Roadside sensor calibration method and device, electronic equipment and storage medium

Abstract

The invention relates to a roadside sensor calibration method and device, electronic equipment and a storage medium, and the method comprises the steps: obtaining a plurality of pieces of global positioning system (GPS) data and a plurality of pieces of sensor data in the driving process of a target vehicle, the GPS data being collected by a GPS system of the target vehicle, and the sensor data being obtained by tracking the target vehicle through a roadside sensor; time alignment processing is carried out on the GPS data and the sensor data, multiple groups of point pairs are determined, and each group of point pairs comprises the GPS data and the sensor data corresponding to the same time point; based on the mapping relation between the GPS data and the sensor data in each group of point pairs, first external parameters between the GPS system and the sensors are determined, and the first external parameters are used for calibrating the sensors. According to the method, calibration efficiency can be improved, regular and even dynamic calibration updating can be conveniently achieved, actual requirements in a complex roadside environment are met, and the method can be suitable for scenes where sensors are far in distance and different in orientation in an intelligent traffic system.

Description

Technical Field

[0001] The present disclosure relates to the fields of smart transportation, vehicles, and data processing technology, and in particular to a roadside sensor calibration method, device, electronic device, and storage medium. Background Art

[0002] A variety of sensors, such as lidar, millimeter-wave radar, and cameras, are deployed on the road infrastructure of intelligent transportation systems (ITS). These sensors collect data from different angles and dimensions. By calibrating the sensors and determining the relative spatial positions of each sensor, the sensor data can be fused and processed based on the external parameters between the sensors to obtain rich and accurate environmental information, providing strong support for applications such as traffic management and autonomous driving.

[0003] Traditional sensor calibration methods rely on static calibration objects or utilize static feature points in the environment. For example, when calibrating a camera and lidar, a checkerboard pattern is typically placed within the common field of view of both sensors. The pixel coordinates of the checkerboard pattern in the camera image data and the 3D coordinates in the lidar point cloud data are then extracted. By establishing a constraint relationship between the pixel coordinates and the 3D coordinates, the extrinsic parameters between the camera and lidar are determined.

[0004] However, this method is usually only applicable to laboratories or scenarios with large overlapping fields of view. In the scenario of intelligent transportation systems, due to the long distances or different orientations between sensors, their fields of view may have only small or even no overlapping areas. Moreover, complex and changeable roadside environmental factors, such as continuous vibration and drastic temperature changes, may cause the external parameters of the sensor to drift after installation, requiring regular or even dynamic recalibration. This makes the placement of static calibration objects extremely difficult and inefficient, making it difficult to meet actual needs. Summary of the Invention

[0005] The present disclosure provides a roadside sensor calibration system, method, device, electronic device, and storage medium to at least address the problem in related art that roadside multi-sensor calibration methods are difficult to operate and inefficient, making them difficult to meet practical needs. The technical solutions of the present disclosure are as follows:

[0006] According to a first aspect of an embodiment of the present disclosure, a roadside sensor calibration method is provided, comprising:

[0007] Acquiring multiple global positioning system (GPS) data and multiple sensor data of a target vehicle during its travel, wherein the GPS data is collected by the GPS system of the target vehicle, and the sensor data is obtained by tracking the target vehicle through roadside sensors;

[0008] Performing time alignment processing on the GPS data and the sensor data to determine a plurality of point pairs, each point pair including the GPS data and the sensor data corresponding to a same time point;

[0009] Based on the mapping relationship between the GPS data and the sensor data in each point pair, a first external parameter between the GPS system and the sensor is determined, where the first external parameter is used to calibrate the sensor.

[0010] According to a second aspect of an embodiment of the present disclosure, a roadside sensor calibration device is provided, comprising:

[0011] An acquisition module is used to acquire multiple global positioning system (GPS) data and multiple sensor data during the driving process of the target vehicle, wherein the GPS data is collected by the GPS system of the target vehicle, and the sensor data is obtained by tracking the target vehicle through roadside sensors;

[0012] a determination module, configured to perform time alignment processing on the GPS data and the sensor data to determine a plurality of point pairs, each point pair including the GPS data and the sensor data corresponding to a same time point;

[0013] The calibration module is used to determine a first external parameter between the GPS system and the sensor based on a mapping relationship between the GPS data and the sensor data in each point pair, wherein the first external parameter is used to calibrate the sensor.

[0014] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device for calibrating a roadside sensor, including:

[0015] processor;

[0016] a memory for storing instructions executable by the processor;

[0017] The processor is configured to execute the instructions to implement any one of the roadside sensor calibration methods described above.

[0018] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are executed by a processor of a roadside sensor calibration electronic device, the roadside sensor calibration electronic device is enabled to execute any one of the roadside sensor calibration methods described above.

[0019] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program / instruction, which implements any one of the roadside sensor calibration methods described above when executed by a processor.

[0020] The technical solutions provided by the embodiments of the present disclosure bring at least the following beneficial effects:

[0021] By utilizing the target vehicle as a dynamic reference object, there is no need to place static calibration objects, thus avoiding the reliance of traditional calibration methods on static calibration objects. By obtaining the GPS data and sensor data of the target vehicle during its driving process, point pairs are determined through time alignment processing to ensure that the vehicle's position (GPS data) and the vehicle's state observed by the sensor (sensor data) at the same time point can be correctly paired. Then, based on the point pair mapping relationship, the first external parameter between the GPS system and the sensor is determined. This not only improves the efficiency of calibration, but also facilitates regular and even dynamic calibration updates, effectively addressing the problem of sensor external parameter drift caused by factors such as environmental vibration and temperature changes, meeting the actual needs of complex roadside environments, and can be applied to scenarios in intelligent transportation systems where sensors are far apart and facing different directions.

[0022] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute an improper limitation of the present disclosure.

[0024] Figure 1 The figure is a flowchart of a roadside sensor calibration method according to an exemplary embodiment.

[0025] Figure 2 The figure is a block diagram of a roadside sensor calibration device according to an exemplary embodiment.

[0026] Figure 3 The present invention is a block diagram of an electronic device for roadside sensor calibration according to an exemplary embodiment.

[0027] Figure 4 The present invention is a block diagram of a device for calibrating a roadside sensor according to an exemplary embodiment. DETAILED DESCRIPTION

[0028] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0029] It should be noted that the terms "first," "second," and the like in the specification and claims of the present disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the numbers used in this manner are interchangeable where appropriate so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure as detailed in the appended claims.

[0030] Related technologies typically rely on static calibration objects or utilize static feature points in the environment. For example, in the calibration of cameras and lidars, a checkerboard pattern is typically placed within the common field of view of the two sensors. The pixel coordinates of the checkerboard pattern in the image data collected by the camera and the three-dimensional coordinates in the point cloud data collected by the lidar are then extracted. By establishing a constraint relationship between the pixel coordinates and the three-dimensional coordinates, the extrinsic parameters between the camera and lidar are solved.

[0031] However, in intelligent transportation system scenarios, due to the long distances between sensors or their different orientations, their fields of view may have little or no overlap. Furthermore, complex and variable roadside environmental factors, such as constant vibration and drastic temperature fluctuations, can cause the sensor's external parameters to drift after installation, requiring regular or even dynamic recalibration. This makes the placement of static calibration objects extremely difficult and inefficient, making it difficult to meet practical needs. Based on this, the roadside sensor calibration method in this application is proposed to address the above issues.

[0032] Figure 1 FIG. 1 is a flow chart showing a method for calibrating a roadside sensor according to an exemplary embodiment. Figure 1 As shown, the roadside sensor calibration method includes:

[0033] In step S11, multiple global positioning system GPS data and multiple sensor data are obtained during the driving process of the target vehicle. The GPS data is collected by the GPS system of the target vehicle, and the sensor data is obtained by tracking the target vehicle through sensors on the roadside.

[0034] In this step, a dynamically moving target vehicle is used as a natural reference object to collect the data required for sensor calibration, including the target vehicle's GPS (Global Positioning System) data and sensor data. Each GPS data and sensor data has a corresponding time point, which indicates the data collection time or reception time, and can exist in the form of a timestamp or time identifier.

[0035] Specifically, during data collection, the target vehicle must be controlled to travel a certain distance within the effective sensing range of the roadside sensor to be calibrated. The trajectory should cover the sensor's field of view as much as possible and include changes such as turns, acceleration, and deceleration to provide richer constraint information. Sensors can include radar sensors and image sensors. Radar sensors include lidar and millimeter-wave radar, and image sensors include cameras.

[0036] As the target vehicle drives, its GPS system continuously records its geographic location at different points in time. This location information is then used to generate GPS data, which describes the global position change of the target vehicle. For example, the GPS data corresponding to time point t can be expressed as P_GPS(t).

[0037] For example, the geographic location information of the target vehicle usually includes the longitude, latitude and altitude information in the geographic coordinate system. The geographic coordinate system is a curvilinear coordinate system based on the earth's ellipsoid. The coordinate values ​​are a mixed unit of angle (latitude and longitude) and length (altitude), which is not suitable for direct use in sensor data fusion. Therefore, the geographic location information can be further converted to a local Cartesian coordinate system, such as ENU (East-North-Up Coordinate System) or UTM (Universal Transverse Mercator Coordinate System) to obtain three-dimensional coordinate data in the local Cartesian coordinate system as GPS data. The local Cartesian coordinate system used by the GPS data can be used as the global reference coordinate system W.

[0038] At the same time, sensors deployed on the roadside can track and collect data on passing target vehicles based on their own field of view and observation capabilities, thereby generating sensor data of the target vehicles at different time points. The sensor data can describe the performance of the target vehicle within the local field of view of the roadside sensor.

[0039] For example, the sensor data can be the three-dimensional coordinate data in the lidar coordinate system S_Lidar, which can be expressed as PC_Lidar(t) at the time point t; it can also be the three-dimensional coordinate data in the millimeter-wave radar coordinate system S_Radar, which can be expressed as PC_Radar(t) at the time point t; it can also be the three-dimensional coordinate data in the image sensor coordinate system S_Camera, which can be expressed as Image_Cam(t) at the time point t.

[0040] It should be noted that the GPS system and sensors need to achieve time synchronization through the Network Time Protocol, Precision Time Protocol or shared GPS clock signals to ensure that the GPS system of the target vehicle and the roadside sensors have a unified high-precision time base, so that the timestamps or time identifiers determined based on the time base are comparable.

[0041] In this way, the sensor calibration process can get rid of its dependence on traditional static calibration objects, so that the calibration work can be completed in real traffic scenarios using the natural driving process of the vehicle, which greatly improves the convenience and applicability of calibration. It is especially suitable for complex roadside infrastructure environments where sensors are installed in fixed positions, with large spacing and limited field of view overlap.

[0042] In step S12, time alignment processing is performed on the GPS data and the sensor data to determine multiple groups of point pairs, each group of point pairs including GPS data and sensor data corresponding to the same time point.

[0043] The collection time of GPS data and sensor data may deviate and not completely correspond to the same time point. Therefore, after obtaining the GPS data and sensor data, the GPS data and sensor data can be time-aligned, and the GPS data and sensor data corresponding to the same time point can be matched to determine multiple groups of point pairs, each group of points including the GPS data and sensor data of the target vehicle at the same time point.

[0044] Specifically, based on the time points of GPS data and sensor data, data points that are close enough in time can be identified. For example, if the difference between the time points of certain GPS data and certain sensor data is less than a first threshold, the two data points can be regarded as a set of point pairs.

[0045] Alternatively, for each data point, another data point with the same time as the data point can be queried to form a set of point pairs. If no data point with the same time point is found, the missing data point at that time point can be estimated through data alignment to obtain a set of point pairs. For example, for a piece of GPS data, if a certain sensor data and the GPS data have the same time point, these two data points can be considered a set of point pairs. If no sensor data has the same time point as the GPS data, sensor data at other time points can be used, and data alignment methods such as interpolation or Kalman filtering can be used to generate the sensor data at that time point to obtain a set of point pairs.

[0046] In this way, multiple groups of point pairs are obtained, each of which represents the correspondence between the GPS data of the target vehicle in the global GPS coordinate system and the sensor data observed in the local coordinate system of the roadside sensor at a specific moment. This ensures that the data used for subsequent external parameter calculations are synchronized and corresponding in time, laying the foundation for accurately solving the spatial relationship between sensors.

[0047] In step S13, based on the mapping relationship between the GPS data and the sensor data in each set of point pairs, a first external parameter between the GPS system and the sensor is determined, and the first external parameter is used to calibrate the sensor.

[0048] As can be seen from the above, GPS data records the geographic location information of the target vehicle, while sensor data records the relative position and morphological information of the target vehicle within the sensor's perception range. In essence, they describe different representations of the same physical entity. The mapping relationship between the two can be described by rigid body transformation, that is, the sensor data is rotated and translated to match it with the GPS data.

[0049] Each point pair corresponds to the state of the target vehicle at a specific point in time. Therefore, based on these time-aligned point pairs, we can deeply analyze the mapping relationship between GPS data and sensor data. Using specific mathematical models and calculation methods, we can fit and solve the rotation matrix and translation vector, reducing interference caused by factors such as measurement errors. This allows us to determine the first extrinsic parameters between the GPS system and the sensor. These first extrinsic parameters are the rigid body transformation parameters between the GPS system and the sensor, including the rotation matrix and translation vector.

[0050] In this way, by optimizing and solving a large number of point pairs collected at different time points, a more robust and accurate first external parameter can be obtained. Moreover, its calibration characteristic based on the data collected during the driving process of the target vehicle avoids the difficulty of placing static calibration objects in the traditional method, improves the efficiency of calibration, and can adapt to the situation of sensor external parameter drift.

[0051] In one implementation, in step S11, obtaining sensor data of a target vehicle during its travel includes:

[0052] Obtaining the raw data collected by the sensor during the target vehicle's driving process;

[0053] The three-dimensional coordinates of the preset reference points in each raw data are identified to obtain sensor data.

[0054] Under this implementation method, first, various sensors deployed on the roadside can continuously collect raw data while the target vehicle is driving. These raw data contain the surrounding environment information perceived by the sensors at different times, but are often unprocessed and in different formats, and may include point cloud coordinates, image pixel values, radar reflection intensity, distance and other forms.

[0055] For example, raw data may include, but is not limited to, radar data and image data. Radar data can include lidar data collected by lidar sensors, typically a large amount of unordered point cloud data that records information such as the distance and angle of each reflection point relative to the sensor coordinate system. It can also include millimeter-wave radar data collected by millimeter-wave radar sensors, which may include point cloud data and / or target lists. Image data typically consists of images and / or video frames captured by image sensors.

[0056] This raw data can then be processed to identify the 3D coordinates of preset reference points within each raw data point, thereby generating sensor data. These reference points are typically pre-selected, distinctive, and easily identifiable locations on the target vehicle, such as the axles of the target vehicle's four wheels, specific roof markings, or contact points. Taking the processing of raw data collected by LiDAR as an example, algorithms such as point cloud clustering and feature extraction are used to filter out the points corresponding to the preset reference points from the massive, unordered point cloud, and their precise 3D coordinates in the sensor coordinate system are calculated.

[0057] In this way, the raw data can be converted into sensor data that can accurately reflect the position of the target vehicle in the sensor coordinate system, providing an effective data basis for subsequent time alignment with GPS data and external parameter determination.

[0058] In one implementation, the raw data includes radar data, and identifying the three-dimensional coordinates of a preset reference point in each raw data to obtain sensor data includes:

[0059] Identify the markers installed on the target vehicle from the radar data to obtain point cluster data;

[0060] Based on the point cluster data, the three-dimensional coordinates of a first preset reference point in the radar coordinate system are determined as sensor data. The first preset reference point is the center of mass or the grounding point of the target vehicle.

[0061] In this implementation, when a target vehicle equipped with a marker travels within the sensor's sensing range, roadside radar sensors, such as lidar and / or millimeter-wave radar, continuously emit electromagnetic waves and receive reflected signals throughout the vehicle's journey, collecting radar data. This radar data contains a wealth of environmental information, including signals related to the target vehicle.

[0062] Using point cloud target detection or segmentation algorithms on radar data, it's possible to identify markers mounted on target vehicles within the radar data. Markers are typically objects with high reflectivity or unique shapes, allowing for clear radar capture. When radar waves strike a marker and reflect back, they create specific signal signatures within the radar data. The algorithm then filters out data related to the marker, generating point cluster data that represents the marker's distribution within the radar coordinate system.

[0063] After acquiring the point cluster data, the three-dimensional coordinates of a first preset reference point in the radar coordinate system can be determined based on the point cluster data as sensor data. The first preset reference point can be the center of mass or ground point of the target vehicle and is a key reference point for describing the vehicle's position.

[0064] The target vehicle's center of mass is used as the first preset reference point. Since the point cluster data reflects the location information of the marker, and there is a fixed relative position relationship between the marker and the vehicle's center of mass, by performing cluster analysis and geometric calculations on these point cluster data, combined with the pre-set position parameters of the marker and the center of mass, the precise three-dimensional coordinates of the center of mass in the radar coordinate system can be calculated;

[0065] For example, using LiDAR data as an example, a point cloud object detection or segmentation algorithm is applied to the LiDAR data at each time point t. This algorithm combines a priori information about the shape and reflectivity of the landmark to identify and segment point clusters belonging to the target vehicle (or landmark). Based on this point cluster data, the three-dimensional coordinates of the target vehicle's center of mass or ground contact point in the LiDAR coordinate system, P_Lidar(t) = (x_L(t), y_L(t), z_L(t)), are then determined.

[0066] Taking millimeter-wave radar data as an example, a target detection and tracking algorithm is applied to the millimeter-wave radar data at each time point t. This algorithm leverages the high RCS (Radar Cross Section) characteristics of the marker or known radar signatures to identify the point cluster data of the target vehicle (or marker). Based on this point cluster data, the three-dimensional coordinates of the target vehicle's center of mass or ground contact point in the millimeter-wave radar coordinate system, P_Radar(t) = (x_R(t), y_R(t), z_R(t)), are determined.

[0067] In this way, the obtained sensor data accurately locates the position of the target vehicle in the radar coordinate system, providing data support for subsequent fusion analysis with GPS data and calibration of roadside sensors.

[0068] In one implementation, the marker is a metal cube or a three-sided corner reflector covered with a preset color and pattern.

[0069] The regular edges and corners of the metal cube and the unique angular reflection structure of the three-sided corner reflector provide clear and identifiable features for the lidar data, which helps to use the point cloud segmentation algorithm to quickly and accurately identify the point clusters formed by these markers, thereby achieving precise positioning of the target vehicle.

[0070] Metal materials give markers a naturally high RCS, meaning they can strongly reflect radar waves back to the sensor. Even in complex and changing electromagnetic environments or at long distances, millimeter-wave radar can stably capture and continuously track the position of the target vehicle with its significant echo signal.

[0071] In addition, the overlaid preset colors (such as bright yellow) and patterns (such as eye-catching geometric figures or QR codes) can also provide significant image features for the visual detection of the image sensor. Whether through traditional image processing algorithms or advanced visual detection models, these markers can be quickly identified from the image data and the target vehicle can be locked.

[0072] Therefore, this sign design, which combines radar detectability, lidar recognizability, and visual conspicuity, can effectively ensure the reliable detection and tracking of target vehicles by multiple roadside sensors. It is an optimized choice that comprehensively considers the characteristics of multiple sensors, and greatly improves the robustness and accuracy of target recognition.

[0073] In one implementation, the raw data includes image data, and identifying the three-dimensional coordinates of a preset reference point in each raw data to obtain sensor data includes:

[0074] The three-dimensional coordinates of the ground contact point of the target vehicle in the image coordinate system are identified from the image data as sensor data.

[0075] As the target vehicle drives, the image sensor continuously captures the scene, generating image data containing the target vehicle's information. This image data presents the vehicle's appearance and surroundings in a two-dimensional format. To extract useful sensor data from this data, key preset reference points can be precisely located within the image data. The target vehicle's ground contact point can be selected as a preset reference point. The ground contact point, where the target vehicle makes contact with the ground, is a crucial reference point for describing the vehicle's posture and position.

[0076] Specifically, image recognition algorithms, such as deep learning-based semantic segmentation and target detection technologies, can first identify the target vehicle's outline and the area of ​​contact with the ground in the image data. Then, combining a pre-established vehicle geometry model, the image sensor (including parameters such as focal length and optical center position), and distortion parameters, and utilizing the principle of perspective transformation, the two-dimensional pixel coordinates of the ground contact point in the image data are converted into three-dimensional coordinates in the image sensor coordinate system through complex mathematical calculations to obtain sensor data. This provides image dimension information for subsequent time alignment with GPS data and sensor calibration, helping to achieve multi-source data fusion and precise sensor calibration.

[0077] For example, a monocular 3D target detection model can be applied to process the image data, outputting the pixel coordinates and predicted depth of the target vehicle's ground contact point. Then, combined with the intrinsic parameters of the image sensor, the three-dimensional coordinates of the target vehicle's ground contact point in the image sensor coordinate system are obtained: P_Camera(t) = (x_C(t), y_C(t), z_C(t)).

[0078] In one implementation, the GPS system and the sensor use the same time base, and both the GPS data and the sensor data have timestamps determined based on the time base. In step S12, time alignment processing is performed on the GPS data and the sensor data to determine multiple groups of point pairs, including:

[0079] Determine multiple target time points based on the timestamps of GPS data and sensor data;

[0080] For each target time point, if there is no corresponding GPS data at the target time point, interpolation processing is performed on the GPS data within a preset time range from the target time point to obtain the GPS data corresponding to the target time point;

[0081] For each target time point, if there is no corresponding sensor data at the target time point, interpolation processing is performed on the sensor data within a preset time range from the target time point to obtain the sensor data corresponding to the target time point;

[0082] The GPS data and sensor data corresponding to each target time point are regarded as a set of point pairs.

[0083] Assuming that the GPS system and sensors use the same time base, that is, the data collected by both are timestamped based on this unified base, time alignment can be used to accurately match GPS data with sensor data, forming multiple point pairs for subsequent analysis.

[0084] First, you can filter and determine multiple target time points based on the timestamps carried by GPS and sensor data. The target time point is the time frame in which the data is aligned. For example, you can use every timestamp carried by GPS and sensor data as the target time point, or you can use the timestamp of either GPS or sensor data as the target time point. Alternatively, you can traverse according to a preset time window and determine a target time point in each time window. There are no specific restrictions.

[0085] Since data may be missing during the actual collection process, that is, there may be no corresponding GPS data or sensor data at some target time points, for this situation, for each target time point, it is possible to check whether there is directly corresponding GPS data and sensor data at that moment.

[0086] If no GPS data exists for a target time point, the system can retrieve GPS data within a preset time range from the target time point and apply an appropriate interpolation algorithm, such as linear, spline, or polynomial interpolation, to fill in the data gaps and obtain the GPS data corresponding to the target time point. Similarly, if no sensor data exists for a target time point, a similar interpolation process can be performed on the sensor data within the preset time range to generate the sensor data for that time point.

[0087] After this bidirectional interpolation and filling, each target time point has corresponding GPS data and sensor data. The pair of GPS data and sensor data matched to each target time point is bundled together to form a set of point pairs. For example, the point pair at target time point t_k can be represented as {P_GPS(t_k),P_Lidar(t_k),P_Radar(t_k),P_Camera(t_k)}, where P_GPS(t_k) represents the GPS data in the point pair, P_Lidar(t_k) represents the sensor data collected by the lidar in the point pair, P_Radar(t_k) represents the sensor data collected by the millimeter-wave radar in the point pair, and P_Camera(t_k) represents the sensor data collected by the image sensor in the point pair.

[0088] Each set of point pairs represents the status information of the target vehicle at the same time and in different dimensions. These point pairs provide an accurate and synchronized data basis for the subsequent determination of the first external parameter between the GPS system and the sensor, ensuring that the sensor calibration work based on this data can be more accurate and reliable.

[0089] In some cases, in order to align data more efficiently and accurately, the GPS data and sensor data acquired by each sensor at consecutive timestamps can be connected in chronological order to form trajectories in their respective coordinate systems, including but not limited to:

[0090] The GPS trajectory, denoted as T_GPS = {P_GPS(t_k)}, is used to reflect the movement path of the target vehicle in the macroscopic geographic space;

[0091] The LiDAR trajectory, represented by T_Lidar = {P_Lidar(t_k)}, is used to reflect the target vehicle's motion trajectory within the LiDAR sensing range;

[0092] The millimeter-wave radar trajectory, represented by T_Radar = {P_Radar(t_k)}, is used to reflect the movement information of the target vehicle within the millimeter-wave radar detection space;

[0093] The 3D image trajectory, represented as T_Camera={P_Camera(t_k)}, is used to reflect the spatial motion trajectory of the target vehicle under the viewing angle of the image sensor.

[0094] Then, based on the target time point, the trajectories from different sensors and in different coordinate systems can be time-aligned to better utilize the time series characteristics of the data, smooth the impact of errors or missing individual data points, and achieve more accurate and stable data synchronization by comprehensively analyzing the changing trends of the trajectories in the time dimension.

[0095] In one implementation, in step S13, determining a first external parameter between the GPS system and the sensor based on a mapping relationship between GPS data and sensor data in each point pair includes:

[0096] Build a calibration model, which is used to define the mapping relationship between GPS data and sensor data;

[0097] For each point pair, the corresponding sensor data is input into the calibration model to obtain the corresponding predicted data, and the distance between the predicted data and the corresponding GPS data is calculated;

[0098] With the goal of minimizing the distance between multiple point pairs, an optimization function is constructed and solved to obtain the model parameters of the calibration model, which serve as the first external parameters between the GPS system and the sensor.

[0099] Specifically, a calibration model can be pre-defined to define how sensor data is converted to GPS data. This model is typically based on the principle of rigid body transformation, transforming the sensor data into the GPS coordinate system through rotation and translation. This model describes the relationship between the two types of data in terms of spatial position and attitude. This model includes parameters such as the rotation matrix and translation vector, which are also known as the first extrinsic parameters.

[0100] Then, each set of point pairs can be used to verify the calibration model. The sensor data is input into the calibration model, and the calibration model obtains the corresponding predicted data through calculation. The predicted data is the estimated value of the calibration model based on the current parameter values ​​and the sensor data is converted into the GPS coordinate system.

[0101] For example, for sensor i (which can be a lidar, millimeter-wave radar or image sensor), the point P_i in its own coordinate system can be converted to the predicted data P_W in the global reference coordinate system W through the rotation matrix R_i and the translation vector T_i, and the relationship is: P_W = R_i*P_i+T_i.

[0102] The distance between the predicted data and the GPS data for the same set of point pairs can then be calculated. This distance directly reflects the degree of deviation between the calibration model's conversion results and the actual GPS data under the current parameters. By iterating over all point pairs, multiple sets of distance data reflecting the accuracy of the model's predictions can be obtained. The distance between the predicted data and the GPS data can be measured using the Euclidean distance metric.

[0103] Furthermore, the distance errors of all point pairs can be taken into account, and an optimization function can be constructed with the goal of minimizing the distances of multiple point pairs. For example, the goal can be to minimize the sum of the squares of all distances. This optimization function can then be considered a least-squares problem. During the solution process, methods such as SVD (Singular Value Decomposition) or iterative optimization can be used to continuously adjust the calibration model parameters to minimize the overall error between the predicted data and the GPS data. When the optimization function converges, the resulting calibration model parameters serve as the first external parameters between the GPS system and the sensor.

[0104] For example, we can independently solve the extrinsic parameters M_{W<-i}=(R_i,T_i) for each sensor i to W. Using the time-synchronized point pairs (P_GPS(t_k), P_i(t_k)), we construct a least squares optimization function. The goal is to minimize the overall error between the predicted data and the GPS data for all point pairs by adjusting the parameters of the calibration model (R_i and T_i). The least squares optimization function can be expressed as:

[0105] argmin_{R_i,T_i}Σ_k||P_GPS(t_k)-(R_i*P_i(t_k)+T_i)||^2

[0106] Among them, ||P_GPS(t_k)-(R_i*P_i(t_k)+T_i|| represents the Euclidean distance between the predicted data and the GPS data, Σ_k represents the summation of all synchronized time points t_k, and R_i must satisfy the orthogonality constraint of the rotation matrix to ensure that it indeed represents a valid rotation.

[0107] The first extrinsic parameter accurately describes the spatial relative position and attitude relationship between the GPS system and the sensor. It can be used to solve the absolute orientation problem (Absolute Orientation Problem) or point set registration problem between the GPS system and the sensor, and accurately map the sensor data to the global reference coordinate system where the GPS data is located through rigid body transformation, providing a core basis for the subsequent efficient fusion and accurate application of multi-sensor data.

[0108] In one implementation, the sensor includes a plurality of sensors. In step S13, after determining a first external parameter between the GPS system and the sensor based on a mapping relationship between GPS data and sensor data in each point pair, the method further includes:

[0109] For any two sensors, a chain transformation is performed based on the first extrinsic parameters of the any two sensors to determine the second extrinsic parameter between the any two sensors.

[0110] In an intelligent transportation system with multiple sensors working together, simply determining the first external parameter between each sensor and the GPS system cannot fully meet the needs of all application scenarios. It is also necessary to clarify the relative position relationship between each sensor to directly fuse data from different sensors or compare data between sensors.

[0111] Specifically, each sensor has obtained the transformation parameters (rotation matrix and translation vector) from its own coordinate system to the global reference coordinate system W through calibration with the GPS system. For example, the first extrinsic parameter of sensor i defines the transformation from the coordinate system of sensor i to the global reference coordinate system W, while the first extrinsic parameter of sensor j defines the transformation from the coordinate system of sensor j to the global reference coordinate system W.

[0112] Based on the determined first extrinsic parameters of each sensor and the GPS system, the second extrinsic parameter between any two sensors can be efficiently determined through a chained transformation. First, the first extrinsic parameter of sensor i is used to transform the points in sensor i's coordinate system to the global reference coordinate system. Then, the inverse transformation of the first extrinsic parameter of sensor j is used to transform the points in the global reference coordinate system to the coordinate system of sensor j. Combining these two transformation steps forms a complete transformation relationship from sensor i directly to sensor j, which is the second extrinsic parameter between them.

[0113] For example, for any two sensors i and j, their first extrinsic parameters can be expressed as M_{W<-i}=(R_i,T_i) and M_{W<-j}=(R_j,T_j), respectively. Then the second extrinsic parameter M_{j<-i}=(R_{j<-i},T_{j<-i}) between sensors i and j can be calculated using the chain rule:

[0114] M_{j<-i}=M_{j<-W}*M_{W<-i}=(M_{W<-j})^{-1}*M_{W<-i}

[0115] Among them, M_{j<-W}=(M_{W<-j})^{-1} represents the transformation from the global reference coordinate system W to the coordinate system S_j of sensor j.

[0116] Through this chain transformation method, a direct coordinate transformation relationship is established between any two sensors, which not only avoids the tedious process of individually calibrating each pair of sensors, but also utilizes the accurate calibration relationship established by the GPS system, improves the accuracy and efficiency of determining the second external parameter, and provides a basis for the fusion of multi-sensor data.

[0117] In one implementation, for any two sensors, after chain transformation is performed based on first extrinsic parameters of the any two sensors to determine a second extrinsic parameter between the any two sensors, the method further includes:

[0118] The first external parameter and the second external parameter are transmitted to the corresponding sensor, so that the sensor performs coordinate conversion on the collected data based on the first external parameter and the second external parameter.

[0119] After accurately determining the second extrinsic parameter between any two sensors through chain transformation, in order to achieve efficient fusion and collaborative work of multi-sensor data, it is necessary to transmit the first and second extrinsic parameters to the corresponding sensors so that they can understand the spatial connection between themselves, other sensors and the GPS system.

[0120] Furthermore, when any sensor receives the corresponding first and second external parameters, it can accurately convert the sensor data in its own coordinate system into a coordinate system compatible with other sensors or GPS systems based on the first and second external parameters when collecting data.

[0121] For example, when the lidar and camera work together to detect targets, the lidar can use the second external parameter received between it and the camera to convert the three-dimensional point cloud data it collects into a coordinate system that is compatible with the camera image data, thereby achieving seamless fusion of two different modal data and more accurately identifying and locating targets.

[0122] In this way, not only does each sensor have the ability to express data in a unified manner, but the entire multi-sensor system can also operate in a coordinated manner, giving full play to the advantages of each sensor, providing richer and more accurate data support for the environmental perception and decision-making analysis of the intelligent transportation system, and promoting the efficient operation of applications such as autonomous driving and traffic flow monitoring.

[0123] As can be seen from the above, the technical solution provided by the embodiment of the present disclosure utilizes the target vehicle as a dynamic reference object, and does not require the placement of static calibration objects, thereby avoiding the dependence of traditional calibration methods on static calibration objects. By obtaining the GPS data and sensor data of the target vehicle during its driving process, point pairs are determined through time alignment processing to ensure that the position of the vehicle (GPS data) and the state of the vehicle observed by the sensor (sensor data) at the same time point can be correctly paired, and then, the first external parameter between the GPS system and the sensor is determined based on the point pair mapping relationship. In this way, not only can the efficiency of calibration be improved, but also regular or even dynamic calibration updates can be easily achieved, effectively dealing with the problem of sensor external parameter drift caused by factors such as environmental vibration and temperature changes, meeting the actual needs in complex roadside environments, and can be applied to scenarios in intelligent transportation systems where sensors are far apart and facing different directions.

[0124] Figure 2 The following is a block diagram of a roadside sensor calibration device according to an exemplary embodiment, including:

[0125] An acquisition module 201 is configured to acquire multiple global positioning system (GPS) data and multiple sensor data during the travel of a target vehicle, wherein the GPS data is collected by the GPS system of the target vehicle, and the sensor data is obtained by tracking the target vehicle using roadside sensors;

[0126] a determination module 202 configured to perform time alignment processing on the GPS data and the sensor data to determine a plurality of point pairs, each point pair including the GPS data and the sensor data corresponding to a same time point;

[0127] The calibration module 203 is configured to determine a first external parameter between the GPS system and the sensor based on a mapping relationship between the GPS data and the sensor data in each point pair, wherein the first external parameter is used to calibrate the sensor.

[0128] In one implementation, the acquisition module 201 is specifically configured to:

[0129] Acquiring raw data collected by the sensor during the driving of the target vehicle;

[0130] The three-dimensional coordinates of a preset reference point in each of the raw data are identified to obtain sensor data.

[0131] In one implementation, the raw data includes radar data and / or image data, and the acquisition module 201 is specifically configured to:

[0132] Identifying a marker mounted on the target vehicle from the radar data to obtain point cluster data, and determining, based on the point cluster data, the three-dimensional coordinates of a first preset reference point in the radar coordinate system as sensor data, where the first preset reference point is the center of mass or ground point of the target vehicle; and / or,

[0133] The three-dimensional coordinates of the ground contact point of the target vehicle in the image coordinate system are identified from the image data as sensor data.

[0134] In one implementation, the marker is a metal cube or a three-sided corner reflector covered with a preset color and pattern.

[0135] In one implementation, the GPS system and the sensor use the same time reference, and both the GPS data and the sensor data have timestamps determined based on the time reference. The determining module 202 is specifically configured to:

[0136] determining a plurality of target time points based on the timestamps of the GPS data and the sensor data;

[0137] For each target time point, if there is no corresponding GPS data at the target time point, interpolation processing is performed on the GPS data within a preset time range from the target time point to obtain the GPS data corresponding to the target time point;

[0138] For each target time point, if there is no corresponding sensor data at the target time point, interpolation processing is performed on the sensor data within a preset time range from the target time point to obtain the sensor data corresponding to the target time point;

[0139] The GPS data and the sensor data corresponding to each target time point are taken as a set of point pairs.

[0140] In one implementation, the calibration module 203 is specifically configured to:

[0141] Constructing a calibration model, wherein the calibration model is used to define a mapping relationship between the GPS data and the sensor data;

[0142] For each point pair, the corresponding sensor data is input into the calibration model to obtain the corresponding predicted data, and the distance between the predicted data and the corresponding GPS data is calculated;

[0143] With the goal of minimizing the distances of the multiple point pairs, an optimization function is constructed and solved to obtain model parameters of the calibration model as the first external parameters between the GPS system and the sensor.

[0144] In one implementation, the sensor includes multiple sensors, and the calibration module 203 is further configured to:

[0145] For any two sensors, a chain transformation is performed based on first extrinsic parameters of the any two sensors to determine a second extrinsic parameter between the any two sensors.

[0146] In one implementation, the calibration module 203 is further configured to:

[0147] The first external parameter and the second external parameter are transmitted to the corresponding sensor, so that the sensor performs coordinate transformation on the collected data based on the first external parameter and the second external parameter.

[0148] As can be seen from the above, the technical solution provided by the embodiment of the present disclosure utilizes the target vehicle as a dynamic reference object, and does not require the placement of static calibration objects, thereby avoiding the dependence of traditional calibration methods on static calibration objects. By obtaining the GPS data and sensor data of the target vehicle during its driving process, point pairs are determined through time alignment processing to ensure that the position of the vehicle (GPS data) and the state of the vehicle observed by the sensor (sensor data) at the same time point can be correctly paired, and then, the first external parameter between the GPS system and the sensor is determined based on the point pair mapping relationship. In this way, not only can the efficiency of calibration be improved, but also regular or even dynamic calibration updates can be easily achieved, effectively dealing with the problem of sensor external parameter drift caused by factors such as environmental vibration and temperature changes, meeting the actual needs in complex roadside environments, and can be applied to scenarios in intelligent transportation systems where sensors are far apart and facing different directions.

[0149] Figure 3 The present invention is a block diagram of an electronic device for roadside sensor calibration according to an exemplary embodiment.

[0150] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory including instructions, wherein the instructions are executable by a processor of an electronic device to perform the method. Alternatively, the computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0151] In an exemplary embodiment, a computer program product is also provided. When the computer program product is executed on a computer, the computer implements the roadside sensor calibration method.

[0152] As can be seen from the above, the technical solution provided by the embodiment of the present disclosure utilizes the target vehicle as a dynamic reference object, and does not require the placement of static calibration objects, thereby avoiding the dependence of traditional calibration methods on static calibration objects. By obtaining the GPS data and sensor data of the target vehicle during its driving process, point pairs are determined through time alignment processing to ensure that the position of the vehicle (GPS data) and the state of the vehicle observed by the sensor (sensor data) at the same time point can be correctly paired, and then, the first external parameter between the GPS system and the sensor is determined based on the point pair mapping relationship. In this way, not only can the efficiency of calibration be improved, but also regular or even dynamic calibration updates can be easily achieved, effectively dealing with the problem of sensor external parameter drift caused by factors such as environmental vibration and temperature changes, meeting the actual needs in complex roadside environments, and can be applied to scenarios in intelligent transportation systems where sensors are far apart and facing different directions.

[0153] Figure 4 It is a block diagram of a device 800 for roadside sensor calibration according to an exemplary embodiment.

[0154] For example, apparatus 800 may be a mobile phone, a computer, a digital broadcast electronic device, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, or the like.

[0155] Reference Figure 4 , the apparatus 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0156] The processing component 802 generally controls the overall operation of the device 800, such as operations associated with display, phone calls, data communications, camera operation, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to perform all or part of the steps of the described methods. In addition, the processing component 802 may include one or more modules to facilitate interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate interaction between the multimedia component 808 and the processing component 802.

[0157] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0158] The power supply component 807 provides power to the various components of the device 800. The power supply component 807 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device 800.

[0159] The multimedia component 808 includes a screen that provides an output interface between the device 800 and the account. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the account. The touch panel includes one or more touch sensors to sense touches, slides, and gestures on the touch panel. The touch sensor can not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.

[0160] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), which is configured to receive external audio signals when the device 800 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signal can be further stored in the memory 804 or transmitted via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0161] The I / O interface 812 provides an interface between the processing component 802 and peripheral interface modules, such as a keyboard, a click wheel, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.

[0162] The sensor assembly 814 includes one or more sensors for providing various aspects of the status assessment of the device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the device 800. The sensor assembly 814 can also detect changes in the position of the device 800 or a component of the device 800, the presence or absence of contact with the device 800, the orientation or acceleration / deceleration of the device 800, and changes in the temperature of the device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0163] The communication component 816 is configured to facilitate wired or wireless communication between the device 800 and other devices. The device 800 can access a wireless network based on a communication standard, such as WiFi, an operator network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0164] In an exemplary embodiment, the apparatus 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the methods described in the first and second aspects.

[0165] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the device 800 to perform the method. Alternatively, for example, the storage medium can be a non-transitory computer-readable storage medium, such as a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0166] In an exemplary embodiment, a computer program product including instructions is further provided, which, when executed on a computer, enables the computer to execute the roadside sensor calibration method described in any one of the embodiments.

[0167] As can be seen from the above, the technical solution provided by the embodiment of the present disclosure utilizes the target vehicle as a dynamic reference object, and does not require the placement of static calibration objects, thereby avoiding the dependence of traditional calibration methods on static calibration objects. By obtaining the GPS data and sensor data of the target vehicle during its driving process, point pairs are determined through time alignment processing to ensure that the position of the vehicle (GPS data) and the state of the vehicle observed by the sensor (sensor data) at the same time point can be correctly paired, and then, the first external parameter between the GPS system and the sensor is determined based on the point pair mapping relationship. In this way, not only can the efficiency of calibration be improved, but also regular or even dynamic calibration updates can be easily achieved, effectively dealing with the problem of sensor external parameter drift caused by factors such as environmental vibration and temperature changes, meeting the actual needs in complex roadside environments, and can be applied to scenarios in intelligent transportation systems where sensors are far apart and facing different directions.

[0168] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the following claims.

[0169] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes can be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A roadside sensor calibration method, characterized in that: include: Acquiring multiple global positioning system (GPS) data and multiple sensor data of a target vehicle during its travel, wherein the GPS data is collected by the GPS system of the target vehicle, and the sensor data is obtained by tracking the target vehicle through roadside sensors; Performing time alignment processing on the GPS data and the sensor data to determine a plurality of point pairs, each point pair including the GPS data and the sensor data corresponding to a same time point; Based on the mapping relationship between the GPS data and the sensor data in each point pair, a first external parameter between the GPS system and the sensor is determined, where the first external parameter is used to calibrate the sensor.

2. The roadside sensor calibration method according to claim 1, characterized in that: Acquire sensor data of the target vehicle during its driving process, including: Acquiring raw data collected by the sensor during the driving of the target vehicle; The three-dimensional coordinates of a preset reference point in each of the raw data are identified to obtain sensor data.

3. The roadside sensor calibration method according to claim 2, characterized in that: The raw data includes radar data and / or image data, and identifying the three-dimensional coordinates of a preset reference point in each of the raw data to obtain sensor data includes: Identifying a marker mounted on the target vehicle from the radar data to obtain point cluster data, and determining, based on the point cluster data, the three-dimensional coordinates of a first preset reference point in the radar coordinate system as sensor data, where the first preset reference point is the center of mass or ground point of the target vehicle; and / or, The three-dimensional coordinates of the ground contact point of the target vehicle in the image coordinate system are identified from the image data as sensor data.

4. The roadside sensor calibration method according to claim 3, characterized in that: The marker is a metal cube or a three-sided corner reflector covered with preset colors and patterns.

5. The roadside sensor calibration method according to claim 1, characterized in that: The GPS system and the sensor use the same time reference, the GPS data and the sensor data both have time stamps determined based on the time reference, and the time alignment processing of the GPS data and the sensor data to determine multiple groups of point pairs includes: determining a plurality of target time points based on the timestamps of the GPS data and the sensor data; For each target time point, if there is no corresponding GPS data at the target time point, interpolation processing is performed on the GPS data within a preset time range from the target time point to obtain the GPS data corresponding to the target time point; For each target time point, if there is no corresponding sensor data at the target time point, interpolation processing is performed on the sensor data within a preset time range from the target time point to obtain the sensor data corresponding to the target time point; The GPS data and the sensor data corresponding to each target time point are taken as a set of point pairs.

6. The roadside sensor calibration method according to claim 1, characterized in that: The determining, based on the mapping relationship between the GPS data and the sensor data in each point pair, a first external parameter between the GPS system and the sensor, comprises: Constructing a calibration model, wherein the calibration model is used to define a mapping relationship between the GPS data and the sensor data; For each point pair, the corresponding sensor data is input into the calibration model to obtain the corresponding predicted data, and the distance between the predicted data and the corresponding GPS data is calculated; With the goal of minimizing the distances of the multiple point pairs, an optimization function is constructed and solved to obtain model parameters of the calibration model as the first external parameters between the GPS system and the sensor.

7. The roadside sensor calibration method according to claim 1, characterized in that: The sensors include a plurality of sensors, and after determining a first external parameter between the GPS system and the sensor based on a mapping relationship between the GPS data and the sensor data in each point pair, the method further includes: For any two sensors, a chain transformation is performed based on first extrinsic parameters of the any two sensors to determine a second extrinsic parameter between the any two sensors.

8. The roadside sensor calibration method according to claim 7, characterized in that: After performing chain transformation on any two sensors based on the first extrinsic parameters of the any two sensors to determine the second extrinsic parameter between the any two sensors, the method further includes: The first external parameter and the second external parameter are transmitted to the corresponding sensor, so that the sensor performs coordinate transformation on the collected data based on the first external parameter and the second external parameter.

9. A roadside sensor calibration device, characterized in that: include: An acquisition module is used to acquire multiple global positioning system (GPS) data and multiple sensor data during the driving process of the target vehicle, wherein the GPS data is collected by the GPS system of the target vehicle, and the sensor data is obtained by tracking the target vehicle through roadside sensors; a determination module, configured to perform time alignment processing on the GPS data and the sensor data to determine a plurality of point pairs, each point pair including the GPS data and the sensor data corresponding to a same time point; The calibration module is used to determine a first external parameter between the GPS system and the sensor based on a mapping relationship between the GPS data and the sensor data in each point pair, wherein the first external parameter is used to calibrate the sensor.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the roadside sensor calibration method according to any one of claims 1 to 8.

Images