coordinate system axis of the vehicle under test to ensure the consistency of the direction of motion data acquisition.

[0101] Subsequently, the portable computing terminal 2 is placed in a stable position in the driver's cab of the vehicle under test for the driver to observe and operate. The portable computing terminal 2 is then started and a communication connection is established with the optical flow sensor component 1 under the chassis via the wireless link 3 to complete the initialization of the data transmission channel.

[0102] After the detector is powered on, the optical flow sensing component 1 automatically enters the working state, and the multispectral illumination units 10 located on both sides of the high frame rate imaging unit 8 are turned on, emitting a constant light power illumination beam to the road surface, eliminating the shadow area formed by the vehicle body under the chassis, and creating a constant lighting environment. At the same time, the structured light projection unit 9 is turned on, projecting an artificial texture pattern 4 with pseudo-random speckle characteristics into the field of view 5 of the high frame rate imaging unit 8, artificially enhancing the contrast of the road surface texture in the absence of road surface texture features, and providing a feature basis for visual measurement.

[0103] During the driving and braking process of the vehicle under test, the optical flow sensor component 1 performs multi-source data acquisition tasks. The high frame rate imaging unit 8, located at the bottom center of the protective housing 7, continuously exposes through the working window to capture road surface image data including the artificial texture pattern 4. The microelectromechanical inertial measurement unit 12, which is close to the back of the circuit board of the high frame rate imaging unit 8, collects the vehicle acceleration data and vehicle pitch rate data in real time. The laser ranging unit 11, which is located next to the high frame rate imaging unit 8, measures the vertical height data of the optical flow sensor component 1 relative to the road surface in real time. The above data is transmitted to the portable computing terminal 2 in the driver's cab for processing via the wireless link 3.

[0104] Portable computing terminal 2 receives and processes the collected data to determine whether the braking performance is qualified. If qualified, the human-machine interface touch screen of portable computing terminal 2 displays a green pass mark. If unqualified, portable computing terminal 2 drives the built-in alarm buzzer to sound an alarm and displays a red warning mark and the value exceeding the standard on the human-machine interface touch screen to prompt the tester that the braking performance is abnormal. After the test is completed, the optical flow sensor component 1 is removed from the chassis of the vehicle under test to complete the removal.

[0105] The following verification example evaluates the detection performance of the brake performance tester based on the optical flow sensor:

[0106] like Figure 4-6 As shown, a wheeled tractor with an initial braking velocity of 12 m / s was used as the test object, and an unstructured road with superimposed random mud and water obstruction interference, high-frequency mechanical vibration noise, and braking head-diving effect was used as the test scenario. The simulation results are as follows. Figures 4 to 6 As shown.

[0107] The simulation experiment can be implemented based on a combination of platforms such as Matlab / Simulink, supplemented by tools such as CarSim, OpenCV, and Python (NumPy / SciPy). By building vehicle dynamics models (simulating motion states such as initial braking velocity and pitch attitude) and working condition disturbance models (simulating complex scenarios such as mud and water obstruction, mechanical vibration, and insufficient lighting) on ​​the platform, core algorithms such as optical flow extraction, geometric attitude compensation, and adaptive weight allocation are integrated to complete the decomposition and collaborative integration of each module, test scenario parameter setting, simulation data generation, and result analysis. Finally, the effectiveness of the algorithm and the system detection performance under complex working conditions are verified.

[0108] Figure 4 The imaging status and sensing results of the optical flow sensor 1 under different environmental conditions are visually demonstrated. Figure 5 The dynamic response process of the adaptive weight allocation strategy within the braking assessment module is demonstrated. Figure 6 The comparison curves of multi-source velocity data and the calculation results of braking distance are shown in the full-process simulation.

[0109] Figure 4 This refers to the environmental perception and feature calculation stage in this embodiment. Figure 4 (a) shows that under normal operating conditions, the structured light projection unit 9 in the optical flow sensing component 1 projects a clear artificial texture pattern 4 onto the road surface. The visual confidence factor extracted by the corresponding feature calculation module is at a high level (close to 1.0), and the detector is in a stable observation state. Figure 4 (b) shows the scene when a vehicle passes through a muddy road and mud splashes up and obscures the lens of the high frame rate imaging unit 8. At this time, the image texture features are severely lacking and the real-time generated visual confidence factor drops sharply to 0.1 or below, which verifies the sensitivity of the detector to harsh environments.

[0110] Figure 5 For the adaptive weighting stage of the braking assessment module, Figure 5(a) in the diagram demonstrates how, in a state of "visual failure (such as mud and water obstruction)," the visual weight is automatically adjusted to an extremely low value (0.1 or below), while the inertial weight is elevated to a dominant value (0.9 or above), thereby seamlessly switching to inertial extrapolation mode to maintain velocity output. Figure 5 (b) illustrates that under "strong vibration" interference, the feature calculation module identifies the dominant vibration frequency and generates a higher vibration interference factor, thereby reducing the visual weight to suppress optical flow noise caused by high frame rate imaging unit jitter. Figure 5 (c) in the figure demonstrates that under “normal working” conditions, a higher visual weight is assigned to make full use of the high precision of optical flow measurement, verifying the robustness of the detector’s adaptive weight assignment strategy under varying working conditions.

[0111] Figure 6 This is the stage for optical flow correction and braking evaluation. Figure 6 The solid black line represents the vehicle's actual speed change curve, the dashed line represents the original optical flow velocity curve without geometric attitude compensation, and the dashed line represents the vehicle's real-time speed curve after geometric correction and weighted fusion. It can be seen that during the emergency braking process 2.0 seconds later, the vehicle "nods" and the body pitch angle changes. The uncorrected dashed line is significantly lower than the actual speed (producing a negative measurement deviation), while the dashed line corrects the projection error in real time by introducing a triangular geometric transformation matrix, always closely following the actual speed curve. Finally, the braking distance calculated by integration is 13.80m, and after comparison with the braking distance threshold, the braking performance is judged to be "qualified".

[0112] The detector in this embodiment effectively solves the measurement error problems caused by missing road texture and braking pitch by combining structured light active texture projection and geometric attitude compensation. Through adaptive weighting based on visual confidence factor and vibration interference factor, it ensures data continuity and accuracy under mud and water obstruction and strong vibration interference.

[0113] In summary, the braking performance tester based on optical flow sensors proposed in this embodiment demonstrates excellent environmental adaptability and high-precision measurement capabilities under complex lighting and harsh road conditions in the field. At the same time, through multi-source information fusion and intelligent judgment logic, an all-weather, highly reliable agricultural machinery braking performance testing system is constructed, which has significant engineering application value.

Claims

1. A braking performance testing instrument based on an optical flow sensor, characterized in that, It includes an optical flow sensing component and a portable computing terminal. The optical flow sensing component collects and preprocesses road surface image data, acceleration data, vehicle pitch rate data and vertical height data when the vehicle is moving. The optical flow sensing component consists of a high frame rate imaging unit, a structured light projection unit, a multispectral illumination unit, a laser ranging unit, and a microelectromechanical inertial measurement unit. It adopts an integrated packaging structure and is attached to the chassis of the vehicle under test, including a protective housing and a magnetic mounting base set on the top of the protective housing. The high frame rate imaging unit is located at the center of the bottom of the protective housing and is used to acquire road surface image data containing artificial texture patterns; The structured light projection unit and the multispectral illumination unit are symmetrically arranged on both sides of the high frame rate imaging unit, and are used to project artificial texture patterns onto the road surface and provide a constant lighting environment, respectively. The laser ranging unit is positioned adjacent to the high frame rate imaging unit and is used to collect vertical height data relative to the road surface. The working windows of the high frame rate imaging unit, structured light projection unit, multispectral illumination unit, and laser ranging unit all face the ground. The microelectromechanical inertial measurement unit is encapsulated inside a protective housing and closely attached to the back of the circuit board of the high frame rate imaging unit, and is used to collect acceleration data and vehicle pitch rate data; The portable computing terminal includes a human-computer interaction touchscreen and an alarm buzzer, and is connected to an optical flow sensor via a wireless link. The portable computing terminal has the following built-in features: The feature calculation module, based on the preprocessed road surface image data, uses the Laplacian operator to generate a visual confidence factor and uses a corner detection algorithm to extract optical flow vectors. It identifies the vibration dominant frequency of the preprocessed acceleration data through fast Fourier transform to generate a vibration interference factor, and integrates the preprocessed vehicle pitch rate data to obtain the vehicle pitch angle. The optical flow correction module uses a trigonometric transformation matrix to correct the optical flow vector projection error based on the preprocessed vertical height data and the vehicle body pitch angle, thus obtaining the geometric compensation speed. The inertial estimation module integrates the preprocessed acceleration data to obtain the inertial estimation velocity. The braking assessment module dynamically allocates visual and inertial weights based on visual confidence factors and vibration interference factors. It calculates the real-time vehicle speed by weighted summation of geometric compensation speed and inertial estimation speed, and performs integration to obtain the braking distance. The braking distance is then compared with a preset braking distance threshold to determine whether the braking performance is qualified.

2. The braking performance testing instrument based on an optical flow sensor according to claim 1, characterized in that, The process by which the optical flow sensor acquires and preprocesses road surface image data, acceleration data, vehicle pitch rate data, and vertical height data during vehicle motion includes: The optical flow sensor component is magnetically attached to a rigid plane near the centerline of the chassis of the vehicle under test. The installation posture is adjusted so that the working windows of the high frame rate imaging unit, structured light projection unit, multispectral illumination unit and laser ranging unit are perpendicular to the ground. The measurement axis of the microelectromechanical inertial measurement unit is aligned with the body coordinate system axis of the vehicle under test. The portable computing terminal is placed in the driver's cab of the vehicle under test to establish a data transmission link with the optical flow sensor component. The multispectral illumination unit and structured light projection unit are turned on to emit an illumination beam with constant light power and project an artificial texture pattern with pseudo-random speckle characteristics into the field of view of the high frame rate imaging unit. The high frame rate imaging unit continuously exposes at a preset image sampling frequency to capture road surface images containing artificial texture patterns, collects road surface image data, and uses a Gaussian smoothing filter algorithm to suppress shot noise from the image sensor and generate preprocessed road surface image data. The microelectromechanical inertial measurement unit measures the linear acceleration signal along the driving direction and the angular rate signal around the lateral axis of the vehicle under test according to the preset inertial sampling frequency. These signals are used as acceleration data and vehicle pitch rate data, respectively. The infinite impulse response low-pass filtering algorithm is used to filter out high-frequency mechanical vibration noise and generate pre-processed acceleration data and pre-processed vehicle pitch rate data, respectively. The laser ranging unit measures the instantaneous vertical distance between the optical flow sensing component and the road surface according to the preset ranging sampling frequency, which is used as the vertical height data. The sliding window midpoint filtering algorithm is used to remove sudden pulse interference caused by uneven road surface, and the preprocessed vertical height data is generated.

3. The braking performance testing instrument based on an optical flow sensor according to claim 1, characterized in that, In the feature extraction module, the process of generating visual confidence factors using the Laplacian operator and extracting optical flow vectors using a corner detection algorithm based on preprocessed road surface image data includes: The Laplacian convolution kernel is used to perform sliding convolution on the preprocessed road surface image data to calculate the second derivative of the image pixel gray value and obtain the image edge gradient map. Calculate the variance of pixel values ​​in the image edge gradient map to obtain the image edge energy value, and map the image edge energy value to a dimensionless interval of 0 to 1 as a visual confidence factor; The Harris corner detection algorithm is used to traverse the preprocessed road surface image data, calculate the corner response function value of each pixel, and select pixels with corner response function values ​​greater than a preset threshold as strong feature points. Using the Pyramid Lucas-Cannard optical flow algorithm, strong feature points are tracked between two consecutive preprocessed road surface image data, and optical flow constraint equations are constructed. The overdetermined equation system composed of the optical flow constraint equations of each pixel in the neighborhood of the strong feature point is solved by the least squares method, and the horizontal and vertical displacement components of the strong feature points in the image pixel coordinate system are obtained. These components are then combined to form the optical flow vector.

4. The braking performance tester based on an optical flow sensor according to claim 1, characterized in that, In the feature calculation module, the vibration dominant frequency of the preprocessed acceleration data is identified by fast Fourier transform to generate a vibration interference factor. The process of integrating the preprocessed vehicle pitch rate data to obtain the vehicle pitch angle includes: Preprocessed acceleration data of a preset time window length is extracted to construct a time-domain acceleration sequence; The time-domain acceleration sequence is converted into a frequency-domain power spectral density sequence using the discrete fast Fourier transform algorithm. The maximum amplitude point in the frequency-domain power spectral density sequence is searched, and the frequency corresponding to the maximum amplitude point is determined as the vibration dominant frequency. The amplitude energy corresponding to the dominant vibration frequency is converted into a value in the closed interval from 0 to 1 using a normalized mapping function, which serves as the vibration disturbance factor. The initial pitch angle at the moment braking begins is obtained. The trapezoidal numerical integration algorithm is used to perform a cumulative integral operation on the preprocessed vehicle pitch rate data with respect to time. The cumulative integral operation result is added to the initial pitch angle to obtain the vehicle pitch angle at the current moment.

5. The braking performance testing instrument based on an optical flow sensor according to claim 1, characterized in that, In the optical flow correction module, the process of correcting the optical flow vector projection error and obtaining the geometric compensation speed by using a trigonometric geometric transformation matrix based on the preprocessed vertical height data and vehicle pitch angle includes: Using the reciprocal of the cosine of the vehicle pitch angle as the longitudinal correction factor, a triangular geometric transformation matrix with the longitudinal correction factor as the diagonal element is constructed. Obtain the pre-calibrated equivalent focal length parameters and optical flow vectors extracted by the feature calculation module in the high frame rate imaging unit; The preprocessed vertical height data is divided by the equivalent focal length parameter of the high frame rate imaging unit to obtain the scaling factor from the image pixel coordinate system to the road surface physical coordinate system. The scaling factor, the triangular geometric transformation matrix and the optical flow vector are then multiplied together to convert the optical flow motion in the image plane into the geometric compensation velocity in the road surface plane.

6. The braking performance tester based on an optical flow sensor according to claim 1, characterized in that, In the inertial estimation module, the process of integrating the preprocessed acceleration data to obtain the inertial estimated velocity includes: Read the inertial velocity calculated at the previous sampling time. If the current time is the start time of the detection cycle, retrieve the geometric compensation velocity at the current time as the initial value of the inertial velocity. Using a numerical integration algorithm, the product of the preprocessed acceleration data at the current sampling time and the inertial sampling time interval is calculated to obtain the velocity increment of the current sampling period. The velocity increment is then added to the inertial inference velocity at the previous sampling time to obtain the inertial inference velocity at the current sampling time.

7. The braking performance tester based on an optical flow sensor according to claim 1, characterized in that, In the braking assessment module, the process of dynamically allocating visual and inertial weights based on visual confidence factors and vibration interference factors includes: An adaptive weight generation model based on the Sigmoid function is constructed. The difference between the visual confidence factor and the vibration interference factor is used as the input variable of the adaptive weight generation model. The visual weight is calculated by nonlinear mapping and the value between 0 and 1 is used as the visual weight. The inertial weight is obtained by subtracting the visual weight from 1, ensuring that the sum of the visual weight and the inertial weight at the same time is equal to 1.

8. The braking performance tester based on an optical flow sensor according to claim 1, characterized in that, In the braking assessment module, the process of calculating the vehicle's real-time speed by weighted summation of the geometrically compensated speed and the inertial-calculated speed, and then integrating the results to obtain the braking distance, includes: Using the calculated visual weight and inertial weight, the product of geometric compensation speed and visual weight, and the product of inertial calculated speed and inertial weight are calculated respectively. The two product results are added together to obtain the real-time speed of the vehicle. By monitoring the real-time speed change characteristics of the vehicle to determine the braking start time and the vehicle stopping time, a numerical integration algorithm is used to perform definite integral calculation on the real-time speed of the vehicle over the time interval from the braking start time to the vehicle stopping time. The area under the speed-time curve is calculated to obtain the braking distance.

9. The braking performance tester based on an optical flow sensor according to claim 1, characterized in that, In the braking assessment module, the process of comparing the braking performance with a preset braking distance threshold to determine whether the braking performance is qualified includes: Read the vehicle type and initial braking speed of the vehicle under test, and retrieve the braking distance threshold corresponding to the vehicle type and initial braking speed from the built-in standard database; Calculate the difference between the braking distance and the retrieved braking distance threshold. If the braking distance is less than or equal to the braking distance threshold, generate a braking performance qualified judgment signal. If the braking distance is greater than the braking distance threshold, generate a braking performance unqualified judgment signal. The portable computing terminal receives the judgment signal. If it receives a qualified braking performance judgment signal, it displays a green pass mark on the human-machine interaction touch screen. If it receives a failed braking performance judgment signal, it drives the alarm buzzer to sound an alarm and displays a red warning mark and the value exceeding the standard on the human-machine interaction touch screen.

Images