Method, device, electronic device and computer storage medium for determining target state

Abstract

The embodiment of the present application provides a method, device, electronic device and computer storage medium for determining the target state. The method for determining the target state includes: obtaining point cloud data of the target to be determined at the current moment; using a preset point cloud registration algorithm to perform registration on the point cloud data at the current moment, which can reduce the error during point cloud registration and improve the accuracy of the registration processing results. Based on the registration process, the speed distribution of the target to be determined at the current moment can be accurately obtained; based on the speed distribution of the target to be determined at the current moment, the point cloud position change of the target to be determined at the next moment is predicted. This allows the autonomous driving vehicle to more accurately perceive the state of its surrounding objects, i.e., the target to be determined, so that the autonomous driving vehicle system can accurately perceive its surrounding objects, thereby executing more accurate and reasonable driving decisions, and providing protection for the safe driving of the autonomous driving vehicle.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of autonomous driving technology, and in particular to a method, device, electronic device, and computer storage medium for determining a target state. Background Art

[0002] Autonomous driving is a technology that uses computer systems to control vehicles to achieve unmanned driving. It has great advantages in improving driving safety and traffic efficiency, and has become a research hotspot in the industry.

[0003] For high-level autonomous vehicle systems, which operate in complex environments and rely largely on independent driver input, accurate perception of the vehicle's surroundings is crucial. Accurate motion perception of surrounding objects effectively ensures the safety of autonomous vehicle systems and lays the foundation for subsequent decision-making and control. Currently, autonomous vehicle systems often use three-dimensional lidar (LiDAR) as a key sensor to obtain reliable three-dimensional information about surrounding objects. However, 3D LiDAR sampling data is sparse, making it inadequate for speed measurement.

[0004] Therefore, how to enable the autonomous driving vehicle system to accurately perceive the motion of its surrounding objects has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, an embodiment of the present application provides a solution for determining a target state to at least partially solve the above-mentioned problem.

[0006] According to a first aspect of an embodiment of the present application, a method for determining a target state is provided, comprising: obtaining point cloud data of a target to be determined at a current moment; determining a velocity distribution of the target to be determined using a preset point cloud registration algorithm for the point cloud data at the current moment; and predicting a point cloud position change of the target to be determined at the next moment based on the velocity distribution of the target to be determined.

[0007] According to a second aspect of an embodiment of the present application, a device for determining a target state is provided, comprising an acquisition module, a registration module and a prediction module; the acquisition module is used to obtain point cloud data of a target to be determined at a current moment; the registration module is used to determine a velocity distribution of the target to be determined using a preset point cloud registration algorithm for the point cloud data at the current moment; the prediction module is used to predict a change in the point cloud position of the target to be determined at the next moment based on the velocity distribution of the target to be determined.

[0008] According to a third aspect of an embodiment of the present application, an electronic device is provided, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; the memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform an operation corresponding to the target state determination method as described in the first aspect.

[0009] According to a fourth aspect of an embodiment of the present application, a computer storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the method for determining the target state as described in the first aspect is implemented.

[0010] According to the target state determination scheme provided in the embodiment of the present application, the point cloud data of the target to be determined at the current moment is obtained; the point cloud data at the current moment is registered using a preset point cloud registration algorithm, which can reduce the error during point cloud registration and improve the accuracy of the registration processing results. Through this registration process, the speed distribution of the target to be determined can be determined; based on the speed distribution of the target to be determined, the point cloud position change of the target to be determined at the next moment is predicted. Because the registration of the point cloud position is an important aspect in the registration process, and because the time interval for collecting point cloud data is short, based on this, the point cloud position change of the target to be determined at the next moment can be predicted based on the speed distribution at the current moment. As a result, the autonomous driving vehicle can more accurately perceive the state of the surrounding objects, i.e., the target to be determined, thereby executing more accurate and reasonable driving decisions, and providing guarantees for the safe driving of the autonomous driving vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0012] Figure 1 A flowchart of a method for determining a target state provided in an embodiment of the present application;

[0013] Figure 2 A schematic diagram illustrating an example application scenario provided by an embodiment of the present application;

[0014] Figure 3 A flowchart of another method for determining a target state provided in an embodiment of the present application;

[0015] Figure 4 A structural block diagram of a target state determination device provided in an embodiment of the present application;

[0016] Figure 5 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to enable those skilled in the art to better understand the technical solutions in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field should fall within the scope of protection of the embodiments of the present application.

[0018] It should be noted that the first and second in this application are only for distinguishing names and do not represent an order relationship. They cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated, for example, the first state, the second state, the first category, the second category.

[0019] The specific implementation of the embodiment of the present application is further explained below in conjunction with the accompanying drawings of the embodiment of the present application.

[0020] Example 1

[0021] The first embodiment of the present application provides a method for determining a target state, such as Figure 1 As shown, Figure 1 A flowchart of a method for determining a target state provided in an embodiment of the present application, the method for determining a target state includes the following steps:

[0022] Step S101: Obtain point cloud data of the target to be determined at the current moment.

[0023] In the embodiments of the present application, the target to be determined can be any object in relative motion. Taking the autonomous driving scenario as an example, the target to be determined can be any object around the autonomous driving vehicle. Any object that moves relative to the autonomous driving vehicle includes but is not limited to people, other vehicles, road facilities, obstacles, and pole-shaped facilities (such as telephone poles or traffic light poles). Point cloud data is a point-based record of the three-dimensional coordinates or color information (RGB) or reflection intensity information (Intensity), etc., representing each point of an object. In this example, point cloud data of the target to be determined around the vehicle can be collected in real time. The point cloud data at the current moment can be point cloud data at any moment in the real-time collected point cloud data. Taking the autonomous driving scenario as an example, the autonomous driving vehicle is equipped with at least one sensor, which can be a lidar, millimeter-wave radar, camera, etc. The example of using the vehicle-mounted three-dimensional lidar to collect point cloud data of obstacles around the vehicle in real time is used for explanation. The three-dimensional lidar can be used to collect point cloud data of obstacles around the vehicle at different times in real time.

[0024] It should be noted that the target state determination method provided in the embodiment of the present application is applied in the autonomous driving scenario, such as Figure 2 As shown, Figure 2 This is a schematic diagram illustrating an example of an application scenario provided in an embodiment of the present application. Figure 2 The driving system includes an autonomous driving vehicle 21 and a network 22. The autonomous driving vehicle 21 is provided with a driving control device 211 and a three-dimensional laser radar 212. The driving control device 211 is used to control the driving path, speed, direction and other information of the autonomous driving vehicle 21. The three-dimensional laser radar 212 is a sensor of the autonomous driving system and can provide three-dimensional information of obstacles around the vehicle. The driving control device 211 communicates interactively with the three-dimensional laser radar 212 through the network 22. The network 22 can be a network 22 of various connection types, such as wired, wireless communication links or optical fiber cables. The target state determination method provided in the embodiment of the present application is executed by the driving control device 211. Accordingly, the target state determination device is provided in the driving control device 211. It can be understood that Figure 2 The number of the driving control device 211, the network 22, and the three-dimensional laser radar 212 is only a schematic representation. The embodiment of the present application does not limit the number of the driving control device 211, the network 22, and the three-dimensional laser radar. Figure 2 The three-dimensional laser radar is used as an example for explanation. Of course, other sensors for collecting point cloud data of obstacles around the vehicle can also be used, and this embodiment of the application does not limit this.

[0025] Step S102: Using a preset point cloud registration algorithm for the point cloud data at the current moment, determine the velocity distribution of the target to be determined.

[0026] The preset point cloud registration algorithm in the embodiment of the present application can be implemented by an appropriate algorithm, such as: simulated annealing algorithm, grid search algorithm, genetic algorithm, taboo search algorithm, particle swarm algorithm, ant colony algorithm, etc., and this embodiment of the present application does not limit this. Because the scanning beam emitted by the three-dimensional laser radar will be blocked by obstacles or other objects, it is impossible to complete the acquisition of the three-dimensional point cloud data of the entire obstacle through a single scan, so it is necessary to scan the object from different positions and angles. Point cloud registration refers to the point cloud data obtained at different times for the same obstacle, and the point cloud data scanned at adjacent moments or preset intervals are spliced together. In the embodiment of the present application, the velocity distribution can be characterized by a velocity probability density distribution function (Probability Density Function, referred to as PDF).

[0027] For the point cloud data at the current moment, using the preset point cloud registration algorithm for registration can reduce the error during point cloud registration and improve the accuracy of the registration processing results. Through this registration process, the velocity distribution of the target at the current moment can be determined.

[0028] Step S103: predicting the point cloud position change of the target to be determined at the next moment based on the velocity distribution of the target to be determined.

[0029] Because the 3D LiDAR collects point cloud data at relatively short intervals, the velocity distribution at the current moment can be used to predict the point cloud position change at the next moment. For example, the point cloud data collected at different moments also carries time series information. When predicting the point cloud position change at the next moment, this application performs time series smoothing on the point cloud data at the current moment, and can obtain the point cloud position change of the target to be determined at the next moment through smoothing prediction.

[0030] Optionally, an embodiment of the present application predicts the point cloud position change of the target to be determined at the next moment based on the speed distribution at the current moment and preset rules. The preset rules can be appropriately set by those skilled in the art according to actual needs. In one feasible manner, the preset rules can be set according to the rule conditions when training historical point cloud data. In another feasible manner, the preset rules can be determined by analyzing a large number of preset rules used when predicting the point cloud position change of a large number of targets to be determined at the current moment.

[0031] It should be noted that the time interval between the current moment and the next moment in the embodiment of the present application can be set by technical personnel in this field according to the actual application scenario or the sampling interval of the three-dimensional lidar, or it can be set by the user according to his or her own needs. The embodiment of the present application does not limit this.

[0032] The method for determining the target state provided in the first embodiment of the present application obtains the point cloud data of the target to be determined at the current moment; for the point cloud data at the current moment, a preset point cloud registration algorithm is used for registration, which can reduce the error during point cloud registration and improve the accuracy of the registration processing results. Through this registration process, the speed distribution of the target to be determined at the current moment can be obtained. According to the speed distribution of the target to be determined, the point cloud position change of the target to be determined at the next moment is predicted. Because the registration of the point cloud position is an important aspect in the registration process, and because the time interval for collecting point cloud data is short, based on this, the point cloud position change of the target to be determined at the next moment can be predicted based on the speed distribution at the current moment. As a result, the autonomous driving vehicle can more accurately perceive the state of the surrounding objects, i.e., the target to be determined, and thus execute more accurate and reasonable driving decisions, providing protection for the safe driving of the autonomous driving vehicle.

[0033] Example 2

[0034] Example 2 of the present application is based on the solution of Example 1, wherein step S102 can be implemented through steps S102a to S102c. Step S102a: according to a preset point cloud registration algorithm, determine the registration time before the current time; step S102b: obtain the point cloud data at the registration time; step S102c: use the preset point cloud registration algorithm to register the point cloud data at the current time with the point cloud data at the registration time, and obtain the speed distribution of the target to be determined at the current time according to the registration processing result.

[0035] It should be noted that in the embodiment of the present application, the registration moment is determined according to the registration accuracy of the preset point cloud registration algorithm, but in actual applications, the registration moment before the current moment may also be the previous moment of point cloud data collection (the previous point cloud data collection moment), or it may be any moment before the current moment determined according to actual conditions.

[0036] If the registration moment is the previous moment of the current moment, then when the target to be determined moves slowly, the distinction between the point cloud data at the current moment and the point cloud data at the previous moment is not obvious enough. The point cloud data at the current moment and the point cloud data at the previous moment are registered, and the error of the registration processing result obtained will be relatively large (compared to when the target to be determined moves faster). Therefore, the embodiment of the present application adopts a cross-time registration method to align the point cloud data of the target to be determined at different moments. The registration moment is determined by the registration accuracy of the point cloud registration algorithm. The embodiment of the present application uses a preset point cloud registration algorithm to align the point cloud data at the current moment with the point cloud data at the registration moment, which reduces the error in point cloud registration and improves the accuracy of the registration processing results. Based on more accurate registration processing results, a more accurate speed distribution of the target to be determined at the current moment can be obtained.

[0037] Optionally, in one embodiment of the present application, the registration moment in step S102a is determined according to the registration accuracy of the point cloud registration algorithm, the minimum instantaneous speed of the target to be determined, and the collection time interval for collecting point cloud data.

[0038] When the embodiment of the present application adopts the cross-time registration method to determine the registration time, the registration accuracy of the point cloud registration algorithm, the minimum instantaneous velocity of the target to be determined, and the collection time interval of the point cloud data are comprehensively considered, thereby further improving the accuracy of the registration time, thereby further reducing the error when aligning the point cloud data at the current moment with the point cloud data at the registration moment.

[0039] When determining the registration time to be registered with the point cloud data at the current moment, optionally, in one embodiment of the present application, the registration time to be registered with the point cloud data at the current moment is determined based on the result of (registration accuracy of the point cloud registration algorithm / collection time interval of the point cloud data / minimum instantaneous velocity of the target to be determined). Wherein, " / " represents division. For example, the accuracy of the point cloud registration algorithm is set to s cm, it is necessary to distinguish that the minimum instantaneous velocity of the target to be determined is v cm / s, and the collection time interval of the 3D laser radar sampling point cloud data is delta_ts. Taking the current moment as t and the previous moment as t-1 as an example, the registration time is ts / delta_ts / v, that is, the registration time is no longer the previous moment of the current moment, but is a moment determined by the registration accuracy of the point cloud registration algorithm, the collection time interval of the point cloud data, and the minimum instantaneous velocity of the target to be determined. Wherein, s cm, the minimum instantaneous velocity v cm / s, and the collection time interval delta_ts can be set by those skilled in the art according to the specific situation, and this embodiment of the application does not limit this. That is, in the embodiment of the present application, before the point cloud data is registered with the point cloud data at the registration moment before the current moment, the registration moment to be registered with the point cloud data at the current moment is determined according to the actual situation. The registration moment is no longer the previous moment of the current moment, thereby reducing the potential error in the point cloud position caused by the slow movement of the target to be determined, and reducing the impact of the perspective change on the point cloud registration.

[0040] Optionally, in one embodiment of the present application, step S102c can be implemented by the following steps: densifying the point cloud data at the current moment to obtain dense point cloud data at the current moment; densifying the point cloud data at the registration moment to obtain dense point cloud data at the registration moment; and using a preset point cloud registration algorithm to register the dense point cloud data at the current moment with the dense point cloud data at the registration moment.

[0041] In actual application scenarios, the reasons for changes in obstacle point cloud data may include the following two: one is the relative motion between obstacles, and the other is the change in perspective. The point cloud data collected by 3D laser radar is usually sparse, so changes in perspective lead to drastic changes in point cloud data. Taking an autonomous vehicle used in an autonomous driving scenario, which can be equipped with at least one 3D laser radar as an example, the 3D laser radar installed on the autonomous driving vehicle can collect point cloud data of obstacles around the vehicle. For the target to be determined, when the geometric information of the target to be determined is known, the point cloud registration of the previous moment and the current moment can be achieved. However, due to the sparse sampling of the 3D laser radar, the complete geometric information of the target to be determined is unknown. Therefore, the embodiment of the present application also densifies the point cloud data collected at the current moment to obtain dense point cloud data corresponding to the current moment. Compared with the point cloud data directly collected by the 3D laser radar, the dense point cloud data carries richer 3D information of the target to be determined, which makes the registration result more accurate when the point cloud registration algorithm is used for subsequent registration processing. The embodiment of the present application densifies the point cloud data, that is, densifies the geometric information of the target to be determined. Based on the more accurate registration result, a more accurate velocity distribution at the current moment can be obtained.

[0042] The densification of the point cloud data in the above steps can be achieved through the following two exemplary methods. In the first example, the point cloud data is binned and dense point cloud data is obtained based on the binning results. In the second example, the point cloud data is meshed and dense point cloud data is obtained based on the meshing results.

[0043] In the first example, point cloud data is densified using a surface element format. For example, each point cloud is modeled as a local facet, and a surface element is constructed, i.e., point cloud surface reconstruction. The reconstructed point cloud surface elements are then used to densify the point cloud data to obtain dense point cloud data. The specific method and process for constructing surface elements can be implemented by those skilled in the art using any appropriate method or algorithm based on actual needs, and this embodiment of the application does not impose any limitation thereto.

[0044] In the second example, point cloud data is densified using a gridding method, for example, triangular facets, where the point clouds are connected based on neighborhood relationships to form a grid representation, and dense point cloud data is output. The specific implementation of the gridding algorithm can also be implemented by those skilled in the art using any appropriate method or algorithm based on actual needs, and this embodiment of the application is not limited thereto.

[0045] The embodiment of the present application densifies the point cloud data of the target to be determined at the current moment and the point cloud data at the registration moment, so that the geometric information of the target to be determined is densified. Compared with the point cloud data directly collected by the three-dimensional lidar, the dense point cloud data carries richer three-dimensional information of the target to be determined, so that when the point cloud registration algorithm is subsequently used for registration processing, the registration result is more accurate.

[0046] Optionally, in one embodiment of the present application, the preset point cloud registration algorithm is a general iterative closest point (GICP) algorithm.

[0047] The preset point cloud registration algorithm in the embodiment of the present application can be an iterative closest point algorithm ICP or a generalized iterative closest point GICP (Generalized-ICP) algorithm, etc.

[0048] The embodiment of the present application adopts the energy function of GICP for modeling, determines the point cloud registration energy function, and performs registration processing on the point cloud data of the target to be determined at the current moment and the point cloud data at the registration moment based on the point cloud registration energy function, thereby improving the accuracy of the registration processing results, and thus obtaining the velocity distribution of the target to be determined at the current moment based on the registration processing results.

[0049] The second embodiment of the present application is based on the solution of the first embodiment. After step S103, the embodiment of the present application can also process the point cloud data whose point cloud position changes exceed the change threshold. Optionally, the point cloud data whose point cloud position changes exceed the change threshold can be weighted down or eliminated.

[0050] Based on the current velocity distribution, predicting the point cloud position change of the target at the next moment can be achieved through point cloud registration. Point cloud registration involves aligning the current point cloud with the point cloud at the registration moment as closely as possible. However, there may be point clouds with poor overlap, such as those with drastic changes. These poorly aligned point clouds can be identified and processed to prevent them from adversely affecting the acquisition and registration of point cloud data at the next moment. If the point cloud position changes significantly, such as exceeding a change threshold, it indicates that the point cloud data collected at the current moment may not be accurate and needs to be processed to obtain more accurate results. For example, point cloud data with point cloud position changes exceeding the change threshold is filtered out and ignored when registering the point cloud data at the next moment to improve the accuracy of the registration processing results. This allows the autonomous vehicle to more accurately perceive the state of its surrounding objects, i.e., the target, and make more accurate and reasonable driving decisions, thus ensuring the safe operation of the autonomous vehicle.

[0051] In an embodiment of the present application, when processing point cloud data whose point cloud position changes exceed a change threshold, in one implementation method, the point cloud data whose point cloud position changes exceed the change threshold is eliminated, and when the point cloud data at the next moment is aligned, the eliminated point cloud data no longer participates in the point cloud alignment processing at the next moment. In another implementation method, the point cloud whose point cloud position changes exceed the change threshold is matched with a lower weight value to reduce its influence on the point cloud alignment at the next moment, so as to improve the accuracy of the alignment processing results.

[0052] Example 3:

[0053] The second embodiment of the present application is based on the solution of the first embodiment. Optionally, in one embodiment of the present application, after obtaining the current speed distribution of the target to be determined according to any of the first to second embodiments, the target state determination method further includes the following steps: step S301 and step S302. Figure 3 As shown, Figure 3 A flowchart of another method for determining a target state provided in an embodiment of the present application. The method for determining a target state includes the following steps:

[0054] Step S301 : extracting dynamic and static features of the velocity distribution at the registration moment and the velocity distribution at the current moment respectively, to obtain the dynamic and static features of the target to be determined at the current moment.

[0055] Among them, the dynamic and static features include: the dynamic probability and static probability of the target to be determined.

[0056] It should be noted that, taking the autonomous driving system used in autonomous driving scenarios as an example, the normal operation of the autonomous driving system requires the coordinated cooperation of multiple modules. Among them, the perception module, as the eyes of the autonomous driving system, plays a vital role in the safety of the autonomous driving system. The speed estimation and dynamic and static judgment of obstacles around the vehicle are important links in the perception module, which effectively ensures the safety of the autonomous driving system and provides a reliable basis for subsequent decision-making and control. Therefore, after obtaining the speed distribution of the target to be determined at the current moment, the embodiment of the present application also judges the dynamic and static characteristics of the target at the current moment based on the speed distribution of the target to be determined at the current moment.

[0057] In an embodiment of the present application, the speed distribution can be characterized by a speed probability density distribution function, and dynamic and static features are extracted for the speed distribution at the registration moment and the speed probability density distribution function at the current moment respectively. The dynamic and static features of the target to be determined are calculated based on the extracted dynamic and static features. In an embodiment of the present application, the dynamic and static features include the probability that the target to be determined is in a dynamic state and the probability that it is in a static state at the current moment. Optionally, after normalizing the probability of being in a dynamic state and the probability of being in a static state, the sum of the probability of the target to be determined being in a dynamic state and the probability of being in a static state at the current moment is 1.

[0058] Optionally, in an embodiment of the present application, the dynamic and static characteristics further include: an uncertainty probability for indicating the uncertainty of the target state to be determined.

[0059] For ease of understanding, in this example, the target to be determined is any obstacle around the vehicle. Figure 2 To illustrate, in actual application scenarios, there may be obstacle states that the driving control device in the autonomous vehicle cannot determine. For example, there may be a sudden change in the point cloud data at a certain obstacle registration moment and the current moment. This causes a large error in the registration result when the point cloud data at the current moment is registered with the point cloud data at the registration moment using a preset point cloud registration algorithm. The speed distribution of the obstacle obtained at the current moment is inaccurate, which in turn causes the dynamic and static characteristics of the obstacle at the current moment to include uncertainty. In the embodiment of the present application, the dynamic and static characteristics of the obstacle at the current moment also include an uncertainty probability for indicating the uncertainty of the obstacle state. This uncertainty probability can be used to prompt or alert the driver of the autonomous vehicle, allowing the driver of the autonomous vehicle to manually judge the obstacle based on the actual road conditions and determine whether the autonomous vehicle needs to avoid or execute other more accurate and reasonable driving decisions, thereby providing protection for the safe driving of the autonomous vehicle.

[0060] In an embodiment of the present application, the dynamic and static characteristics include the probability that the target to be determined is in dynamic state at the current moment, the probability that it is in static state, and the probability that it is in uncertainty. Optionally, after normalizing the probability that it is in dynamic state, the probability that it is in static state, and the probability that it is in uncertainty, the sum of the probabilities that the target to be determined is in dynamic state, the probability that it is in static state, and the probability that it is in uncertainty at the current moment is 1.

[0061] Optionally, in an embodiment of the present application, step S301 may be implemented as step S301a and step S301b.

[0062] Step S301a: Obtain a velocity direction consistency measure, a state uncertainty measure, and a static state confidence of the target to be determined based on the velocity distribution at the registration moment and the velocity distribution at the current moment.

[0063] The consistency of the speed direction of the target at the registration moment and the current moment, that is, the variance of the speed direction, can be determined. For example, the circular statistics technology can be used to express the consistency of the speed direction. It is understandable that other technologies for expressing the consistency of the speed direction can also be used. The embodiment of the present application is only an exemplary representation and does not mean that the embodiment of the present application is limited to this. When determining the speed direction consistency measure, in one implementable method, the weighted average of the speed probability density distribution function is adopted. In another implementable method, the topk form is adopted to obtain the speed direction consistency measure according to the variance of the speed direction of the top K largest speed probability density distribution functions. In the embodiment of the present application, the obtained speed direction consistency measure is set to h, the h of the dynamic features and the static features are counted, and the distribution of the speed direction consistency is fitted by at least one method. According to the fitted distribution, h is normalized to between 0 and 1 by a normalization function, and the normalized h is set to ph. It is understandable that the embodiment of the present application can also use ph to characterize the speed direction consistency measure.

[0064] The state uncertainty measurement of the registration moment and the current moment of the target to be determined, the embodiment of the present application models the unimodal characteristics of the velocity probability density distribution function through uncertainty modeling, and determines the state uncertainty measurement u according to the standard deviation of the velocity probability density distribution function, and then normalizes u to between 0 and 1 through the normalization function, and sets the normalized u to pu. It can be understood that the embodiment of the present application can also use pu to represent the state uncertainty measurement.

[0065] The registration moment of the target to be determined and the static state confidence at the current moment, that is, in the velocity probability density distribution function, the static state confidence s is determined according to the probability density of the speed being 0, and then s is normalized to between 0-1 through the normalization function, and the normalized s is set to ps. It can be understood that the embodiment of the present application can also use ps to represent the static state confidence.

[0066] The normalization function in the embodiment of the present application can be a softmax function or other mapping functions used for normalization, and the embodiment of the present application does not limit this.

[0067] Step S301b: Obtain the dynamic and static characteristics of the target to be determined at the current moment based on the speed direction consistency measurement, the state uncertainty measurement and the static state confidence.

[0068] Optionally, in one embodiment of the present application, step S301b can be implemented in at least one of the following ways: determining the difference between a preset standard value and the product of a state uncertainty measure and a static state confidence; determining the dynamic probability of the target to be determined based on the product of the difference and the speed direction consistency measure; and / or determining the difference between the preset standard value and the speed direction consistency measure, and determining the static probability of the target to be determined based on the difference; and / or determining the uncertainty probability of the target to be determined based on the product of the speed direction consistency measure, the state uncertainty measure and the static state confidence.

[0069] It should be noted that, if the embodiment of the present application uses ph to represent the speed direction consistency measure, pu to represent the state uncertainty measure, and ps to represent the static state confidence, because the above ph, pu and ps are all normalized values, then 1 is used to represent the preset standard value in the embodiment of the present application. It can be understood that, if the embodiment of the present application uses h to represent the speed direction consistency measure, u to represent the state uncertainty measure, and s to represent the static state confidence, then the preset standard value can be set by a technician in this field after a comprehensive analysis of h, u and s.

[0070] In the embodiment of the present application, ph is used to represent the speed direction consistency measurement, pu is used to represent the state uncertainty measurement, ps is used to represent the static state confidence, and 1 is used to represent the preset standard value. If the dynamic and static characteristics include the dynamic probability, static probability and uncertainty probability of the target to be determined, the dynamic probability is expressed as (1-pu×ps)×ph, the static probability is expressed as 1-ph, and the uncertainty probability is expressed as pu×ps×ph.

[0071] If the dynamic and static features include the dynamic probability and static probability of the target to be determined, the dynamic probability is expressed as [(1-pu×ps)×ph] / (1-pu×ps×ph), and the dynamic probability is expressed as (1-ph) / (1-pu×ps×ph), where the symbol “ / ” represents division.

[0072] Step S302: Determine the current state of the target to be determined based on the dynamic and static characteristics of the target to be determined at the current moment.

[0073] Step S302 may be implemented by a first state determination operation and a second state determination operation. For example, the first state determination operation includes: performing a first classification process on the dynamic and static features of the target to be determined at the current moment using a hidden Markov model, and determining, based on the first classification result, whether the target to be determined is in a moving state, a stationary state, or an uncertain state at the current moment.

[0074] In this embodiment, a state sequence is modeled using a Hidden Markov Model (HMM). The HMM is used to perform a first classification process on the dynamic and static features. A transition probability matrix is set based on the dynamic and static features, and the HMM is used to classify and solve the dynamic and static features. Based on the first classification results, it is determined whether the target is currently in motion, stationary, or uncertain.

[0075] For example, the second state determination operation includes: if the first classification results of multiple consecutive moments all indicate that the target to be determined is in a set state, then the last moment of the multiple consecutive moments is used as the reference moment, where the set state includes a stationary state or an uncertain state; after the reference moment, the current moment is updated at a preset time interval, and the real-time speed distribution of the current moment is obtained, and the state of the target to be determined is determined according to the real-time speed distribution and the speed distribution corresponding to the reference moment.

[0076] In embodiments of the present application, the set state can include a stationary state or an uncertain state. If the first classification results for multiple consecutive moments indicate that the target to be determined is in the set state, it is unknown whether the target to be determined is in motion. The first state determination operation is no longer necessary to determine the target's state, and the second state determination operation is required to determine the target's state. For example, if point cloud registration is performed on the point cloud data of the target to be determined at the current moment and the registration moment using a 0.1s interval, and multiple consecutive 0.1s first classification results all indicate that the target to be determined is in the set state, the last 0.1s moment in the multiple 0.1s intervals is used as the reference moment. After the reference moment, the current moment is updated at 1s intervals, and the point cloud data at the current moment is registered with the point cloud data at the reference moment to obtain the real-time velocity distribution at the current moment. The state of the target to be determined is determined based on the real-time velocity distribution and the velocity distribution corresponding to the reference moment. The time span between the real-time velocity distribution at the current moment and the reference moment is relatively long, which not only saves time and computing resources, but also makes the state of the target to be determined determined based on the real-time velocity distribution at the current moment and the velocity distribution at the set state more accurate. It is understandable that the preset time interval can be set by those skilled in the art according to actual conditions. The embodiments of the present application are described using 0.1s and 1s as examples, which does not mean that the embodiments of the present application are limited thereto.

[0077] When the second state determination operation is started, optionally, in one embodiment of the present application, the method for determining the target state further includes: using the state of the determined target to be determined to update the state of the target to be determined determined by the first state determination operation.

[0078] In the embodiment of the present application, when the second state determination operation starts working, the state of the target to be determined determined by the first state determination operation is replaced by the state determined by the second state operation. That is to say, it is impossible to judge whether the target to be determined is moving according to the first state determination operation, and its corresponding target state to be determined needs to be replaced with the target state to be determined determined by the second state operation, so that the determined target state to be determined is more accurate, which helps the driving control device of the autonomous driving vehicle to execute accurate and reasonable driving decisions according to the accurate target state to be determined, and provides protection for the safe driving of the autonomous driving vehicle.

[0079] Optionally, in one embodiment of the present application, the method for determining the target state further includes: monitoring the state of the target to be determined determined by the first state determination operation, and if it is monitored that the target to be determined changes from a stationary state or an uncertain state to a moving state, ending the second state determination operation.

[0080] The second state determination operation is an operation used only when the state of the target to be determined determined by the first state determination operation is a stationary state or an uncertain state. The embodiment of the present application also monitors the state of the target to be determined determined by the first state determination operation. If it is monitored that the target to be determined changes from a stationary state or an uncertain state to a moving state, it means that the second state determination operation is no longer applicable when determining the state of the target to be determined. In this case, the second state determination operation is terminated, which not only saves computing resources, but also makes the state of the target to be determined determined according to the first state determination operation more accurate in subsequent moments, which helps the driving control device of the autonomous driving vehicle to execute accurate and reasonable driving decisions according to the accurate state of the target to be determined, and provides protection for the safe driving of the autonomous driving vehicle.

[0081] Furthermore, a specific example is given to illustrate the determination of the target state in the embodiment of the present application. The data processing includes speed estimation and dynamic and static classification. For ease of understanding, this example takes the target to be determined as any obstacle around the vehicle as an example for explanation, as follows.

[0082] (1) Speed estimation

[0083] When estimating the speed of an obstacle at the current moment, the velocity probability density distribution function of the obstacle at the current moment is obtained through the point cloud registration algorithm and the point cloud data registration error. The velocity probability density distribution function represents the velocity distribution, and the point cloud position change of the obstacle at the next moment is predicted based on the current velocity distribution, as shown below.

[0084] (1) Global point cloud registration

[0085] This embodiment estimates the speed of an obstacle based on changes in its current point cloud data, position, and rotation. For example, a point cloud registration algorithm is used to register the current point cloud data with the point cloud data from a previous registration moment. The registration results are used to determine the obstacle's current velocity distribution. Based on this velocity distribution, the obstacle's point cloud position change at the next moment is predicted.

[0086] Among them, the point cloud registration algorithm can adopt a variety of global optimization algorithms, such as simulated annealing algorithm and grid search algorithm.

[0087] The embodiment of the present application uses a 3D laser radar installed on an autonomous vehicle to collect point cloud data of obstacles around the vehicle. For obstacles, when the geometric information of the obstacle is known, point cloud registration can be achieved between the registration time and the current time. However, due to the sparse sampling of the 3D laser radar, the complete geometric information of the obstacle is unknown. Therefore, the embodiment of the present application also densifies the point cloud data collected at the current time to obtain dense point cloud data corresponding to the current time. Compared with the point cloud data directly collected by the 3D laser radar, the dense point cloud data carries richer 3D information about the obstacle, which makes the registration result more accurate when the point cloud registration algorithm is used for subsequent registration processing.

[0088] In the embodiment of the present application, point cloud data is densified in the following ways to obtain dense point cloud data: (1) facet form, each point cloud is modeled as a local small plane, and facets are constructed, that is, point cloud surface reconstruction, and the reconstructed point cloud facets are used to densify the point cloud data to obtain dense point cloud data; (2) meshing, such as triangular facets, the point clouds are connected to each other according to the neighborhood relationship to form a mesh expression, thereby outputting dense point cloud data.

[0089] This embodiment of the application uses a point cloud registration algorithm to register the dense point cloud data at the current moment with the dense point cloud data at the previous registration moment. The velocity distribution of the obstacle at the current moment is obtained based on the registration results. Compared to point cloud data directly collected by a 3D lidar, this dense point cloud data carries richer 3D information about the obstacle, resulting in more accurate registration results when the point cloud registration algorithm is subsequently used.

[0090] (2) Point cloud data registration error

[0091] If the above-mentioned registration moment is the previous moment of the current moment, then when the obstacle moves slowly, the distinction between the point cloud data at the current moment and the point cloud data at the previous moment is not obvious enough. The point cloud data at the current moment and the point cloud data at the previous moment are registered, and the error of the registration processing result obtained will be relatively large (compared to when the obstacle moves faster). Therefore, the embodiment of the present application adopts a cross-time registration method to align the point cloud data of the obstacle at different moments.

[0092] In practical applications, the reasons for changes in obstacle point cloud data can include the following two: one is the relative motion between obstacles, and the other is the change in perspective. 3D LiDARs are sparse, so changes in perspective lead to drastic changes in point cloud data, which in turn leads to inaccurate obstacle velocity probability density distribution functions. Therefore, the embodiments of this application model the potential errors in point cloud position to reduce the impact of perspective changes on the alignment of current point cloud data with previous point cloud data. The registration moment is not just the moment before the current moment, that is, the registration moment and the current moment are across time. The accuracy of the point cloud registration algorithm is set to s cm, and the minimum instantaneous speed of obstacles to be distinguished is v cm / s. The collection time interval of the 3D lidar sampling point cloud data is delta_ts. Taking the current moment as t and the previous moment as t-1 as an example, the registration moment is (ts / delta_ts / v). The symbol " / " in the formula represents division, that is, the registration moment is no longer the moment before the current moment, but the moment determined by the registration accuracy of the point cloud registration algorithm, the collection time interval of the point cloud data, and the minimum instantaneous speed of the obstacle.

[0093] The embodiment of the present application adopts the GICP (Generalized-ICP) energy function for modeling to obtain the point cloud registration energy function. According to the point cloud registration energy function, the point cloud data of the obstacle at the current moment is registered with the point cloud data at the registration moment, so as to obtain the velocity distribution of the obstacle at the current moment according to the registration processing result.

[0094] Furthermore, because the time interval for collecting point cloud data by the three-dimensional laser radar is short, the embodiment of the present application can predict the change in the point cloud position at the next moment based on the speed distribution of the obstacle at the current moment.

[0095] (2) Dynamic and static classification

[0096] The embodiment of the present application extracts dynamic and static features from the velocity probability density distribution function at the current moment and the registration moment, and classifies the dynamic and static features to obtain the current state of the obstacle, as follows.

[0097] (1) Dynamic and static feature extraction

[0098] The embodiment of the present application also extracts dynamic and static features from the velocity probability density distribution function. If the dynamic and static features include the dynamic probability, static probability, and uncertainty probability of the obstacle, the dynamic and static features are expressed as a three-dimensional vector [dynamic probability, static probability, uncertainty probability], and the cumulative probability of the three is 1.0; if the dynamic and static features include the dynamic probability and static probability of the obstacle, the dynamic and static features are expressed as [dynamic probability, static probability], and the cumulative probability of the two is 1.0. The dynamic and static features extracted by this application are as follows:

[0099] 1.1. The speed direction consistency of the two frames before and after (twice collected point cloud data), that is, the variance of the speed direction. The embodiment of the present application uses circular statistics technology to express the speed direction consistency. Similar technology can also be used. When determining the speed direction consistency measure, in one implementable method, the weighted average of the speed probability density distribution is adopted. In another implementable method, the topk form is adopted to obtain the speed direction consistency measure based on the variance of the speed direction of the top K largest speed probability density distributions. First, the speed direction consistency measure is set to h, and the h of the dynamic features and static features are statistically analyzed. The distribution of the speed direction consistency is fitted in a variety of ways. The fitted distribution is used to normalize h to between 0 and 1, and the normalized h is set to ph, and ph is used to characterize the final speed direction consistency measure.

[0100] 1.2 Uncertainty modeling is mainly used to model the unimodal characteristics of the velocity probability density distribution function. The standard deviation of the velocity distribution is used to express the uncertainty measure u. Then, u is mapped to the range of 0-1 through a mapping function. The mapping function can be any normalization function, such as the softmax function. The normalized u is set to pu, and pu is used to represent the final state uncertainty measure.

[0101] 1.3. The confidence level of the stationary state. That is, in the velocity probability density distribution function, the stationary state confidence level s is determined based on the probability density of a velocity of 0. Then, s is mapped to a value between 0 and 1 through a mapping function. The mapping function can be any normalization function, such as the softmax function. The normalized s is set to ps, and ps is used to represent the final stationary state confidence level.

[0102] If the dynamic and static features include the dynamic probability, static probability, and uncertainty probability of an obstacle, the dynamic probability is expressed as (1-pu×ps)×ph, the static probability is expressed as 1-ph, and the uncertainty probability is expressed as pu×ps×ph. If the dynamic and static features include the dynamic and static probabilities of an obstacle, the dynamic probability is expressed as [(1-pu×ps)×ph] / (1-pu×ps×ph), and the dynamic probability is expressed as (1-ph) / (1-pu×ps×ph), where the " / " symbol indicates division.

[0103] (2) Dynamic and static classification

[0104] The embodiment of the present application models the state sequence based on the Hidden Markov Model (HMM), uses the dynamic and static features extracted above, sets an appropriate transition probability matrix, and uses an existing algorithm to solve the dynamic and static features. This process can be called short-term dynamic and static classification. The embodiment of the present application also proposes long-term dynamic and static classification to detect ultra-low-speed obstacles. The specific process is as follows:

[0105] When the obstacle is stationary or uncertain in the short-term dynamic and static classification results, the long-term dynamic and static classification begins. The long-term dynamic and static classification process is as follows:

[0106] 2.1. Count the velocity direction consistency features at all static moments starting from the static state, and obtain the velocity direction consistency measure ph.

[0107] 2.2. Record the static moment of the first frame, and then register the point cloud of the first frame at each subsequent moment or at regular intervals. Registering the point cloud at regular intervals, i.e., selecting point clouds with longer intervals to register with the point cloud of the first frame, can save time and improve efficiency. The uncertainty measure pu and the static state confidence ps are statistically obtained.

[0108] 2.3. When the long-term dynamic and static classification starts, the final result of the short-term dynamic and static classification is replaced by the long-term dynamic and static classification result.

[0109] 2.4. If the classification result of the short-term dynamic and static classification is a moving state, the long-term dynamic and static classification is terminated.

[0110] Example 4:

[0111] The embodiment of the present application provides a device for determining a target state, such as Figure 4 As shown, Figure 4An apparatus for determining a target state provided in an embodiment of the present application, the apparatus for determining a target state 40 includes an acquisition module 401, a registration module 402, and a prediction module 403;

[0112] The acquisition module 401 is used to obtain the point cloud data of the target to be determined at the current moment;

[0113] The registration module 402 is used to determine the velocity distribution of the target to be determined using a preset point cloud registration algorithm for the point cloud data at the current moment;

[0114] The prediction module 403 is used to predict the point cloud position change of the target to be determined at the next moment according to the velocity distribution of the target to be determined.

[0115] Optionally, in one embodiment of the present application, the registration module 402 is also used to determine the registration time before the current time according to a preset point cloud registration algorithm; obtain the point cloud data at the registration time; use the preset point cloud registration algorithm to align the point cloud data at the current time with the point cloud data at the registration time, and obtain the speed distribution of the target to be determined at the current time based on the registration processing result.

[0116] Optionally, in an embodiment of the present application, the registration time is determined according to the registration accuracy of the point cloud registration algorithm, the minimum instantaneous speed of the target to be determined, and the collection time interval for collecting point cloud data.

[0117] Optionally, in one embodiment of the present application, the target state determination device 40 also includes a densification processing module, which is used to densify the point cloud data at the current moment to obtain dense point cloud data at the current moment; densify the point cloud data at the registration moment to obtain dense point cloud data at the registration moment; and the registration module 402 is used to use a preset point cloud registration algorithm to perform registration processing on the dense point cloud data at the current moment and the dense point cloud data at the registration moment.

[0118] Optionally, in one embodiment of the present application, the densification processing module is also used to perform faceting processing on the point cloud data, and obtain dense point cloud data based on the faceting processing results; or, to perform gridding processing on the point cloud data, and obtain dense point cloud data based on the networking processing results.

[0119] Optionally, in one embodiment of the present application, the preset point cloud registration algorithm is a general iterative closest point (GICP) algorithm.

[0120] Optionally, in one embodiment of the present application, the target state determination device 40 also includes a feature module and a state determination module, the feature module is used to extract dynamic and static features of the velocity distribution at the registration moment and the velocity distribution at the current moment, respectively, to obtain the dynamic and static features of the target to be determined at the current moment, wherein the dynamic and static features include: the dynamic probability and static probability of the target to be determined; the state determination module is used to determine the state of the target to be determined at the current moment based on the dynamic and static features of the target to be determined at the current moment.

[0121] Optionally, in an embodiment of the present application, the dynamic and static characteristics further include: an uncertainty probability for indicating the uncertainty of the target state to be determined.

[0122] Optionally, in one embodiment of the present application, the feature module is also used to obtain the velocity direction consistency measurement, state uncertainty measurement and static state confidence of the target to be determined based on the velocity distribution at the registration moment and the velocity distribution at the current moment; and obtain the dynamic and static characteristics of the target to be determined at the current moment based on the velocity direction consistency measurement, state uncertainty measurement and static state confidence.

[0123] Optionally, in one embodiment of the present application, the feature module is also used to determine the difference between a preset standard value and the product of a state uncertainty measure and a static state confidence; determine the dynamic probability of the target to be determined based on the product of the difference and the speed direction consistency measure; and / or determine the difference between the preset standard value and the speed direction consistency measure, and determine the static probability of the target to be determined based on the difference; and / or determine the uncertainty probability of the target to be determined based on the product of the speed direction consistency measure, the state uncertainty measure and the static state confidence.

[0124] Optionally, in one embodiment of the present application, the state determination module also includes a first state determination unit, which is used for a first state determination operation, including: performing a first classification process on the dynamic and static features of the target to be determined at the current moment through a hidden Markov model, and determining whether the target to be determined is in a motion state, a stationary state, or an uncertain state at the current moment according to the first classification result.

[0125] Optionally, in one embodiment of the present application, the state determination module also includes a second determination unit, and the second determination unit is used to take the last moment of the multiple consecutive moments as the reference moment if the first classification results of multiple consecutive moments all indicate that the target to be determined is in a set state, wherein the set state includes a stationary state or an uncertain state; after the reference moment, update the current moment at a preset time interval, and obtain the real-time speed distribution of the current moment, and determine the state of the target to be determined based on the real-time speed distribution and the speed distribution corresponding to the reference moment.

[0126] Optionally, in an embodiment of the present application, the target state determination device 40 further includes an updating module, and the updating module is used to update the state of the target to be determined determined by the first state determination operation using the determined state of the target to be determined.

[0127] Optionally, in one embodiment of the present application, the target state determination device 40 also includes an ending module, which is used to monitor the state of the target to be determined determined by the first state determination operation. If it is monitored that the target to be determined changes from a stationary state or an uncertain state to a moving state, the second state determination operation is ended.

[0128] Optionally, in an embodiment of the present application, the target state determination device 40 further includes a processing module, and the processing module is used to reduce the weight of or eliminate point cloud data whose point cloud position changes exceed a change threshold.

[0129] The target state determination device 40 of the embodiment of the present application is used to implement the target state determination methods corresponding to the aforementioned multiple method embodiments and has the beneficial effects of the corresponding method embodiments, which will not be described in detail here. In addition, the functional implementation of each module in the target state determination device of the embodiment of the present application can refer to the description of the corresponding parts in the aforementioned method embodiments, which will not be described in detail here.

[0130] Of course, these algorithm modules will vary depending on the type of autonomous vehicle. For example, different algorithm modules will be used for logistics vehicles, public service vehicles, medical service vehicles, and terminal service vehicles. The following examples illustrate the algorithm modules for these four types of autonomous vehicles:

[0131] Logistics vehicles refer to vehicles used in logistics scenarios, such as those with automatic sorting functions, refrigeration and insulation functions, and measurement functions. These logistics vehicles involve different algorithm modules.

[0132] For example, logistics vehicles can be equipped with automated sorting devices that automatically retrieve, transport, sort, and store goods upon arrival. This involves an algorithm module for cargo sorting, which primarily implements logical control over cargo retrieval, transportation, sorting, and storage.

[0133] For another example, for cold chain logistics scenarios, logistics vehicles can also be equipped with refrigeration and insulation devices, which can refrigerate or insulate the transported fruits, vegetables, aquatic products, frozen foods, and other perishable foods, so that they are in a suitable temperature environment, solving the problem of long-distance transportation of perishable foods. This involves an algorithm module for refrigeration and insulation control, which is mainly used to dynamically and adaptively calculate the appropriate temperature for cold meals or insulation based on information such as the nature of the food (or item), perishability, transportation time, current season, and climate, and automatically adjust the refrigeration and insulation device according to the appropriate temperature. In this way, when the vehicle transports different foods or items, the transportation personnel do not need to manually adjust the temperature, freeing the transportation personnel from the tedious temperature control and improving the efficiency of refrigerated and insulated transportation.

[0134] For example, in most logistics scenarios, charges are based on the volume and / or weight of the package. However, the number of logistics packages is very large. Simply relying on couriers to measure the volume and / or weight of the packages is very inefficient and has high labor costs. Therefore, in some logistics vehicles, measuring devices are added to automatically measure the volume and / or weight of logistics packages and calculate the fees for logistics packages. This involves an algorithm module for logistics package measurement, which is mainly used to identify the type of logistics package and determine the measurement method of the logistics package, such as volume measurement or weight measurement or a combination of volume and weight measurement. It can also complete the volume and / or weight measurement according to the determined measurement method, and complete the fee calculation based on the measurement results.

[0135] Public service vehicles refer to vehicles that provide certain public services, such as fire trucks, de-icing trucks, sprinkler trucks, snowplows, garbage disposal vehicles, traffic control vehicles, etc. These public service vehicles involve different algorithm modules.

[0136] For example, for an autonomous fire truck, its main task is to carry out reasonable fire-fighting tasks at the fire scene. This involves an algorithm module for fire-fighting tasks. The algorithm module must at least implement logic such as fire condition identification, fire-fighting plan planning, and automatic control of fire-fighting equipment.

[0137] For example, the main task of a de-icing vehicle is to clear ice and snow from the road surface, which involves a de-icing algorithm module. This algorithm module must at least be able to identify the ice and snow conditions on the road surface, formulate a de-icing plan based on the ice and snow conditions, such as which sections of the road require de-icing and which sections do not, whether to use salting and the amount of salt to spread, etc., as well as the logic for automatic control of the de-icing device when the de-icing plan is determined.

[0138] Among them, medical service vehicles refer to self-driving vehicles that can provide one or more medical services. Such vehicles can provide medical services such as disinfection, temperature measurement, medication, and isolation. This involves algorithm modules that provide various self-service medical services. These algorithm modules mainly realize the identification of disinfection needs and the control of disinfection devices so that the disinfection devices can disinfect patients, or identify the patient's position and control the temperature measuring device to automatically approach the patient's forehead and other positions to measure the patient's temperature, or are used to realize the judgment of the disease, give a prescription based on the judgment result, and need to realize the identification of drugs / drug containers, as well as the control of the drug-taking robot so that it can grab drugs for patients according to the prescription, etc.

[0139] Among them, terminal service vehicles refer to self-service autonomous driving vehicles that can replace some terminal devices to provide certain convenient services to users. For example, these vehicles can provide users with printing, attendance, scanning, unlocking, payment, retail and other services.

[0140] For example, in some application scenarios, users often need to go to a specific location to print or scan documents, which is time-consuming and labor-intensive. Therefore, a terminal service vehicle that can provide users with printing / scanning services has emerged. These service vehicles can be interconnected with the user's terminal device. The user issues a print instruction through the terminal device, and the service vehicle responds to the print instruction, automatically prints the document required by the user, and can automatically deliver the printed document to the user's location. The user does not need to queue at the printer, which can greatly improve printing efficiency. Alternatively, it can respond to the scanning instruction issued by the user through the terminal device and move to the user's location. The user places the document to be scanned on the scanning tool of the service vehicle to complete the scanning, without having to queue at the printer / scanner, saving time and effort. This involves an algorithm module that provides printing / scanning services. The algorithm module at least needs to identify the connection with the user's terminal device, the response to the print / scan instruction, the positioning of the user's location, and the travel control.

[0141] For example, with the development of new retail businesses, more and more e-commerce companies are using self-service vending machines to deliver goods to office buildings and public areas. However, these vending machines are stationary and immovable, requiring users to visit them to purchase their desired items, making them relatively inconvenient. Consequently, self-driving vehicles have emerged to provide retail services. These vehicles can carry goods and move autonomously, offering a corresponding self-service shopping app or portal. Users can use their mobile phones or other devices to place orders with the self-driving vehicles through the app or portal. The order includes the product name, quantity, and user location. After receiving the order, the vehicle can determine whether the requested item is available and whether the quantity is sufficient. If the requested item is available and sufficient, it can automatically move to the user's location with the item and deliver it to the user, further improving shopping convenience and saving time, allowing users to focus on more important tasks. This involves the algorithm modules that provide retail services, which primarily implement logic for responding to user order requests, processing orders, maintaining product information, locating the user, and managing payments.

[0142] Example 5

[0143] Based on any one of the target state determination methods described in the above embodiments 1 to 3, the present application embodiment provides an electronic device. It should be noted that the target state determination method of the present application embodiment can be executed by any appropriate electronic device with the ability to determine the target state, including but not limited to: servers, mobile terminals (such as mobile phones, PADs, etc.) and PCs, etc. Figure 5 As shown, Figure 5 This is a block diagram of an electronic device provided in an embodiment of the present application. The specific embodiments of the present application do not limit the specific implementation of the electronic device. The electronic device 50 may include: a processor 502, a communications interface 504, a memory 506, and a communication bus 508.

[0144] The processor 502 , the communication interface 504 , and the memory 506 communicate with each other via a communication bus 508 .

[0145] The communication interface 504 is used to communicate with other electronic devices or servers.

[0146] The processor 502 is configured to execute the computer program 510 , and specifically may execute the relevant steps in the above-mentioned target state determination method embodiment.

[0147] Specifically, the computer program 510 may include computer program codes including computer operating instructions.

[0148] Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application. The one or more processors included in the smart device may be processors of the same type, such as one or more CPUs, or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0149] The memory 506 is used to store the computer program 510. The memory 506 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0150] The methods for determining the target state in the aforementioned embodiments provided in the embodiments of the present application can all be executed by the electronic device 50 . Accordingly, the target state determination device in the aforementioned embodiments is provided in the electronic device 50 .

[0151] The specific implementation of each step in program 510 can be found in the corresponding descriptions of the corresponding steps and units in the above-mentioned target state determination method embodiment, and will not be repeated here. Those skilled in the art will clearly understand that for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding process descriptions in the above-mentioned method embodiment, and will not be repeated here.

[0152] It should be pointed out that, according to the needs of implementation, the various components / steps described in the embodiments of the present application can be split into more components / steps, or two or more components / steps or partial operations of components / steps can be combined into new components / steps to achieve the purpose of the embodiments of the present application.

[0153] The above-mentioned method according to the embodiment of the present application can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded via a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor or hardware, the method for determining the target state described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method for determining the target state shown here, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method for determining the target state shown here.

[0154] Those skilled in the art will appreciate that the various exemplary units and method steps described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the embodiments of this application.

[0155] The above implementation methods are only used to illustrate the embodiments of the present application, and are not intended to limit the embodiments of the present application. Ordinary technicians in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the embodiments of the present application. Therefore, all equivalent technical solutions also fall within the scope of the embodiments of the present application, and the scope of patent protection of the embodiments of the present application should be defined by the claims.

Claims

1. A method for determining a target state, comprising: Obtain the point cloud data of the target to be determined at the current moment; Determining a registration time before the current time according to a preset point cloud registration algorithm; Obtaining point cloud data at the registration moment; Using a preset point cloud registration algorithm, the point cloud data at the current moment is registered with the point cloud data at the registration moment, and the velocity distribution of the target to be determined at the current moment is obtained according to the registration result; Predicting a point cloud position change of the target to be determined at a next moment based on the velocity distribution of the target to be determined; The registration time is determined according to the registration accuracy of the point cloud registration algorithm, the minimum instantaneous speed of the target to be determined, and the collection time interval of the point cloud data when performing cross-time registration.

2. The method according to claim 1, wherein Using the preset point cloud registration algorithm, registering the point cloud data at the current moment with the point cloud data at the registration moment, including: Densifying the point cloud data at the current moment to obtain dense point cloud data at the current moment; Densifying the point cloud data at the registration moment to obtain dense point cloud data at the registration moment; The preset point cloud registration algorithm is used to perform registration processing on the dense point cloud data at the current moment and the dense point cloud data at the registration moment.

3. The method according to claim 2, wherein: The densification process includes: Performing faceting processing on the point cloud data, and obtaining dense point cloud data according to the faceting processing results; or, The point cloud data is gridded and dense point cloud data is obtained based on the network processing results.

4. The method according to any one of claims 1 to 3, wherein The preset point cloud registration algorithm is the general iterative closest point (GICP) algorithm.

5. The method according to claim 1, wherein The method further comprises: Extracting dynamic and static features from the velocity distribution at the registration moment and the velocity distribution at the current moment, respectively, to obtain dynamic and static features of the target to be determined at the current moment, wherein the dynamic and static features include: a dynamic probability and a static probability of the target to be determined; The state of the target to be determined at the current moment is determined according to the dynamic and static characteristics of the target to be determined at the current moment.

6. The method according to claim 5, wherein: The dynamic and static characteristics further include: an uncertainty probability for indicating the uncertainty of the target state to be determined.

7. The method according to claim 5 or 6, wherein: The extracting of dynamic and static features of the velocity distribution at the registration moment and the velocity distribution at the current moment respectively to obtain the dynamic and static features of the target to be determined at the current moment includes: Obtaining a velocity direction consistency measure, a state uncertainty measure, and a static state confidence of the target to be determined according to the velocity distribution at the registration moment and the velocity distribution at the current moment; The dynamic and static characteristics of the target to be determined at the current moment are obtained according to the speed direction consistency measure, the state uncertainty measure and the static state confidence.

8. The method according to claim 7, wherein: The obtaining of the dynamic and static characteristics of the target to be determined at the current moment according to the velocity direction consistency metric, the state uncertainty metric, and the static state confidence level includes: Determine a difference between a preset standard value and a product of the state uncertainty measure and the static state confidence; and determine a dynamic probability of the target to be determined based on the product of the difference and the velocity direction consistency measure; and / or, Determining a difference between the preset standard value and the velocity direction consistency measure, and determining a static probability of the target to be determined based on the difference; and / or, The uncertainty probability of the target to be determined is determined according to the product of the velocity direction consistency measure, the state uncertainty measure and the static state confidence.

9. The method according to claim 6, wherein: The determining the state of the target to be determined at the current moment according to the dynamic and static characteristics of the target to be determined at the current moment includes: The first state determination operation includes: performing a first classification process on the dynamic and static features of the target to be determined at the current moment through a hidden Markov model, and determining whether the target to be determined is in a moving state, a stationary state, or an uncertain state at the current moment according to the first classification result.

10. The method according to claim 9, wherein: The method further comprises: The second state determination operation includes: If the first classification results at a plurality of consecutive moments all indicate that the target to be determined is in a set state, then the last moment of the plurality of consecutive moments is used as a reference moment, wherein the set state includes the static state or the uncertain state; After the reference moment, the current moment is updated at preset time intervals, and the real-time speed distribution of the current moment is obtained. The state of the target to be determined is determined based on the real-time speed distribution and the speed distribution corresponding to the reference moment.

11. The method according to claim 10, wherein: The method further comprises: The determined state of the target to be determined is used to update the state of the target to be determined determined by the first state determination operation.

12. The method according to claim 10, wherein: The method further comprises: The state of the target to be determined determined by the first state determination operation is monitored, and if it is monitored that the target to be determined changes from a stationary state or an uncertain state to a moving state, the second state determination operation is terminated.

13. The method according to claim 1, wherein The method further includes: reducing the weight of or eliminating the point cloud data whose point cloud position changes exceed a change threshold.

14. A device for determining a target state, comprising an acquisition module, a registration module, and a prediction module; The acquisition module is used to obtain the point cloud data of the target to be determined at the current moment; The registration module is used to determine the registration time before the current time according to a preset point cloud registration algorithm; obtain the point cloud data at the registration time; use the preset point cloud registration algorithm to perform registration processing on the point cloud data at the current time and the point cloud data at the registration time, and obtain the velocity distribution of the target to be determined at the current time according to the registration processing result; wherein, The registration time is determined according to the registration accuracy of the point cloud registration algorithm, the minimum instantaneous speed of the target to be determined, and the collection time interval of the point cloud data when performing cross-time registration; The prediction module is used to predict the point cloud position change of the target to be determined at the next moment according to the speed distribution of the target to be determined.

15. An electronic device comprising: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform an operation corresponding to the method for determining a target state according to any one of claims 1 to 13.

16. A computer storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method for determining a target state according to any one of claims 1 to 13 is implemented.

Images