An object alignment method, device, electronic device and medium for intelligent driving

Abstract

The present invention relates to a target alignment method, device, electronic device and medium for intelligent driving, belonging to the technical field of autonomous driving. The method includes: obtaining a data set collected by a plurality of acquisition devices; using the target object in one of the devices as a reference target object, performing time alignment on the target objects in other devices and the reference target object to obtain a time-aligned data set; determining a flag bit based on the time-aligned data set, and grouping the time-aligned data set according to a preset category condition according to the flag bit to obtain a plurality of grouped data; determining a variable for distinguishing a target object from a non-target object, performing a preset arithmetic solution on the variables in the plurality of grouped data, and determining an index parameter for distinguishing a target object from a non-target object; aligning the target objects in the plurality of acquisition devices according to the index parameter. The present invention solves the technical problems in the prior art that when aligning the acquisition data of multiple sensors or other measurement devices, the calculation amount is large and it is difficult to process multiple targets.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and particularly to a method, device, electronic device and medium for target alignment in intelligent driving. Background Art

[0002] In the current target recognition system, when multiple sensors are set to collect data for the same target to be recognized, there is often a problem of asynchronous data collection. Especially for different types of sensors, due to different scanning mechanisms / modes of each sensor, the time corresponding to the data collected by the sensors is not unified, resulting in different moments when the target to be recognized is collected / scanned by different sensors. In the coordinate systems of different sensors, its corresponding position and size are also inconsistent. As a result, most of the current target recognition is based on the collected data of each sensor to obtain their respective recognition results, and then the respective recognition results are fused to obtain the final target recognition result. In practical applications, there are many cases of false alarms and missed detections.

[0003] The method for target alignment in the industry is as follows: after obtaining the data of sensors or measurement devices, the time synchronization of multiple data sources is performed, and then the Euclidean distance is calculated frame by frame for the data in multiple data sources to obtain the optimal matching target pairs for each frame, so as to complete the task of target alignment.

[0004] However, the above method requires calculating the Euclidean distance frame by frame. If each frame contains multiple target objects or the data volume is large, it will consume a large amount of computing resources, making the task difficult to complete in a short time; and for a single sensor or measurement tool, only one optimal target pair can be found for each frame, and it is difficult to handle the situation where there are multiple target pairs in one frame; for the targets output by a certain sensor or measurement device, if there are two or more targets that are very close to each other, it will interfere with its matching with other sensors or measurement devices, and it is difficult to accurately match to a specific target. Summary of the Invention

[0005] In view of this, it is necessary to provide a method, device, electronic device and medium for target alignment in intelligent driving to solve the technical problems in the prior art that involve large amounts of calculation and difficulty in processing multiple targets when aligning the collected data of multiple sensors or other measurement devices.

[0006] To solve the above problems, the present invention provides a method for target alignment in intelligent driving, including:

[0007] Obtaining a data set collected by multiple acquisition devices;

[0008] Performing time alignment on the data set collected by multiple acquisition devices to obtain a time-aligned data set;

[0009] Determine a flag bit based on the time-aligned dataset, and group the time-aligned dataset according to preset category conditions based on the flag bit to obtain multiple grouped data;

[0010] Determine a variable for distinguishing the target object from the non-target object, perform a preset arithmetic solution on the variables in the multiple grouped data, and determine an index parameter for distinguishing the target object from the non-target object;

[0011] Align the target objects in multiple acquisition devices according to the index parameter.

[0012] In a possible implementation manner, the time alignment of the datasets collected by multiple acquisition devices to obtain a time-aligned dataset includes:

[0013] Obtain the data frequency of each device;

[0014] Compare the differences between the data frequencies of each device, determine the device corresponding to the data frequency with the largest difference as the reference device, and use the target object in the reference device as the reference target object;

[0015] Perform a difference process on the dataset corresponding to the reference device so that the dataset corresponding to the reference device is consistent with the datasets corresponding to the non-reference devices in time to obtain a time-aligned dataset.

[0016] In a possible implementation manner, the determination of the flag bit based on the time-aligned dataset, and the grouping of the time-aligned dataset according to preset category conditions based on the flag bit to obtain multiple grouped data includes:

[0017] Based on the time-aligned data, use the characteristics of the target object as the flag bit, and calibrate the data segments where the same target object is located in each acquisition device according to the flag bit;

[0018] Group the data segments according to preset category conditions to obtain multiple grouped data.

[0019] In a possible implementation manner, the preset category conditions at least include sensor position, target object ID, and target object ID group number; before grouping the data segments according to the preset category conditions, it further includes:

[0020] According to the characteristics of the reference points of the data collected by different acquisition devices, convert the reference points of the target objects collected by multiple acquisition devices into the same points.

[0021] In a possible implementation manner, the determination of the variable for distinguishing the target object from the non-target object, performing a preset arithmetic solution on the variables in the multiple grouped data, and determining the index parameter for distinguishing the target object from the non-target object includes:

[0022] Determine that the variables that have a great influence on distinguishing the target object from the non-target are the longitudinal distance and the lateral distance between the acquisition device and the target object; determine the statistical relationship values of the variables according to the longitudinal distance and the lateral distance, where the statistical relationship values are the mean longitudinal difference, the mean lateral difference, the standard deviation of the longitudinal difference, and the standard deviation of the lateral difference;

[0023] Obtain the longitudinal distance and the lateral distance between the reference device and the target object, and obtain the longitudinal distance and the lateral distance between the non-reference device and the target object;

[0024] Determine the longitudinal differences between the longitudinal distance between the reference device and the target object and the longitudinal distances between each non-reference device and the target object, and determine the mean longitudinal difference and the standard deviation of the longitudinal difference according to each longitudinal difference;

[0025] Determine the lateral differences between the lateral distance between the reference device and the target object and the lateral distances between each non-reference device and the target object, and determine the mean lateral difference and the standard deviation of the lateral difference according to each lateral difference;

[0026] Determine the index parameters for distinguishing the target object from the non-target object according to the correlation relationship among the mean longitudinal difference, the standard deviation of the longitudinal difference, the mean lateral difference, and the standard deviation of the lateral difference.

[0027] In a possible implementation manner, the correlation relationship among the mean longitudinal difference, the standard deviation of the longitudinal difference, the mean lateral difference, and the standard deviation of the lateral difference can be expressed by the following formula:

[0028]

[0029] Wherein, represents the index parameter, represents the weight parameter, represents the statistical relationship value, represents the variable, i represents the variable statistical value number, and N represents the total number of variable statistical values.

[0030] In a possible implementation manner, it further includes:

[0031] Perform target alignment in a loop by adjusting the weight parameter.

[0032] In a second aspect, the present invention further provides a target alignment device for intelligent driving, including:

[0033] An acquisition module, configured to acquire a data set collected by a plurality of acquisition devices;

[0034] A time alignment module, configured to perform time alignment on the target objects in other devices and the reference target object with the target object in one of the devices as the reference target object to obtain a time-aligned data set;

[0035] A grouping module, configured to determine flag bits based on a time-aligned data set, group the time-aligned data set according to preset category conditions based on the flag bits, and obtain multiple grouped data;

[0036] An index parameter determination module, configured to determine variables for distinguishing a target object from a non-target object, perform preset arithmetic operations on the variables in multiple grouped data, and determine index parameters for distinguishing a target object from a non-target object;

[0037] A target alignment module, configured to align target objects in multiple acquisition devices according to the index parameters.

[0038] In a third aspect, the present invention further provides an electronic device, including: a processor and a memory;

[0039] A computer-readable program executable by the processor is stored on the memory;

[0040] When the processor executes the computer-readable program, the steps in the target alignment method for intelligent driving described above are implemented.

[0041] In a fourth aspect, the present invention further provides a computer-readable storage medium, where the computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the target alignment method for intelligent driving described above.

[0042] The beneficial effects of the present invention are as follows: First, a data set collected by multiple acquisition devices is obtained. Subsequently, taking the target object in one of the devices as a reference target object, the target objects in other devices are time-aligned with the reference target object to obtain a time-aligned data set; and flag bits are determined based on the time-aligned data set, and the time-aligned data set is grouped according to preset category conditions based on the flag bits to obtain multiple grouped data; by grouping the data, the consistency of the data is ensured and the computational complexity of frame-by-frame calculation is reduced. Finally, variables for distinguishing a target object from a non-target object are determined, and preset arithmetic operations are performed on the variables in multiple grouped data to perform relevant statistics on the variables, and index parameters for distinguishing a target object from a non-target object are determined; the target object and the non-target object are distinguished by the index parameters, so as to eliminate the non-target object, align the target object, reduce the problems of missed alignment and misalignment, and align the target objects in multiple acquisition devices according to the index parameters, achieving the purpose of efficient and accurate target alignment. Description of the Drawings

[0043] FIG. 1 is a flowchart of a method according to an embodiment of the target alignment method for intelligent driving provided by the present invention;

[0044] Figure 2 FIG. is a flowchart of a method according to an embodiment of step S102 in the target alignment method for intelligent driving provided by the present invention;

[0045] Figure 3 In the target alignment method for intelligent driving provided by the present invention, it is a flowchart of a method for an embodiment of step S104;

[0046] Figure 4 It is a schematic diagram of an embodiment of the target alignment device for intelligent driving provided by the present invention;

[0047] Figure 5 It is a schematic diagram of the operating environment of an embodiment of the electronic device provided by the present invention. Specific Embodiments

[0048] The following will specifically describe the preferred embodiments of the present invention in conjunction with the accompanying drawings. The accompanying drawings form a part of this application and are used together with the embodiments of the present invention to explain the principles of the present invention, and are not used to limit the scope of the present invention.

[0049] A specific embodiment of the present invention discloses a target alignment method for intelligent driving. Please refer to Figure 1 , including:

[0050] S101. Obtain a data set collected by multiple collection devices;

[0051] S102. Perform time alignment on the data set collected by multiple collection devices to obtain a time-aligned data set;

[0052] S103. Determine a flag bit based on the time-aligned data set, and group the time-aligned data set according to a preset category condition according to the flag bit to obtain multiple grouped data;

[0053] S104. Determine a variable for distinguishing a target object from a non-target object, perform a preset arithmetic solution on the variables in the multiple grouped data, and determine an index parameter for distinguishing a target object from a non-target object;

[0054] S105. Align the target objects in the multiple collection devices according to the index parameter.

[0055] In this embodiment, first, a data set collected by multiple acquisition devices is obtained. Subsequently, taking the target object in one of the devices as the reference target object, the target objects in other devices are time-aligned with the reference target object to obtain a time-aligned data set. And a flag bit is determined based on the time-aligned data set, and the time-aligned data set is grouped according to a preset category condition according to the flag bit to obtain multiple grouped data. By grouping the data, the consistency of the data is ensured and the computational complexity of frame-by-frame calculation is reduced. Finally, a variable for distinguishing the target object from the non-target object is determined, and a preset arithmetic solution is performed on the variables in the multiple grouped data to determine an index parameter for distinguishing the target object from the non-target object. By using the index parameter to distinguish the target object from the non-target object, the problems of missed alignment and misalignment are reduced, and according to the index parameter, the target objects in multiple acquisition devices are aligned, achieving the purpose of efficiently and accurately aligning the target.

[0056] It should be noted that the acquisition devices include various sensors or measuring devices, specifically including but not limited to surround cameras and ground truth device RT-Range. In a specific embodiment, if it is necessary to evaluate the characteristics such as the distance or speed of 5 surround cameras (front fisheye, front side left and right fisheye, rear side left and right long focal), the ground truth device RT-Range (or a device similar to the method for obtaining the true distance and speed of the target) is used to compare the output data of the surround cameras to check the difference between it and the ground truth.

[0057] Furthermore, the purpose of this solution is to find the target object in each acquisition device. In this embodiment, according to the test target, the target alignment is mainly performed by collecting the data of surround cameras, R devices, and vehicle body data. The acquisition devices can be sensors such as cameras, millimeter wave radars, and lidar, and the data set can be related data such as fusion, planning, control, decision-making, or vehicle body.

[0058] Even further, after collecting the data, it is also necessary to parse the collected data into readable text or pictures while ensuring the uniqueness, integrity, and correctness of the data. Among them:

[0059] Uniqueness: The parsed text data all contain timestamp information, and the timestamp of each frame of data is unique;

[0060] Integrity: The data parsing should include the necessary information of the evaluation index, such as distance, speed, and acceleration, etc.;

[0061] Correctness: After parsing the data, it should be sampled to check whether the values in it are consistent with the original data.

[0062] It should be noted that in this embodiment, the target object is a vehicle, and the number of target objects is one vehicle. The non-target objects are other objects that appear in the data collected by the acquisition device except the target vehicle. It can be understood that the target object is not limited to a vehicle, and the number of target objects is not limited either.

[0063] In step S103, the flag bit represents a feature that can uniquely and accurately identify the target object. The flag bit can be determined according to the target object or the actual application scenario, and is not limited here.

[0064] In some embodiments, for time-aligning the data sets collected by multiple acquisition devices to obtain a time-aligned data set, please refer to Figure 2 , including:

[0065] S201. Obtain the data frequency of each device;

[0066] S202. Compare the differences between the data frequencies of each device with each other, and determine the device corresponding to the data frequency with the largest difference as the reference device, and use the target object in the reference device as the reference target object;

[0067] S203. Perform difference processing on the data set corresponding to the reference device so that the data set corresponding to the reference device is consistent with the data sets corresponding to the non-reference devices in terms of time, and obtain a time-aligned data set.

[0068] It should be noted that time alignment means aligning the output data of multiple sensors or measurement devices according to time information. This step is based on the time synchronization of multiple sensors or measurement devices, that is, the time delay of the devices should be eliminated and multiple devices should be synchronized to the same time source. Among the 5 surround-view cameras involved in this embodiment, there is a time difference between the time of photo exposure and the data collected by the device. This period of time is called time delay. The method to eliminate the time delay is to stamp the time point of photo exposure with a time stamp. The device that stamps the time stamp is the time synchronization device. At the same time, use the same time synchronization device to stamp the detection time of device R with a time stamp. The time stamps used for time alignment are those stamped for both.

[0069] In step S201, obtaining the data frequency of each device means determining the frequency or time interval at which each device generates data. This is crucial for data processing and analysis because different devices may generate data at different frequencies, and these frequencies may change over time. Understanding the data frequency of each device helps ensure that the data generated by different devices can be effectively compared and analyzed within the same time period, thus avoiding data inconsistencies caused by different frequencies.

[0070] In this embodiment, to ensure that each device is aligned in time, it is determined whether each device records data at the same time through the data frequency, and the data sets collected by devices with relatively large frequency differences are processed by interpolation to fill the time difference between the two and ensure that they record data at the same time point.

[0071] It should be noted that time alignment can be achieved after interpolation processing because interpolation can estimate the values between two known data points based on the known data points. In this embodiment, the data frequency of device R is quite different from that of other surround-view devices. Performing interpolation processing on device R can align its data with the data of the surround-view camera in time and ensure that they record data at the same time point. In this way, even if their data frequencies are different, their records can be aligned in time, making subsequent analysis and processing more accurate and reliable.

[0072] It can be understood that if the data frequency of the surround-view device is similar to that of other sensors, no interpolation processing is required for it. However, if the data frequency of device R is significantly different from that of other devices, interpolation processing needs to be performed to ensure that the data is aligned in time. On the other hand, it is also possible to determine the data that needs to be processed by interpolation from the perspective of the stability of device data; if the device data is subject to more noise or fluctuations, additional processing is required to ensure time alignment and data accuracy.

[0073] Furthermore, the non-reference device is a concept relative to the reference device. The non-reference device refers to all other devices except the reference device.

[0074] In a specific embodiment, the surround-view data frequency is 33HZ, and the data frequency of device R is 100HZ. The frequency difference of device R is relatively large. Based on the surround-view data, with device R as the reference device, interpolation is performed on the data of device R for time alignment. Since there are 5 sensor data, and there may be multiple target objects in each sensor data, the target objects detected by device R may appear in them, but it is not known in which sensor the target object appears, or which of the target objects it is. Therefore, when aligning the targets, the target objects of all sensors will be aligned with the detected target objects of device R in time.

[0075] In some embodiments, the flag bit is determined based on the time-aligned data set, and the time-aligned data set is grouped according to preset category conditions according to the flag bit to obtain multiple grouped data, including:

[0076] Based on the time-aligned data, using the characteristics of the target object as the flag bit, the data segments where the same target object is located in each acquisition device are calibrated according to the flag bit;

[0077] Group the data segments according to the preset category conditions to obtain multiple grouped data.

[0078] In this embodiment, by determining the flag bits of the target object in all the acquisition devices, accurate positioning and qualitative understanding of the target object can be achieved. Through the flag bits, it can be determined that the target objects output by a certain acquisition device within a certain period of time are the same target object.

[0079] Furthermore, the flag bit data can be utilized to improve the efficiency and accuracy of the system from the following aspects:

[0080] Collect and analyze the flag bit data output by each acquisition device to comprehensively evaluate the performance of the target object in different devices.

[0081] Associate the target object data output by different devices through timestamps or other methods to confirm whether they represent the same target object.

[0082] Utilize the flag bit data to detect and correct possible errors to improve the accuracy of target object positioning and understanding.

[0083] According to the analysis results of the flag bit data, feedback and adjustment are made to the working mode of the acquisition device to optimize the overall performance of the system.

[0084] Establish a real-time monitoring mechanism to detect and warn of possible abnormal situations or device failures based on the flag bit data to ensure the stable operation of the system.

[0085] In a specific embodiment, the flag bits can be the target object ID, attribute information, device ID, timestamp, etc.

[0086] In some embodiments, the preset category conditions at least include the sensor position, target object ID, and target object ID group number; before grouping the data segments according to the preset category conditions, it further includes:

[0087] Convert the reference points of the target objects collected by multiple acquisition devices into the same points according to the characteristics of the reference points for data collection by different acquisition devices.

[0088] In this embodiment, by converting the reference points, it is ensured that multiple acquisition devices have the same reference framework or coordinate system. By unifying the target reference points, it can be ensured that the data deviations between the panoramic view data and the R device data of multiple acquisition devices are within a controllable range, which is crucial for subsequent target alignment.

[0089] In a specific embodiment, since the target detected by the surround view at the rear side is the nearest point, while the target detected by the R device is a fixed point (the front or rear of the vehicle), if the R device uses the front of the vehicle as the reference point, when the target vehicle drives away backward, the surround view detection value and the R device detection value will differ by the length of the target vehicle. Similarly, there are similar problems for the side front or forward. There are many ways to solve the problem of inconsistent target reference points. For example, the absolute speed of the target can be calculated to determine the orientation of the front of the vehicle, or data can be collected and stored by file according to the same vehicle driving orientation, etc.

[0090] Specifically, calculate the absolute speed of the target: By calculating the absolute speed of the target, the driving direction of the target vehicle can be determined, thereby unifying the reference point. For example, when the target vehicle drives away backward, the surround view detection value and the R device detection value can be adjusted to ensure that they are based on the same reference point.

[0091] Unify file collection and storage: Data can be collected and stored by file according to the same vehicle driving orientation. This can ensure that in the subsequent processing process, data from different devices has the same reference point and consistent data structure, thereby simplifying the process of data alignment and analysis.

[0092] Through these methods, the data processing problems caused by inconsistent target reference points can be effectively solved, ensuring accurate and reliable results in subsequent analysis and applications.

[0093] In this embodiment, the reference point is unified by judging the absolute speed of the target.

[0094] Furthermore, the preset category conditions at least include sensor position, target ID, and target ID group number. This means that when grouping data, the data segments will be divided into different groups according to these conditions. The sensor position can be used to determine the position of the target in space, the target ID is used to uniquely identify each target, and the target ID group number is used to group related targets together for more in-depth analysis and processing. The generation of the ID group number is because in the long-term acquisition process, the same ID may represent different targets. In order to distinguish them, they need to be numbered, which is called the ID group number.

[0095] Among them, the preset category conditions may also include motion state, target type, environmental conditions, target attributes, and data quality indicators.

[0096] In a specific embodiment, grouping is performed according to conditions such as the sensor being the left rear, the ID being 125, and the ID group number being 3, to obtain multiple grouped data.

[0097] In some embodiments, to determine the variables for differentiating the target object from the non-target object, perform a preset arithmetic solution on the variables in multiple grouped data, and determine the index parameters for differentiating the target object from the non-target object. Please refer to Figure 3 , including:

[0098] S301. Determine that the variables having a great influence on differentiating the target object from the non-target object are the longitudinal distance and the lateral distance between the acquisition device and the target object; determine the statistical relationship values of the variables according to the longitudinal distance and the lateral distance, where the statistical relationship values are the mean longitudinal difference, the mean lateral difference, the standard deviation of the longitudinal difference, and the standard deviation of the lateral difference;

[0099] S302. Obtain the longitudinal distance and the lateral distance between the reference device and the target object, and obtain the longitudinal distance and the lateral distance between the non-reference device and the target object;

[0100] S303. Determine the longitudinal differences between the longitudinal distance between the reference device and the target object and the longitudinal distances between each non-reference device and the target object, and determine the mean longitudinal difference and the standard deviation of the longitudinal difference according to each longitudinal difference;

[0101] S304. Determine the lateral differences between the lateral distance between the reference device and the target object and the lateral distances between each non-reference device and the target object, and determine the mean lateral difference and the standard deviation of the lateral difference according to each lateral difference;

[0102] S305. Determine the index parameters for differentiating the target object from the non-target object according to the correlation relationship among the mean longitudinal difference, the standard deviation of the longitudinal difference, the mean lateral difference, and the standard deviation of the lateral difference.

[0103] In this embodiment, first, obtain the data sets collected by multiple acquisition devices. Subsequently, take the target object in one of the devices as the reference target object, perform time alignment on the target objects in other devices and the reference target object to obtain the time-aligned data set; and determine the flag bits based on the time-aligned data set, and group the time-aligned data set according to the preset category conditions to obtain multiple grouped data; by grouping the data, the consistency of the data is ensured and the calculation amount of frame-by-frame calculation is reduced. Finally, determine the variables for differentiating the target object from the non-target object, perform a preset arithmetic solution on the variables in multiple grouped data, and determine the index parameters for differentiating the target object from the non-target object; differentiate the target object from the non-target object through the index parameters, reduce the problems of missed alignment and misalignment, and align the target objects in multiple acquisition devices according to the index parameters, achieving the purpose of efficient and accurate alignment of the target.

[0104] Further, first determine the variables that have a great impact on distinguishing the target object from non-targets as the longitudinal distance and the lateral distance between the acquisition device and the target object; determine the statistical relationship values of the variables according to the longitudinal distance and the lateral distance, where the statistical relationship values are the mean longitudinal difference, the mean lateral difference, the standard deviation of the longitudinal difference, and the standard deviation of the lateral difference; then obtain the longitudinal distance and the lateral distance between the reference device and the target object, and obtain the longitudinal distance and the lateral distance between the non-reference device and the target object; then determine the longitudinal differences between the longitudinal distance between the reference device and the target object and the longitudinal distances between each non-reference device and the target object, and determine the mean longitudinal difference and the standard deviation of the longitudinal difference according to each longitudinal difference; and determine the lateral differences between the lateral distance between the reference device and the target object and the lateral distances between each non-reference device and the target object, and determine the mean lateral difference and the standard deviation of the lateral difference according to each lateral difference; finally, determine the index parameters for distinguishing the target object from non-targets according to the correlation relationship between the mean longitudinal difference, the standard deviation of the longitudinal difference, the mean lateral difference, and the standard deviation of the lateral difference.

[0105] It should be noted that the variables can also be parameters such as speed, shape, area, and reflection characteristics. By statistically analyzing the variables, the index parameters for distinguishing the target object from non-targets are obtained, and the target object is screened according to the index parameters and the target objects in multiple acquisition devices are aligned.

[0106] In a specific embodiment, the mean value μ and the standard deviation σ of the longitudinal distance difference dx and the lateral distance difference dy between the panoramic view and the R device are solved, and then the solved values are weighted (the weight is represented by ω) and added to obtain an index for distinguishing the target from non-targets, and the obtained index parameter is the discrimination number (represented by Δ), and the formula for this index is as follows:

[0107]

[0108] The calculation formulas for the mean value and the standard deviation are as follows:

[0109]

[0110] The above formulas are for this example. For this example, the longitudinal and lateral distance differences and their mean values and standard deviations have a relatively large impact on distinguishing and matching the target, so only these values are included in the formula. Of course, it is not limited to these values, and other variables that have an obvious impact on the discrimination number and their statistical values can also be included. If the variable that has an obvious impact on the discrimination number is set as α (such as the longitudinal and lateral distances or their differences), and the statistical relationship value of the variable is ρ (such as the mean value and the standard deviation), then the more general calculation formula for the discrimination number Δ is:

[0111]

[0112] Among them, represents an index parameter, represents a weight parameter, represents a statistical relationship value, represents a variable, i represents the variable statistical value number, and N represents the total number of variable statistical values.

[0113] It can be understood that the setting and magnitude of the weights directly determine the ability and accuracy of the index parameter to distinguish the target from the non-target. A reasonable weight allocation can improve the ability and accuracy of the index parameter to distinguish the target from the non-target. Usually, the weights can be determined through experiments or based on domain knowledge, or machine learning algorithms can be used to automatically learn and optimize the weights to achieve the best discrimination effect.

[0114] In a specific embodiment, it is confirmed that α is the difference in the horizontal and vertical distances, ρ is the average value and the standard deviation, and the weights are confirmed by presetting an initial value group with a sum of 1. The number of elements in the group is determined according to the number of weights. For example, if there are 4 weights involved in the example, the initial value group can be set to (0.25, 0.25, 0.25, 0.25), and the order is arranged according to the subscript numbers of the weights. Then, for the discrimination numbers of each group obtained by grouping the sensor categories, IDs, and ID group numbers, by analyzing the data, the discrimination number boundary Δt for distinguishing the target from the non-target is obtained. The determined targets need to be retained, and other targets do not need to be retained. The corresponding formula is as follows (where keep represents whether to retain the grouped data, 1 means retain, and 0 means do not retain):

[0115]

[0116] Furthermore, in order to ensure the rationality of the weight setting, another set of data sets is selected for verification. Specifically, through actual data or simulation experiments, the performance under different weight settings is evaluated and compared, and the weight combination with the best effect is selected. Or methods such as cross-validation are used to verify the generalization ability of the model on different data sets to ensure that the weight setting has reasonable applicability in different situations. Or a feedback mechanism is established, and the weight setting is continuously adjusted and optimized according to the feedback information in actual applications to adapt to environmental changes and requirement changes.

[0117] After verification and screening, the final weights are obtained, thereby determining the index parameter. According to the index parameter, the target and non-target are distinguished, and the target is screened to obtain a new target data set.

[0118] After generating the target dataset through the above steps, use a visualization tool or other automated tools to verify whether the targets are aligned, whether there are missing targets or misaligned targets. If necessary, readjust ω and re-enter the target alignment step. Repeat this step multiple times to obtain the weight ω and the discrimination value Δt that conform to the current data.

[0119] Based on the above target alignment method for intelligent driving, an embodiment of the present invention further provides a target alignment device for intelligent driving. Please refer to Figure 4 , including: an acquisition module 410, a time alignment module 420, a grouping module 430, an index parameter determination module 440, and a target alignment module 450.

[0120] The acquisition module 410 is configured to acquire a dataset collected by multiple acquisition devices;

[0121] The time alignment module 420 is configured to use the target in one of the devices as a reference target, and perform time alignment on the targets in other devices with the reference target to obtain a time-aligned dataset;

[0122] The grouping module 430 is configured to determine a flag bit based on the time-aligned dataset, and group the time-aligned dataset according to a preset category condition according to the flag bit to obtain multiple grouped data;

[0123] The index parameter determination module 440 is configured to determine a variable for distinguishing a target from a non-target, perform a preset arithmetic solution on the variables in the multiple grouped data, and determine an index parameter for distinguishing a target from a non-target;

[0124] The target alignment module 450 is configured to align the targets in multiple acquisition devices according to the index parameter.

[0125] As Figure 5 shown, based on the above target alignment method for intelligent driving, the present invention also correspondingly provides an electronic device, which may be a computing electronic device such as a mobile terminal, a desktop computer, a notebook, a palm computer, and a server. The electronic device includes a processor 510, a memory 520, and a display 530. Figure 5 Only some components of the electronic device are shown, but it should be understood that it is not required to implement all the shown components, and more or fewer components can be alternatively implemented.

[0126] The memory 520 may be an internal storage unit of the electronic device in some embodiments, such as the hard disk or memory of the electronic device. The memory 520 may also be an external storage electronic device of the electronic device in other embodiments, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the electronic device. Further, the memory 520 may also include both the internal storage unit of the electronic device and the external storage electronic device. The memory 520 is used to store application software installed on the electronic device and various types of data, such as program codes installed on the electronic device. The memory 520 may also be used to temporarily store data that has been output or will be output. In one embodiment, a target alignment program 540 for intelligent driving is stored on the memory 520, and the target alignment program 540 for intelligent driving can be executed by the processor 510, so as to implement the target alignment method for intelligent driving in various embodiments of the present application.

[0127] The processor 510 may be a central processing unit (CPU), a microprocessor, or other data processing chips in some embodiments, and is used to run the program codes stored in the memory 520 or process data, such as executing the target alignment method for intelligent driving, etc.

[0128] The display 530 may be an LED display, a liquid crystal display, a touch liquid crystal display, and an OLED (Organic Light-Emitting Diode) toucher, etc. in some embodiments. The display 530 is used to display information on the target alignment electronic device for intelligent driving and to display a visual user interface. The components 510 - 530 of the electronic device communicate with each other through a system bus.

[0129] Those skilled in the art can understand that all or part of the processes for implementing the methods of the above embodiments can be completed by instructing relevant hardware through a computer program, and the program can be stored in a computer-readable storage medium. Among them, the computer-readable storage medium is a magnetic disk, an optical disk, a read-only memory, or a random access memory, etc.

[0130] As mentioned above, the above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention.

Claims

1. A target alignment method for intelligent driving, characterized in that, Including: Obtain data sets collected by multiple acquisition devices; Perform time alignment on the data sets collected by multiple acquisition devices to obtain a time-aligned data set; Determine flag bits based on the time-aligned data set, and group the time-aligned data set according to preset category conditions to obtain multiple grouped data; Determine variables for distinguishing the target object from non-target objects, perform preset arithmetic operations on the variables in the multiple grouped data, and determine index parameters for distinguishing the target object from non-target objects; Align the target objects in multiple acquisition devices according to the index parameters.

2. The target alignment method for intelligent driving according to claim 1, wherein, The performing time alignment on the data sets collected by multiple acquisition devices to obtain a time-aligned data set includes: Obtain the data frequency of each device; Compare the differences between the data frequencies of each device, and determine the device corresponding to the data frequency with the largest difference as the reference device, and use the target object in the reference device as the reference target object; Perform difference processing on the data set corresponding to the reference device to make the data set corresponding to the reference device consistent with the data sets corresponding to non-reference devices in time, and obtain a time-aligned data set.

3. The target alignment method for intelligent driving according to claim 1, wherein The determining flag bits based on the time-aligned data set and grouping the time-aligned data set according to preset category conditions to obtain multiple grouped data includes: Based on the time-aligned data, use the characteristics of the target object as flag bits, and calibrate the data segments where the same target object in each acquisition device is located according to the flag bits; Group the data segments according to preset category conditions to obtain multiple grouped data.

4. The target alignment method for intelligent driving according to claim 3, characterized in that The preset category conditions at least include sensor position, target object ID, and target object ID group number; Before grouping the data segments according to preset category conditions, it further includes: Convert the reference points of the target objects collected by multiple acquisition devices into the same point according to the characteristics of the reference points of the data collected by different acquisition devices.

5. The target alignment method for intelligent driving according to claim 2, wherein, The determining variables for distinguishing the target object from non-target objects, performing preset arithmetic operations on the variables in the multiple grouped data, and determining index parameters for distinguishing the target object from non-target objects includes: Determine that the variables that have a great impact on distinguishing the target object from non-target objects are the longitudinal distance and the lateral distance between the acquisition device and the target object; determine the statistical relationship values of the variables according to the longitudinal distance and the lateral distance, where the statistical relationship values are the mean longitudinal difference, the mean lateral difference, the standard deviation of the longitudinal difference, and the standard deviation of the lateral difference; Obtain the longitudinal distance and the lateral distance between the reference device and the target object, and obtain the longitudinal distance and the lateral distance between the non-reference device and the target object; Determine the longitudinal differences between the longitudinal distance between the reference device and the target object and the longitudinal distances between each non-reference device and the target object, and determine the mean longitudinal difference and the standard deviation of the longitudinal difference according to the longitudinal differences; Determine the lateral differences between the lateral distance between the reference device and the target object and the lateral distances between each non-reference device and the target object, and determine the mean lateral difference and the standard deviation of the lateral difference according to the lateral differences; Determine the index parameters for distinguishing the target object from non-target objects according to the correlation relationship between the mean longitudinal difference, the standard deviation of the longitudinal difference, the mean lateral difference, and the standard deviation of the lateral difference.

6. The target alignment method for intelligent driving according to claim 5, characterized in that The correlation relationship among the longitudinal difference mean, the longitudinal difference standard deviation, the lateral difference mean, and the lateral difference standard deviation is expressed by the following formula: Among them, represents the index parameter, represents the weight parameter, represents the statistical relationship value, represents the variable, i represents the variable statistical value number, and N represents the total number of variable statistical values.

7. The target alignment method for intelligent driving according to claim 6, wherein, It further includes: By adjusting the weight parameter, target alignment is performed cyclically.

8. An object alignment device for intelligent driving, characterized in that, It includes: An acquisition module, configured to acquire a data set collected by multiple acquisition devices; A time alignment module, configured to use the target object in one of the devices as a reference target object, perform time alignment on the target objects in other devices and the reference target object, and obtain a time-aligned data set; A grouping module, configured to determine a flag bit based on the time-aligned data set, and group the time-aligned data set according to a preset category condition according to the flag bit to obtain a plurality of grouped data; An index parameter determination module, configured to determine a variable for distinguishing a target object from a non-target object, perform a preset arithmetic solution on the variables in the plurality of grouped data, and determine an index parameter for distinguishing a target object from a non-target object; A target alignment module, configured to align the target objects in multiple acquisition devices according to the index parameter.

9. An electronic device, characterized in that, It includes: A processor and a memory; A computer-readable program executable by the processor is stored on the memory; When the processor executes the computer-readable program, the steps in the target alignment method for intelligent driving according to any one of claims 1-7 are implemented.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps in the target alignment method for intelligent driving according to any one of claims 1-7.

Images